For the last decade the Monte Carlo calculation method has been proven to be an excellent tool for accurately simulating electron-solid interactions in atmospheric individual particles of micrometer size. Although it was designed for application to scanning electron microscopy, in the present study it is demonstrated that the Monte Carlo calculation can also be applied in a quantitative single particle analysis using transmission electron microscopy (TEM) with an ultrathin window energy-dispersive X-ray (EDX) spectrometer with a high accelerating voltage (200 kV). By utilizing an iterative reverse Monte Carlo simulation combined with successive approximation, atomic elemental concentrations (including low-Z elements) of submicrometer standard particles were determined with high accuracy for electron beam refractory particles such as NaCl, KCl, SiO2, Fe2O3, Na2SO4, K2SO4, CaCO3, and CaSO4. On the basis of quantitative X-ray analysis together with morphological information from TEM images, overall 1638 submicrometer individual particles from 10 sets of aerosol samples collected in Incheon, Korea, were identified. The most frequently encountered particle types are carbonaceous and (NH4)2SO4/NH4HSO4-containing particles, followed by mineral (e.g., aluminosilicate, SiO2, CaCO3), sea salt, K-rich (e.g., K2SO4 and KCl), Fe-rich, fly ash, and transition or heavy-metal-containing (e.g., ZnSO4, ZnCl2, PbSO4) particles. The relative abundances of the submicrometer particle types vary among samples collected in different seasons and also depend on different air mass transport routes. This study demonstrates that the quantitative TEM-EDX individual particle analysis is a useful and reliable technique in characterizing urban submicrometer aerosol particles.
 Single particle analysis can provide detailed information on an aerosol's microphysical and chemical properties and can deepen our understanding of the sources, reactivity, transport, and removal of aerosol particles [Choël et al., 2005; Laskin et al., 2006]. Over the past decade the use of energy-dispersive X-ray spectroscopy (EDX) coupled with scanning electron microscopy (SEM) and transmission electron microscopy (TEM) has widely been applied for the analysis of both homogeneous and heterogeneous single aerosol particles [Tsuji et al., 2008]. In particular, quantitative energy-dispersive electron probe X-ray microanalysis (ED-EPMA) method, called low-Z particle EPMA, based on SEM equipped with ultrathin-window EDX detector (SEM-EDX), has been shown to reliably provide quantification results for elemental compositional analysis by using a single-scattering Monte Carlo calculation procedure, which can simulate the generation of characteristic and Bremsstrahlung X-rays under low-energy electron beam conditions (usually less than 30 kV) for individual particles on collecting substrates [Ro, 2006]. Not only can the low-Z particle EPMA simultaneously detect the morphology and constituent elements of a single particle, but it also provides quantitative information on many environmental microscopic particles containing nitrates, sulfates, oxides, or mixtures including a carbon matrix [Ro et al., 2005]. However, owing to its limited spatial resolution, SEM measurements are not practically applicable to the characterization of submicrometer particles [Utsunomiya and Ewing, 2003]. Nanometer-sized structures and features of aerosol particles are often characterized by the use of analytical TEM equipped with a variety of analytical tools such as bright field and dark field imaging, selected area electron diffraction (SAED), electron energy loss spectroscopy (EELS), and EDX [Chen et al., 2006; Pósfai et al., 2004]. For instance, Chen et al.  examined submicrometer particulate matter in the urban atmosphere of Lexington, Kentucky, and identified carbonaceous particles (e.g., soot aggregates, char, organic type, and biogenic particles), sulfur-bearing particles (e.g., ammonium sulfate, and crystalline sodium, potassium, and calcium sulfates), silicon-bearing particles, and considerable amounts of iron oxides with other transition metal elements (e.g., Mn and Zn). Among the analytical techniques available in analytical TEM, TEM-EDX is the most convenient tool that has been widely applied to characterize the size, morphology, behavior under electron beams and elemental compositions of ambient fine and ultrafine individual particles [Pósfai et al., 1995]. In TEM-EDX analysis the use of a high accelerating voltage (between 100 and 400 kV) for a very thin specimen minimizes electron-sample interactions that result in limited beam scatterings and high spatial resolution and allows studies on chemical features of particles of smaller size than in SEM-EDX analysis [Williams and Carter, 1996].
 However, since the characteristic X-ray signals from formvar/carbon-coated TEM grids used as collecting substrate can influence accurate determination of carbon and oxygen contents in airborne single particles, and a high accelerating voltage used in TEM-EDX measurement often results in damage on electron beam-sensitive particles, the majority of results based on TEM-EDX measurement has been qualitative. Even a semiquantitative method for TEM-EDX microanalysis has rarely been utilized for identification of ambient aerosol particles. Therefore, most previous studies have had to combine TEM-EDX detection with other analytical tools such as SAED and EELS, demanding extensive data acquisition and interpretation effort.
