The influence of collecting substrates on the single-particle characterization of real atmospheric aerosols

Analytica Chimica Acta 658 (2010) 120–127

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

The influence of collecting substrates on the single-particle characterization of real atmospheric aerosols Shila Maskey, Marie Choël 1 , Sunni Kang, HeeJin Hwang, HyeKyeong Kim, Chul-Un Ro ∗ Department of Chemistry, Inha University, 253 Yonghyun-dong, Incheon 402-751, Republic of Korea

a r t i c l e

i n f o

Article history: Received 26 August 2009 Received in revised form 13 October 2009 Accepted 3 November 2009 Available online 10 November 2009 Keywords: Substrate effect Single-particle analysis Indoor aerosol Low-Z particle electron probe X-ray microanalysis

a b s t r a c t This work investigated the influence of three different collecting substrate materials, Ag and Al foils and grids for transmission electron microscopy (TEM grid), on the morphological and chemical compositional analysis of individual particles collected at an underground shopping area in Seoul, Korea. The feasibility of using each substrate in a quantitative single-particle analysis was evaluated by comparing particle morphologies, X-ray spectra, and elemental quantification results obtained for the three substrates. The morphologies and the quality of X-ray spectra for crystalline mineral particles were very similar among the three substrates. However, water-soluble, CNO-rich aerosols showed different morphologies among the three substrates, mainly due to the differences in the hygroscopic properties of the substrates. The quality of the X-ray spectra of the CNO-rich particles was optimal when collected on the TEM grid. To reliably assess the characteristic X-rays of the CNO-rich particles collected on the Ag and Al foils, appropriate data analysis had to be applied. Especially, the X-ray spectra of the CNO-rich particles collected on Al foil required a new background subtraction procedure. The overall relative abundances of the chemical species, obtained from the three collecting substrates, were in good agreement with each other and single-particle characterization of the real aerosol sample was feasible on the different substrates. However, the TEM grid substrate was the most appropriate for single-particle analysis of the water-soluble CNO-rich particles as: (i) it retains the original morphology and size of the particles, (ii) it allows high contrast in the backscattered electron image (BSEI) mode, and (iii) it provides a high peak-to-background ratio (P/B) with small and correctable interferences in the X-ray spectra. © 2009 Elsevier B.V. All rights reserved.

1. Introduction A quantitative single-particle analytical technique, low-Z particle electron probe X-ray microanalysis (low-Z particle EPMA), has been successfully applied to characterize various types of individual aerosol particles [1–4]. This technique, which employs a scanning electron microscope (SEM) and an energy-dispersive X-ray (EDX) detector equipped with an ultra-thin window, allows the quantitative determination of the concentration of low-Z elements such as C, N, and O, as well as other chemical elements, in individual particles of micrometer size [5,6]. The quantitative determination of low-Z elements in individual atmospheric particles improves the applicability of single-particles analysis as many environmentally important atmospheric particles contain low-Z elements (e.g., sulfates, nitrates, ammonium, and carbonaceous compounds). In

∗ Corresponding author. Tel.: +82 32 860 7676; fax: +82 32 867 5604. E-mail address: [email protected] (C.-U. Ro). 1 Present address: Univ. Lille 1, LASIR UMR 8516, 59655 Villeneuve d’Ascq, France. 0003-2670/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2009.11.006

low-Z particle EPMA, atmospheric aerosols are collected on a substrate, typically a metallic foil, by using a cascade impactor. The materials of the collecting substrate may alter the morphology (shape and size) and chemical composition of atmospheric particles through their surface interaction with atmospheric aerosols, resulting in a misleading interpretation of the real nature of the aerosol particles [7]. In addition, X-rays from the substrate can interfere with the quantitative determination of chemical elements by overlapping with the characteristic X-rays from atmospheric particles. Especially, the analysis of submicron particles can be seriously hampered by the interference of characteristic and Bremsstrahlung (continuum) background X-rays emitted from the underlying substrate since electron beams used in SEM/EDX probe a volume of a few cubic micrometers. The choice of optimal substrate material has been a critical issue for the successful application of singleparticle analysis. Many researchers have investigated the feasibility of diverse substrate materials, such as Ag and Al foils, B and Be plates, Nucleopore filter, Si wafer, and grids for transmission electron microscopy (TEM grids), for single-particle analysis using SEM/EDX. Before the advent of low-Z particle EPMA, a polycar-