 For the last decade, the Monte Carlo calculation method has been proven to be an excellent tool for accurately simulating electron-solid interactions in atmospheric individual particles of micrometer size [Ro et al., 2003; Choël et al., 2005; Kang et al., 2008]. In Monte Carlo calculations, electron trajectories are simulated; generated characteristic and Bremsstrahlung X-rays are calculated for spherical, hemispherical, and hexahedral particles sitting on a flat surface; and the concentrations of chemical elements (including low-Z elements such as carbon, nitrogen, and oxygen) of single particles can be accurately determined. Although it was designed for SEM application, the Monte Carlo calculation is herein applied to a quantitative single particle analysis using TEM with an ultrathin-window EDX with a high accelerating voltage (200 kV).
 The aim of this study is to develop a quantitative TEM-EDX single particle analysis technique, based on a modified Monte Carlo calculation, for measuring elemental atomic concentrations of ambient aerosol particles and for identifying individual aerosol particles collected in Incheon, Korea. By the combined use of quantitative X-ray spectral data and morphological information from TEM images, overall 1638 individual submicrometer particles from 10 samples collected in different seasons were characterized.
2. Materials and Methods
2.1. Preparation of Standard Particle Samples
 Chemical compounds such as chlorides, oxides, carbonates, sulfates, and nitrates, prepared from analytical-grade powders with 99.99% purity (Aldrich Chemical Co., Inc., Milwaukee, Wisconsin), were used to evaluate the concentration calculation method based on Monte Carlo calculations combined with iterative successive approximations. Each of them (∼0.2 g) was dissolved in 10 mL distilled water, and the standard aerosols were generated from their corresponding solutions/suspensions by the use of a homemade nebulizer driven by N2 gas (99.99%). These standard aerosol particles were collected on Ni or Cu TEM grids coated with formvar/carbon films (200 mesh grid; Ted Pella Inc., Redding, California) by using a seven-stage May cascade impactor. The particles on stages 6 and 7 were analyzed, where the aerodynamic diameters of collected particles ranged from 0.5–1.0 to 0.25–0.5 μm at the flow rate of 20 L/min. The overall number of standard particles was 225.
2.2. Aerosol Samples Collected in Incheon, Korea
 Incheon is the third largest city in Korea, densely populated with 2.5 million people in an area of 958 km2. The sampling site was located on the roof of a five-storey building (about 20 m above the ground) at Inha University, Incheon (latitude 37.45°N, longitude 126.73°E; see Figure 1). Samples were collected on 200-mesh Cu TEM grids coated with formvar/carbon films by using the seven-stage May impactor, and particles on stage 6 and 7 samples were analyzed.
 In this work, ten samples collected in different seasons and different air mass transport routes were selected. Sampling times, temperature, and relative humidity during samplings, as well as mass concentrations of CO, SO2, NO2, PM2.5 and PM10 on the sampling days, provided by Incheon Meteorological Administration, are shown in Table 1. On 1 April and 26 May 2007, Asian dust storm events occurred with highly elevated PM10 and PM2.5 levels. Backward air mass trajectories were produced using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT4) model available at the NOAA Air Resources Laboratory's web server (http://www.arl.noaa.gov/ready/hysplit4.html) (Figure 2). The 72 hour backward air mass trajectories at a sampling height (around 40 m above the sea level) show that most of the air masses originated from east and northeast China and passed over the Yellow Sea (e.g., those for the samples collected on 23 December 2006 and 25 January, 1 and 19 April, 26 May, and 22 July 2007), while one originated from Sea of Japan (31 July 2007), and two likely from southeast China (23 and 24 October 2007).
Table 1. Sampling Times and Daily Concentrations of Air Pollutantsa
Local Time at the Start of Sampling (KST)
Daily Concentrations of Air Pollutants
Abbreviations are as follows: RH, relative humidity; T, temperature.
 A 200 kV field emission analytical TEM (JEOL JEM-2100F) equipped with an ultrathin-window Oxford EDX spectrometer was used to investigate the standard and Incheon aerosol particles. The shape and size of individual particles were evaluated from their TEM images. Acquisition of X-ray spectral data was carried out in point analysis mode. The X-ray counting time was 20 live seconds. The beam current was 120 μA, the electron beam diameter was 0.5 nm, the take-off angle was 18.7°, and the magnification used was 10,000–15,000 ×. Characteristic X-ray intensities of chemical elements present in individual submicrometer particles were obtained with a nonlinear least-squares fitting program called AXIL [Vekemans et al., 1994]. The dimensional (equivalent projected diameters of particles) and some instrumental information were used as input parameters in a Monte Carlo simulation [Ro et al., 1999, 2003].
 On the basis of morphological data from TEM images and chemical compositional data from X-ray spectra, submicrometer airborne particles collected in Incheon, Korea, were identified and classified into many types. The following parameters guided the classification. First, elements with an atomic concentration of less than 1.0% were not included in the procedure of chemical speciation because elements at trace levels could not be reliably investigated. Second, a particle was regarded as being composed of just one chemical species if this species accounted for at least 90% atomic fraction, and the particle was regarded as a “pure” one. Third, efforts were made to specify chemical species even for particles internally mixed with two or more species. Since many measured particles are heterogeneous in compositions and the spot size of the beam is very small, chemical compositional variation within one particle was sometimes observed. In this case we identified the main part of a particle and classified the particle according to the compositions of the main part. X-ray spectra of beam-sensitive particles are not representative for their original chemical species, so that the identification of those particles mainly relies on their TEM image data with the aid of their EDX data.