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

bonate membrane or Nucleopore filter had been the standard choice as a collecting substrate for the computer-controlled, automated elemental analysis in the atomic range of Z > 12, mainly due to its perfectly flat surface [8,9]. However, a polycarbonate membrane filter is inadequate for the analysis of low-Z elements because the filter is composed of C and O elements that produce strong interfering C and O X-rays. The conductive coating required to minimize the charging of samples hinders the determination of low-Z elements as well. Since the influence of Bremsstrahlung background from the filter increases significantly with decreasing particle size, the polycarbonate membrane filter is not appropriate for the analysis of submicron particles [10]. In addition to its feasibility for the analysis of low-Z elements, an ideal collecting substrate needs to be suitable for automated particle analysis by SEM/EDX. In order to obtain environmentally meaningful results for the characterization of airborne particles, several thousands of individual particles have to be analyzed in a relatively short period with automatic data acquisition. A beryllium plate was considered an appropriate collecting substrate due to its suitable physical properties such as low backscattering coefficient, low Bremsstrahlung background, and the absence of interfering X-ray peaks [9,11]. The influence of the collecting substrates on the analysis of electron beam-sensitive particles such as ammonium nitrate and sulfate was studied using artificially generated standard particles. The results demonstrated that the high thermal conductivity of the beryllium plate is helpful for reducing the beam damage [9,11]. However, its toxicity, high cost, and reactivities with sulfates [12] and chlorides [13] limit its use as a practical collecting substrate. Boron plate was reported to be a promising candidate substrate because of its minor spectral interference, low Bremsstrahlung background, and non-toxicity [14,15]. In addition, its polished flat surface is suitable for automated X-ray data acquisition. It is not practical for routine use, however, because it is quite expensive and requires the time-consuming preparation of laboratory-made, polished boron substrate. Si wafer has often been utilized as a substrate material, especially for manual data acquisition, owing to its very smooth surface without defects. However, airborne particles are observed with a poor contrast in backscattered electron (BSE) image, and thus automated detection of particles is not feasible on Si wafer. In addition, the large interfering characteristic X-ray peak of Si and high Bremsstrahlung backgrounds make the interpretation of X-ray spectra difficult [9,11,12]. In this study, we investigated three collecting substrates, Ag and Al foils and a TEM grid, which have been widely used in practice. Ag foil has been successfully used for automatic particle measurement and quantitative analysis of low-Z elements [1,16,17]. Al foil was reported to form a protective and resistant oxide layer on its pure surface, which is normally insoluble in water and other common solvents, thereby reducing the probability of chemical interaction between particles and the substrate [12]. The use of thin-film substrates such as TEM grids was reported to be useful to minimize X-ray interferences from the substrate in single-particle analysis [18,19]. All the previous studies evaluating various collecting substrates for their feasibility in single-particle analysis have been performed with laboratory-generated particles and no published work has evaluated the performance of collecting substrates with real atmospheric aerosols. Since atmospheric aerosols comprise a complex mixture of various chemical species with different sources and sizes, the influence of the collecting substrates on the characterization of real atmospheric aerosols is complicated. The objective of this study is to investigate the influence of the three collecting substrates on the morphological and chemical analyses of real atmospheric aerosols.

121

2. Experimental 2.1. Samples To investigate the influence of three collecting substrates, Al and Ag foils and a TEM grid, on the single-particle characterization of real atmospheric aerosol particles, indoor atmospheric aerosols were collected at an underground shopping area nearby a subway station (Dongdaemun station, Seoul, Korea). The underground sampling site is surrounded by a number of shops selling garments, footwear, leather goods, cosmetics, snacks, and fast foods. Sampling was done from 10 a.m. to 1 p.m. on 13 January 2007. Using a three-stage cascade impactor sampler (PM10 Impactor, Dekati Inc.), particles were sampled on Ag foil (Goodfellow Inc., UK, 99.95% purity, 0.025 mm thickness), Al foil (Sigma–Aldrich, USA, 99.8% purity, 0.05 mm thickness), and TEM grids (Ted Pella Inc., USA, Formvar® /Carbon 200 mesh Cu grid, 30–50 nm thickness). The general properties of the three substrates are presented in Table 1, including the supplier, cost, purity, thickness, and roughness. The three collecting substrates were mounted together in parallel on the impactor stage for simultaneous collection of the particles from the indoor environment. The impactor had aerodynamic cut-off diameters of 10, 2.5, and 1 ␮m for stages 1–3, respectively, at a sampling flow of 10 L min−1 . The samples collected on stage 2 of the impactor (2.5–10 ␮m) were analyzed. The collected samples were put in plastic carriers, sealed, and stored in desiccators before EPMA measurements.

2.2. EPMA measurements The low-Z particle EPMA measurements were carried out on a Hitachi S-4300 cold field emission SEM equipped with an Oxford Link SATW ultra-thin window EDX detector. The resolution of the detector was 133 eV for Mn K␣ X-rays. The X-ray spectra were recorded under the control of EMAX software (Horiba). To achieve optimal experimental conditions, such as low background level in X-ray spectra and good sensitivity for low-Z element analysis, a 10 kV accelerating voltage was used. The beam current was 0.5 nA and the measuring time was 10 s. A more detailed discussion on the measurement conditions is given elsewhere [5]. X-ray data acquisition for individual particles was carried out manually in point analysis mode, i.e., the electron beam was focused at the centre of each particle, and X-rays were acquired while the beam remained fixed on this single spot. Morphological parameters, such as diameter and shape factor, were calculated by an image processing routine. These estimated geometrical data were set as input parameters for the quantification procedure.

2.3. Data analysis The net X-ray intensities for chemical elements were obtained by non-linear, least-square fitting of the collected spectra using AXIL program [20]. The elemental concentrations of individual particles were determined from their X-ray intensities by the application of Monte Carlo calculation combined with reverse successive approximations [6,21,22]. The low-Z particle EPMA can provide quantitative information on the chemical composition. Based on the chemical composition, particle morphology, and X-ray spectra, the particles were classified into different groups. The analytical procedure for determining the chemical species, and the assignment of particles to specific types, is described in more detail elsewhere [23,24].

122

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

Table 1 Characteristics of three different substrate materials for single-particle analysis. Substrate property

Silver foil

Aluminum foil

Carbon-coated copper grid

Supplier Cost per surface unit (D cm−2 ) Purity (wt.%) Thickness Roughness

Goodfellow Inc., UK Low (0.17) 99.95 0.025 mm Parallel ridges

Ted Pella Inc., USA Medium (5.68) 99.99 ∼30–60 nm Smooth

Physical resistance

Strong

Sigma–Aldrich, USA Low (0.08) 99.8 0.05 mm Bumpy surface Ridges on the shiny side Strong

Electrical conductivity Sample preparation

Yes Not necessary

Yes Not necessary

Feasible in BSEI mode only (owing to high atomic number of substrate)

Not feasible (due to atomic number of substrate and rough surface)

Very high Ag–L on Cl and K and Ag–M on C (correction factor Ag–L/Ag–M)

High Mg, Al, Si

Very fragile (can be strengthened with a thin layer of Formvar® ) Yes TEM grid sample holder adapted for SEM–EDX analysis is required

Secondary electron image (SEI) of blank substrate

Secondary electron image (SEI) of atmospheric particles

Backscattered electron image (BSEI) of atmospheric particles

Grey level threshold operation for computer-controlled automated data measurement Continuum background Overlapping lines

High contrast for all types of particles in BSEI mode (owing to very low BSE yield of the film) Very low Very small signal of C (from the film) and Cu (from lateral scattering)

3. Results and discussion 3.1. Characteristics of Ag and Al foils and TEM grids as collecting substrates The three collecting substrates were mounted together in parallel on the same impactor stage for simultaneous collection of the indoor aerosols. The feasibility of each substrate in single-particle analysis was evaluated by comparing X-ray spectra, particle morphologies, and elemental quantification results obtained for the three substrates. High quality X-ray spectra need the minimum interference from continuum (Bremsstrahlung) and characteristic X-rays of the substrate. Fig. 1 shows X-ray spectra obtained for the three bare substrates. The X-ray spectrum of the bare TEM grid showed a remarkably low Bremsstrahlung background as most primary electrons passed through the carbon/Formvar® thin-film (30–50 nm thickness) without generating many X-rays from the film with a low BSE yield. The particles on the TEM grids that provided a very dark background are clearly evident with better contrast (see Table 1). Previous studies also reported that the low

Fig. 1. X-ray spectra of bare collecting substrates collected at an accelerating voltage of 10 kV with an acquisition time of 10 s.