3. Results and Discussion
3.1. Evaluation of Quantification Procedure Based on the Monte Carlo Simulation
 For TEM measurements, formvar/carbon-coated TEM grids have widely been used as a collecting substrate of airborne particles. The formvar/carbon film, consisting of formvar resin for film resilience and a carbon layer for film stability under the electron beam, is composed of C, O, and H and has a density of about 1.2 g/cm3. When the formvar/carbon-coated TEM grids are used, the correction for the interference of C and O X-rays from the formvar/carbon film is critical for quantitative analysis of single particles in TEM-EDX. Monte Carlo calculation has strong merit as a quantification procedure since it can deal with sample systems with complex geometry, such as a single particle sitting on a thin film, which is not feasible for conventional EPMA quantification procedures such as those based on ZAF and ϕ(ρz) corrections. In a Monte Carlo calculation program newly modified for TEM-EDX application, which corrects for the interference of C and O X-rays from the formvar/carbon film and includes modifications of input parameters such as accelerating voltage, beam current, beam size, and take-off angle, electron trajectories for a particle sitting on the thin film under high accelerating voltage can be simulated, and X-rays generated both from the particle and the film can be calculated (Figure 3). Primary electrons, with minimized elastic scatterings, pass through the particles as well as formvar/carbon films owing to their high energy (Figure 3). The simulated X-ray spectra for particles with different chemical compositions and sizes show that the intensities of C and O X-rays are nearly constant for particles without C and O elements such as NaCl and KCl particles, whereas X-ray intensities of chemical elements of the particles (e.g., Na, Cl, and K), as well as the Bremsstrahlung background, increase with increasing particle size (Figures 4a and 4b). For particles containing O and/or C such as SiO2, Fe2O3, K2SO4, and CaCO3, the intensities of O and/or C X-rays increase with increasing particle size, indicating that the contribution is mainly from particles (possibly a small fraction from the formvar/carbon films; see Figures 4c–4f). Thus the C and O X-ray intensities of the particle can be calculated separately from those of the film, and the extent of the contribution of X-rays from the collecting film can be estimated and corrected through the Monte Carlo simulation.
 In the present study, overall, 12 types of standard particles, namely NaCl, KCl, SiO2, Fe2O3, Na2CO3, CaCO3, K2SO4, Na2SO4, CaSO4, NaNO3, Ca(NO3)2·4H2O, and (NH4)2SO4, prepared from analytical-grade powders, were analyzed to evaluate the validity of the quantification procedure based on the modified Monte Carlo calculation combined with successive approximation. A detailed description of the algorithm that iteratively determines the elemental concentrations from measured X-ray is given elsewhere [Ro et al., 2003]. The accuracy of the quantification procedure, including the correction for C and O X-ray intensities of the formvar/carbon film, was evaluated by comparing calculated elemental atomic concentrations with the nominal concentrations of particles with known chemical compositions. The relative differences (denoted as Δ) between the nominal (Cn) and calculated (Cc) concentrations for various standard particles, calculated as Δ = 100%*∣Cc-Cn∣/Cc, are shown in Table 2. As illustrated in Table 2, the relative differences for electron beam refractory particles such as NaCl, KCl, SiO2, Fe2O3, Na2CO3, CaCO3, Na2SO4, K2SO4, and CaSO4 are within 18% with the smallest 4.4% (ΔK in KCl) and the biggest 17.4% (ΔC in CaCO3), which is sufficiently accurate for the reliable identification of chemical species of these particles [Niemi et al., 2006], indicating the validity of the quantification procedure. However, for electron beam-sensitive particles such as NaNO3 and Ca(NO3)2·4H2O, the deviations are quite large; for example, ΔN and ΔNa for NaNO3 and ΔCa for Ca(NO3)2·4H2O particles are >30%. For (NH4)2SO4 particles the situation is worse; that is, ΔN, ΔS, and ΔO are 71.3%, 67.9%, and 47.5%, respectively. These deviations are due to electron beam damage on the particles, as electron beams of high energy and current are employed in TEM-EDX measurements. Figure 5 shows typical TEM images of some standard particles (after X-ray spectral data acquisition) and typical X-ray spectra of KCl, Na2CO3, NaNO3, and (NH4)2SO4 particles. The electron beam refractory particles such as KCl and Na2CO3 appear dark, irregular, and crystalline on their TEM images, and the elemental atomic concentrations obtained from their X-ray spectral data are close to their stoichiometries. However, beam-sensitive NaNO3 and (NH4)2SO4 particles show white spots on their TEM image, resulting from damages by high-energy electrons. For the NaNO3 and (NH4)2SO4 particles the X-ray intensities of N and O elements and the obtained elemental concentrations are highly suppressed, indicating the release of NOx and/or NH3 gases with the electron impact on the particles. For the beam-sensitive particles the combined use of TEM images with X-ray spectral data is helpful for their identification even though the quantification procedure cannot provide reliable elemental concentration results.
Table 2. Relative Differences Between Nominal and Calculated Elemental Atomic Concentrations for Various Standard Particlesa
Abbreviations are as follows: Cc, calculated elemental atomic concentration; Cn, nominal elemental atomic concentration; Δ, relative difference between Cc and Cn (Δ = 100%*∣Cc-Cn∣/Cc).
Mean ± standard deviation.
Beam Refractory Particles
Na: 50 Cl: 50
Na: 54.3±1.8 Cl: 45.7±2.0
ΔNa = 7.9 ΔCl = 9.4
K: 50 Cl:50
K: 52.3±1.2 Cl: 47.6±1.1
ΔK = 4.4 ΔCl = 5.1
Si: 33.3 O: 66.7
Si: 37.2±2.4 O: 61.8±2.4
ΔSi = 10.5 ΔCl = 7.9
Fe: 40 O: 60
Fe: 37.7±6.1 O: 57.2±5.2
ΔFe = 6.1 ΔO = 4.9
Na: 33.3 C: 16.7 O: 50
Na: 36.9±2.9 C: 18.8±4.8 O: 44.4±3.3
ΔNa = 9.6 ΔC = 11.2 ΔO = 12.7
Ca: 20 C: 20 O: 60
Ca: 18.4±5.1 C: 24.2±5.7 O: 57.4±8.4
ΔCa = 8.7 ΔC = 17.4 ΔO = 4.5
Na: 28.6 S: 14.3 O: 57.1
Na: 24.8±2.9 S: 15.0±1.6 O: 60.2±3.1
ΔNa = 15.2 ΔS = 4.8 ΔO = 5.1
K: 28.6 S: 14.3 O: 57.1
K: 33.2±9.4 S: 15.0±4.1 O: 51.8±7.7
ΔK = 13.9 ΔS = 4.7 ΔO = 10.2
Ca: 16.7 S: 16.7 O: 66.6
Ca: 19.1±3.3 S: 18.7±3.3 O: 62.2±7.1
ΔCa = 12.6 ΔS = 10.8 ΔO = 7.1
Na: 20 N: 20 O: 60
Na: 31.7±2.3 N: 15.3±0.9 O: 53.0±1.9
ΔNa = 36.9 ΔN = 30.9 ΔO = 13.1
Ca: 7.7 N: 15.4 O: 76.9
Ca: 11.7±1.2 N: 13.9±2.7 O: 74.4±3.4
ΔCa = 34.2 ΔN = 10.8 ΔO = 3.4
N: 28.6 S: 14.3 O: 57.1
N: 16.7±16.1 S: 44.6±13.8 O: 38.7±15.2
ΔN = 71.3 ΔS = 67.9 ΔO = 47.5
3.2. Characterization of Submicrometer Aerosol Particles Collected in Incheon, Korea
3.2.1. Major Particle Types
 Urban areas generally undergo high-pollution loads, especially from heavy traffic, emissions from industrial plants, and intensive energy consumption, which leads to complicated compositions of atmospheric aerosol particles. In the present investigation, analyzed particles (overall 735 and 903 particles for samples of stages 6 and 7, respectively) were classified into 11 types on the basis of their chemical composition and morphology. The characteristics of each particle type are described below.
184.108.40.206. Carbonaceous Particles
 Carbonaceous aerosols of anthropogenic or natural origins are extensively present in urban atmosphere [Chen et al., 2005b, 2005c] and are generally differentiated into two types, organic carbon (OC) and elemental carbon (EC) [Ro et al., 2005; Arimoto et al., 2006]. EC particles, also known as carbon-rich particles, are defined in EPMA as containing more than 90 at. % of C and O, in which the concentration of C is much larger (sometimes more than ten times) than that of O [Geng et al., 2009], and have complex morphologies including fractal-like chain structures (e.g., soot aggregates such as particles 2, 12, 16, and 24 in Figures 6a and 6b), separate spherules (e.g., tar balls such as particle 6 in Figure 6a), and irregular-shaped carbons (e.g., chars such as particles 21 and 46 in Figures 6b and 6c). The soot aggregates are easily distinguished owing to their unique morphology on TEM images. They vary in size from submicrometer to several microns and are believed to be formed via a vaporization-condensation mechanism during combustion processes (mostly from diesel burning) [Chen et al., 2005c, 2006]. Once emitted into the air, soot particles are subject to adsorption or condensation of gaseous species, coagulation with other preexisting aerosols, and oxidation, tending to form soot aggregates. The irregular geometry and complex microstructure of soot aggregates may provide active sites for deposition of water and other chemical species such as sulfates [Pósfai et al., 1999; Zhang et al., 2008a]. This deposition can sometimes change its shape so that it becomes difficult to distinguish soot from chars or organic matter. For example, fresh soot particles display mainly chain agglomerates with spherical primary particles of about 15–35 nm [Mathis et al., 2005]. TEM measurements show that when they are exposed to H2SO4 vapor, they are coated or surrounded by smaller droplets of sulfuric acid, followed by a marked change in morphology of the particles: soot aggregates exhibit a considerable restructuring and shrinking to a more compact form [Zhang et al., 2008a].