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

Bremsstrahlung background of the TEM grid is a good asset for a collecting substrate [8,10,19,25,26]. As the Ag (25 ␮m) and Al (50 ␮m) foils are much thicker than the ultra-thin TEM grid (30–50 nm), they generate high Bremsstrahlung backgrounds due to X-ray emissions from the bulk substrates. These high backgrounds are problematic in the elemental analysis of small particles. For the bare Al foil, the Bremsstrahlung background had a step-up shelf and a stepdown at energies below and above the Al K␣-line, respectively (Fig. 1). The step-up shelf was due to incomplete charge collection phenomenon resulting in a detector-produced background shelf below each characteristic line and the step-down was due to selfabsorption. This anomaly of the continuum background of Al foil rendered the accurate subtraction of the background, which is necessary in the X-ray intensity evaluation, difficult. The three bare substrates also produced their own characteristic X-ray peaks. In the Ag substrate, the X-ray L-lines of Ag interfered with the accurate determination of Cl and K K␣ X-ray intensities and the M-lines of Ag overlapped with the X-ray K␣-line of C. Similarly, when a particle contained an Al element, the determination of the Al content of the particle was impaired by the characteristic X-rays from the Al substrate itself. As shown in Fig. 1, the characteristic X-ray peak intensity of the C K␣-line of the TEM grid (<60 photon counts) was one or two orders of magnitude smaller than that of the Ag L-line of the Ag foil (>900 photon counts) and Al K␣-line of the Al foil (>2000 photon counts). Although the characteristic X-ray intensities of C and O were very low for the bare TEM grid, the accurate determination of C and O was still hindered, especially for small particles. Recently, our group developed a methodology based on a Monte Carlo calculation technique to correct for the interfering peaks of C and O emitted from the TEM grid, which provided reliable quantification results when applied to the quantification of standard (sub)micron particles such as CaCO3 , CaSO4 , Na2 SO4 , and SiO2 [27]. Although a small peak of Cu–L-line is caused by lateral scattering on Cu bars of the TEM grid and the interference of Cu cannot be compensated for, Cu-containing airborne particles are rarely encountered in ambient aerosol samples. However, it is probable to encounter submicron Cu-containing particles in polluted atmospheric environment, in which case the interference from the Cu bars can be problematic. As the BSE signal from the Ag foil was stronger than that from the airborne particles being analyzed, a BSE image (BSEI) was obtained by inverting the original BSE signal and processing it to binary image, which is a common method for easy recognition of individual particles in computer-controlled SEM/EDX measurements (see Table 1). In the case of Al foil, the BSE signal from the Al substrate can be stronger than that from a particle when the mean Z (Zave ) of a particle is equal to or less than 13, and weaker when Zave is greater than 13, resulting in different BSE contrasts according to the chemical compositions of individual particles (Table 1). The BSEI cannot be used for automatic particle recognition and only manual measurement is allowed using secondary electron image (SEI). As the TEM grid produces a very weak BSE signal owing to its ultra-thin-film structure, all types of particle are distinctly recognized on the dark TEM grid substrate in the BSEI mode. The use of the TEM grid substrate facilitates feasible automatic localization of small particles, which is always not possible using the metallic foils. 3.2. Secondary electron images (SEIs) and X-ray spectra of indoor aerosol particles on the different collecting substrates Fig. 2 shows typical SEIs of particles collected on the Ag foil, Al foil, and TEM grid. The chemical species of each particle obtained from the low-Z particle EPMA were designated together with the number of the particles. For the notation of the chemical species of the particles, a unique notation is devised: e.g., SiO2 /C indicates

123

Fig. 2. Typical secondary electron images (SEIs) of individual particles collected on (a) Ag foil, (b) Al foil, and (c) TEM grid substrates.

that the SiO2 and carbonaceous species are >10% in atomic fraction, while (Ca,Mg)(NO3 ,SO4 ) indicates that the particle is composed of an internal mixture of Ca(NO3 )2 , CaSO4 , Mg(NO3 )2 , and MgSO4 . In Fig. 2a and b, two different particle types can be differentiated on the basis of their SEIs: one type has a bright contrast (e.g., particles #1–3 in Fig. 2a and #1, #3, and #5 in Fig. 2b, to name a few), and the other a very dark contrast (e.g., particles #9 and #15 in Fig. 2a and particles #4, #14, #15, and #18 in Fig. 2b). In Fig. 2c, two different particle types are also observed: one looks bright and angular (e.g., particles #1–3 and #5) and the other looks gray and circular (e.g., particles #4, #7, and #16). The different morphologies of the two types represented their different particle phases