 Chars, which have compact and irregular-shaped morphology and are occasionally intermixed with minor inorganic species such as K and S, are possibly formed as incomplete combustion remnants of liquid or solid carbonaceous fuel particles that have undergone carbonization (i.e., devolatilization and/or fluidization followed by solidification) during combustion [Chen et al., 2006]. Tar balls, which have high C, N, and O content (Figure 7a), usually exhibit dark, amorphous shapes and are stable even under a high-energy electron beam. Some of them contain a Cl peak besides C, N, and O signals (Figure 7b). Tar balls are observed only in winter samples in the present study and are regarded to be generated by various human activities, especially household wood combustion [Pósfai et al., 2003, 2004]. The tar ball particles were reported to play an important role in regional haze and climate forcing [Niemi et al., 2006].
 Organic carbon (OC) particles have irregular shapes and lighter contrast on TEM images compared with tar balls (particles 14 and 60 in Figure 6). For the identification of an OC particle the sum of C and O (sometimes including N or S) concentrations is more than 90% in atomic fraction, but the concentration difference between C and O is not too large. It has been reported that atmospheric organic material can account for 30–90% of urban total particle mass and that up to 90% of organic aerosol mass in urban areas can be secondary organic aerosol, which is not directly emitted but formed in the atmosphere by the oxidation of gaseous precursors [Takahama et al., 2007]. A substantial fraction of the organic aerosol mass is composed of polymers, mainly resulting from reactions of carbonyls and their hydrates [Kalberer et al., 2004].
 In addition, very few of biogenic particles (such as pollen, spore, algae, bacteria, virus, and plant or insect fragments from soil and water bodies, etc.) were encountered. The biogenic particles belong to a special type of organic particles different from combustion-generated organics, as they had the noncarbon characteristic elements of organisms (such as N, P, S, K, and/or Cl) along with an amorphous microstructure [Chen et al., 2006; Ro et al., 2002].
220.127.116.11. (NH4)2SO4/NH4HSO4-Containing Particles
 Particles containing (NH4)2SO4 or NH4HSO4 have unique morphologies and microstructures because they are extremely sensitive to an electron beam, under which a large portion of the particle is vaporized within a few seconds, and the remaining residue has a “bubbly” appearance (Figures 5 and 6). EDX spectra obtained from such particles provide limited information about their compositions: the peaks of C, O, S, and perhaps N (N signals often fail to be detected owing to vaporization). They were regarded as mixtures of organic (or carbon-rich) matter and ammonia sulfate/bisulfate [Pósfai et al., 1995], with usually more than 95% atomic fractions of (C+N+O+S) and, accordingly, they could be denoted as (C, N, O, S)-containing particles, as illustrated in Figure 6. It is known that ambient sulfate often acts as a sink for ammonia, which largely originates from animal waste, fertilizer application, soil release, and industrial emissions. The most common form is (NH4)2SO4; but if ammonia is scarce, sulfate would remain in more acidic forms such as NH4HSO4 or H2SO4 [Millstein et al., 2008]. The acidic NH4HSO4/H2SO4 particles are more hygroscopic than pure (NH4)2SO4 [Pósfai et al., 1998] and can spread on the collecting substrate (formvar/carbon film) more than neutral species can, as shown in particle 64 (Figure 6d). The more acidic the sulfate particles are, the more rings of halos are formed [Buseck and Pósfai, 1999]. As far as carbon signals in the (NH4)2SO4/NH4HSO4-containing particles are concerned, they may be from either carbon-rich or organic matter. Indeed, many studies show that in urban atmosphere, (NH4)2SO4/NH4HSO4 aerosols are often internally mixed with soot and organic matter [Adachi and Buseck, 2008], making soot/sulfate aggregates common in polluted urban environments [Johnson et al., 2005]. Sulfates internally mixed with organic compounds (e.g., particles 41 and 55 in Figure 6) would leave visible residues on the grid when sublimated with the electron beam under the TEM [Buseck and Pósfai, 1999]. However, the C signals from formvar/carbon films would not be excluded, as shown in (NH4)2SO4 standard particles (Figure 5).
18.104.22.168. Mineral Particles
 Silicon oxide, calcium carbonate/phosphate, and aluminosilicates are main components of mineral particles. They are electron beam refractory species and easily identified by the quantitative TEM-EDX individual particle analysis (Figures 6 and 8). Both crystalline and amorphous SiO2-containing particles contain high concentrations of Si and O, in which their ratio of atomic concentration is [Si]:[O] ≈ 1:2, very close to their stoichiometry (e.g., the elemental concentrations of O and Si are 60.8 and 34.4 at. %, respectively, for particle 7 in Figure 6a). Aluminosilicate (denoted as “AlSi-containing”) particles with strong Al, Si, and O X-ray signals (Figure 8a) have a wide range of Si and Al content ratio because of their various chemical compositions and crystalline structures such as feldspar, mica, kaoline, zeolites, etc. Carbonate (phosphate)-containing mineral dusts are often associated with calcium and magnesium, either of natural origin or originating from street abrasion or construction sites [Jankowski et al., 2008]. CaCO3, CaMg(CO3)2, and Ca3(PO4)2/CaHPO4 are the main forms.
 When the genuine mineral particles react or are mixed with the “secondary acids,” especially nitric and sulfuric acids, which are formed in the atmosphere by oxidation of SO2 and NOx, sulfates and nitrates can be produced [Sullivan et al., 2007; Hwang et al., 2008]. These reactions or mixture products of genuine particles with “secondary acids” are herein termed reacted mineral particles. Whether they contain nitrate or sulfate (or both) depends on the environment and reaction conditions. Ca(NO3)2/CaSO4, “AlSi + (N, S),” or “SiO2 + (N, S)” particles are the most common products (e.g., particles 1, 15, 25, 27, and 68 in Figure 6), where (N, S) notation represents compounds containing either nitrates, sulfates, or both [Geng et al., 2009]. Normally, the reacted mineral particles are identified by the presence of N and/or S X-ray signals, as shown in Figure 8b. In our samples, sulfate-containing reacted mineral particles are more frequently encountered than nitrate-containing ones. It has been reported that ammonium sulfate or sulfate-containing secondary particles tend to be smaller in size (often in submicrometer size range) than ammonium nitrate or nitrate-containing species (often in supermicrometer size range) [Sullivan et al., 2007].
22.214.171.124. Marine Aerosols
 Genuine and reacted marine aerosol particles can be clearly identified by the quantitative TEM-EDX analysis. The genuine sea-salt particles, which do not experience chemical reactions after being freshly emitted into the air by the so-called bubble bursting or sea spray process [de Hoog et al., 2005], are identified by the presence of Na and Cl X-ray peaks in their X-ray spectrum with their atomic concentration ratio of nearly 1:1 (e.g., particles 34, 35, 37, 42, and 43 in Figure 6c). Often, minor C and O X-ray peaks are concomitant, suggesting that organic matter is adsorbed on them [Wise et al., 2007] and/or NaOH shell, an alkaline hygroscopic coating around the NaCl, is present [Laskin et al., 2003]. The organic matter is probably a kind of aliphatic hydrocarbons (fatty acids) or particulate organic compounds that adhere to marine aerosols as coatings [Tervahattu et al., 2002; Li and Shao, 2010].
 When the genuine sea-salt particles react with nitrogen and sulfur oxides species in the atmosphere, the reacted (or aged) particles such as NaNO3 and Na2SO4 are formed, which results in chlorine loss [Pósfai et al., 1995; Gard et al., 1998; Laskin et al., 2003] (e.g., particles 31 and 32 in Figure 6c). Also, it was observed that some reacted sea salts are internally mixed with soil-derived species such as CaCO3, Ca(NO3)2, and CaSO4. They are classified into the group of “reacted sea salts and mixture” (e.g., particle 33 in Figures 6c and 9).
126.96.36.199. K-Rich Particles
 Among K-rich particles, K2SO4-containing species are encountered the most often, KCl and KNO3 much less. They look dark and irregular on TEM images (e.g., 11, 23, 29, 38, 45, 47, and 65 in Figure 6). K2SO4-containing particles are regarded to mostly originate from biomass burning and also to be formed by the reaction of KCl or mineral K2CO3 with H2SO4 in the air [Pósfai et al., 2003; Ro et al., 2005].
188.8.131.52. Fly Ash Particles
 Fly ash is a powdery material made up of tiny glass spheres and composed of mostly SiO2, Al2O3, Fe2O3, and/or CaO. The fly ash particles that have undergone high-temperature combustion can be easily distinguished because of their unique spherical shape. Although tar balls also look circular on their two-dimensional TEM images, the chemical composition of the fly ash particles are quite different from that of tar balls (Figure 10). The fly ash particles, largely generated by coal combustion for power plants, can travel long distance in air, and many organic compounds can be adsorbed on their surface [Chen et al., 2005a; Kim et al., 2008].
184.108.40.206. Fe-Rich Particles
 Fe-rich particles are generally in the form of iron ((oxy)hydr) oxides (e.g., particles 30 and 48 in Figure 6), usually interpreted as goethite, hematite, or magnetite in atmospheric aerosols. They look irregular on TEM images, and the contents of Fe and O are >20 and 30–60 at. %, respectively. Sometimes they are chemically combined with minor amounts of transition metals, typically manganese and zinc [Chen et al., 2006]. Human activities such as mining, steel production, or metallurgical industries might lead to distinctly higher loads of Fe/FeOx in the urban aerosol; air masses from rural regions showed lower Fe loads than those from industrial locations [Flament et al., 2008].
220.127.116.11. Transition and Heavy-Metal-Containing Particles
 Anthropogenic sources of submicrometer metal-containing particles are plentiful in urban area. Emission of particulates rich in transition and heavy metals such as Pb, Zn, Mn, and Cr are mostly related to high-temperature processes; for example, smelting, combustion of municipal waste, burning of fossil fuel, and ore processing [Murphy et al., 2007; Moffet et al., 2008]. In addition, these metal-containing particles can be emitted on the streets or road surface as part of brake dust, road paint, diesel exhaust particles, construction materials, or car catalyst materials [Li et al., 2001; Hopke et al., 1980]. For example, Zn is a frequently encountered transition metal in street dust. The primary source of Zn is thought to be tire dust, which arises from friction between tire treads and asphalt pavement. In our samples most of the metal-containing particles are Zn- or (Zn, Pb)-containing species (e.g., particles 3, 4, 8, and 26 in Figures 6 and 11). The particles containing minor Cr, As, Se, and/or Mn are often classified into either mineral, Fe-rich, or fly ash particle according to the major components in these particles.
3.2.2. Relative Number Abundances of Various Particle Types in Submicrometer Particle Samples
 The relative number abundances of various types of particles encountered in ten seasonal samples are shown in Figure 12. The (NH4)2SO4/NH4HSO4-containing particles, often mixed with OC/EC, are the most abundant species in nearly all the samples (except sample 23 October 2007), and their relative abundances in stage 7 fraction outweigh those in stage 6 fraction (on average 55.4% versus 35.6%), where the size ranges of stages 6 and 7 are 0.5–1.0 and 0.25–0.5 μm, respectively. Our observation of predominant (NH4)2SO4/NH4HSO4-containing particles of submicrometer size range is consistent with other studies. Adachi and Buseck  reported that (NH4)2SO4/NH4HSO4 aerosol particles, a secondary species formed from atmospheric reactions, are one of the major aerosol types in urban atmosphere. They are of anthropogenic origin and exist abundantly in submicrometer size range. On the basis of TEM analysis of individual particles, Thomas and Buseck  reported that the relative abundance of (NH4)2SO4 and H2SO4 droplets was more than 95% in submicrometer aerosols of eastern Arizona. Zhang et al.  reported that in Qingdao, China, many droplets containing (NH4)2SO4 were produced when sulfuric acid (formed by the oxidation of SO2 from anthropogenic sources) was neutralized by a large amount of ammonia (emitted from urban areas due to human activities).
 Carbonaceous particles are also abundant in nearly all the samples. The sum of relative abundances of organic and carbon-rich particles accounts for 5–45%. Both for stage 6 and 7 fractions, carbonaceous particles are more abundant in winter and spring samples than in summer and autumn samples, indicating that during the cold seasons, such particles are emitted more from combustion processes (such as heating). In addition, semivolatile organic compounds' partitioning between gases and particles shifts toward particulate phase in cold season, leading to more carbonaceous particles [He and Balasubramanian, 2010]. Char particles (incomplete combustion remnants) outnumber soot aggregates, which in turn greatly outnumber tar balls (Table 3), especially in the summer and autumn samples.
Table 3. Relative Abundances of Carbonaceous Particles and Zn- or Zn/Pb-Containing Particles in Stage 6 and 7 Fractions
 As shown in Figure 12 and Table 1, the relative abundances of mineral particles are not correlated with daily concentrations of airborne PM10 or PM2.5. When Asian dust storms occurred (1 April and 26 May 2007), the daily concentrations of PM10 and PM2.5 (especially PM10) were highly elevated, and yet the relative abundances of mineral particles are not so high, even much lower than those on 23 and 24 October 2007. It is probably because this study focused on particles of submicrometer range, whereas Asian dust size distribution was centered at several micrometers [Wehner et al., 2004]. For mineral particles the reacted ones are more abundant than the genuine ones in most of samples (except samples 26 May 2007 and 23 October 2007). Owing to the presence of abundant precursors such as SO2 and NOx in urban atmosphere (e.g., in most sampling days, daily concentrations of airborne NO2 are close or above the Korean state air quality standard value), these reacted mineral particles can be formed in the local atmosphere. Also, they can be formed in long-range transport (Figure 2), as the formation of reacted particles is favorable over the Yellow Sea, the Bohai Sea, and the areas around them [Geng et al., 2009]. Similarly, genuine sea-salt particles are frequently encountered only in the stage 7 fraction of the 22 July 2007 sample and the stage 6 fraction of the 23 October 2007 sample, whereas reacted sea-salt particles are significantly encountered in almost all the samples, indicating that the reaction of sea salts with SO2 or NOx is common over Incheon and its surrounding regions [Hwang and Ro, 2006; Geng et al., 2009].
 K-rich aerosol particles are ubiquitous, with relative abundances of 10.5% and 7.2% on average in stage 6 and 7 fractions, respectively, of all the samples. They are considered to be largely present in biomass burning plumes and are traditionally used as a tracer of biomass burning in source apportionment studies [Adachi and Buseck, 2008]. The high abundance of K-rich particles in the 22 July 2007 sample can probably be attributed to biomass burning from northeast China [Zhang et al., 2008b]. Li and Shao [2009, 2010] observed abundant K-rich particles internally mixed with organic matter by TEM and concluded that most of them resulted from agricultural biomass burning from Beijing and Shandong, Anhui, and Henan provinces, China. In addition, it was reported that forest fires in border areas between China, Mongolia, and Russia produced fume covering a huge region downwind in East Asia [Lee et al., 2005; Zhang et al., 2008b]. However, recent studies reveal that K-rich particles can be emitted from other sources in cities, such as meat cooking, refuse incinerators, and particularly, the use of coal [Hildemann et al., 1991; Wang et al., 2007]. This would explain why the winter and spring samples exhibit higher abundances of K-rich particles than the summer and autumn samples, as the heating in the cold seasons in north China mainly relies on coal combustion [Chan and Yao, 2008].
 Although their relative abundances are not high, Fe-rich, fly ash, and transition- and heavy-metal-containing particles are encountered in nearly every sample. The Fe-rich and fly ash particles are likely of anthropogenic origin, as there exist big steel factories in northeast China and the west coast of Korea (the coast is located ∼5 km to the west of the sampling site), and air masses for most of the samples passed over the regions with steel factories. The extensive subway system in the Seoul-Incheon metropolis area, and street dust, may also contribute to the emission of Fe-containing particles [Kang et al., 2008].
 Transition- and heavy-metal-containing particles are frequently encountered in the spring and summer samples. Their relative abundances in the stage 6 fraction outweigh those in the stage 7 fraction (on average, 3.9% versus 1.1%), indicating that they tend to be relatively large in size. This is likely related to their water solubility since the majority of these metal-containing particles on their TEM images appear circular (see Figure 6a), which is similar to how internally mixed particles containing Zn, Pb, and Sn collected in Mexico City and adjacent areas appear in typical TEM images [Adachi and Buseck, 2008]. Moreover, ZnCl2- or ZnSO4-containing particles outnumber (Zn, Pb)Cl2 or (Zn, Pb)SO4 particles (Table 3). The high level of Zn- and Pb-containing particles (perhaps anglesite, PbSO4, gunningite, ZnCl2, and ZnSO4) has been attributed to industrial processes or coal-fired power stations [Gieré et al., 2006]. The urban layout of Incheon is such that the industrial area is concentrated in the west and north to the sampling site (Figure 1). During sampling processes most air masses pass over western and northern Incheon (Figure 2). Some of these metal-containing particles are probably produced from the factories in the industrial area. It is worth noting that municipal and hazardous waste incineration are considered to emit particles containing Pb, Zn, and chloride, which electronic waste contains abundantly [Moffet et al., 2008]. Further works are needed to investigate lifetimes, concentration, and emission sources of these metal-containing particles.
 Using analytical TEM with an ultrathin-window EDX detector, we measured submicrometer particles of 0.5–1.0 and 0.25–0.5 μm size ranges, including analytical-grade standard particles (such as NaCl, KCl, SiO2, Fe2O3, Na2CO3, CaCO3, K2SO4, Na2SO4, CaSO4, NaNO3, Ca(NO3)2·4H2O, and (NH4)2SO4), as well as the ambient aerosol particle samples collected in Incheon, Korea, in different seasons. These particulate matters were characterized by a quantitative EDX analysis with the combined use of TEM image data. Results show the following:
 1. The Monte Carlo simulation, which has been initially developed for SEM applications (low-energy beam interaction with solid samples), proves to be satisfactorily applicable for quantitative TEM-EDX individual particle analysis, in which high accelerating voltage (200 kV) is used. For beam refractory particles the relative differences between the nominal (stoichiometric) and calculated concentrations for each element (Δ = ∣Cc-Cn∣/Cc) were within 4–18%, an accuracy sufficient to deduce the particles' chemical compositions and to classify the particles on the basis of their chemical species. For beam-sensitive particles such as Ca(NO3)2, NaNO3, and (NH4)2SO4, chemical species can still be identified, as they have obvious or unique morphologies; for example, bubbles.
 2. All the urban aerosol particles from ten sets of samples collected in Incheon were classified into eight groups and eleven types. Overall, the most abundant types were (NH4)2SO4/NH4HSO4-containing and carbonaceous particles, followed by K-rich and mineral particles, and then the marine-derived, Fe-rich, fly ash, and transition- and heavy-metal-containing (such as ZnSO4, ZnCl2, PbCl2, PbSO4) particles, suggesting that anthropogenic sources of submicrometer particles in urban areas of Incheon were plentiful. It is concluded that the relative abundances of various particle types were influenced by the dimensions of particles, urban layout, climate (seasons), and sources and transport routes of air masses.
 This work was supported by the Korean Science and Engineering Foundation (KOSEF) grant funded by the Korean government (MOST) (ROA-2007-000-20030-0).