124

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

Fig. 3. Typical secondary electron images (SEIs) and X-ray spectra of “core–shell” particles collected on (a) Ag foil, (b) Al foil, and (c) TEM grid substrates.

at the time of their collection on the substrate, i.e., solid particles and water droplets. The solid aerosols were observed with a bright contrast on their SEIs as they are “dense” mineral particles such as aluminosilicates, SiO2 , CaCO3 , and Fe oxides, and thus produced sufficient SE and BSE signals. These particles looked angular as they are primary dispersion aerosols of mostly soil origin. Due to their crystalline nature, the penetration range of primary electrons is low (1–2 ␮m for a 10 kV accelerating voltage for aluminosilicates, SiO2 , CaCO3 , and Fe oxides), resulting in the generation of X-rays mostly from the particles. Indeed, as shown in Supplement Fig. 1, the typical X-ray spectra of aluminosilicates and CaCO3 particles of sizes 1.5–2.8 ␮m showed strong X-ray peaks characteristic of the particles. Although a small amount of Ag L and Al K X-rays from the Ag and Al substrates, respectively, were present in the spectra, the X-ray spectral quality was sufficiently high to obtain reliable quantitative elemental concentrations after applying the procedures to correct for the interferences that are discussed later. As the Bremsstrahlung background X-rays were also mostly generated from solid mineral particles themselves, the peak-to-background (P/B) ratios of the spectra from all three substrates were similar, although slightly better from the TEM grid. The darkly shaded particles on the Al and Ag foils were collected as water droplets at the time of sampling, and then spread out on the hydrophilic metallic foils [3]. The hydrophilicity of a solid surface can be assessed by the contact angle at which the water interface meets the solid surface; i.e., the surfaces with contact angle larger or smaller than 90◦ are hydrophobic or hydrophilic, respectively [28]. The contact angles of deionized water are 85◦ on

the Ag electroplated surface [29] and 87 ± 2◦ on the Al surface [30], which indicates that both substrates are hydrophilic but the Ag surface is more hydrophilic than the Al one. The lower hydrophilicity of the Al foil in comparison to that of the Ag foil caused the compacter spread of the particles on the Al foil (Fig. 2a and b and Supplement Fig. 2). As the TEM grid is hydrophobic with a contact angle of the carbon film of 150◦ [31], this type of particle morphology was not observed in the SEIs of the TEM grid (Fig. 2c and Supplement Fig. 2). Instead, the water-soluble aerosols appeared gray in contrast and were circular. The presence of small holes or cracks observed in the water-soluble aerosols on the TEM grid (Fig. 2c and Supplement Fig. 2c) indicated that the water rapidly evaporated when the sample was placed in the vacuum of the SEM observation chamber. The numerous dark and small circular spots (<0.5 ␮m) observed on the surface of the TEM grid (Fig. 2c) were attributed to the artifact of the TEM grid substrate. A total of 715 particles (approximately 240 for each substrate sample) were analyzed manually, and a significant fraction of the samples consisted of water-soluble aerosols: 21.4%, 22.8%, and 25.5% for the Ag foil, Al foil, and TEM grid samples, respectively. Most of them were rich in C, N, and O. As the low-Z particle EPMA technique cannot detect hydrogen, it was difficult to clearly identify the chemical species of these water-soluble, CNO-rich particles; they were likely to be secondary organic particles internally mixed with NH4 NO3 species. As many CNO-rich particles contain minor S, Mg, and Cl, additional SO4 2− and Cl− combined with Mg2+ and NH4 + are present in them. Supplement Fig. 2 shows typical SEIs and X-ray spectra of the water-soluble, CNO-rich particles collected on

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

the different substrates. The dark shade of the particles on the Ag and Al foils was due to the lower SE and BSE yields of the particles with major C, N, and O elements than those of the substrates. The BSE coefficient (), defined as the ratio of the BSE current passing out of the specimen to the primary electron beam current injected into the specimen, increased with increasing specimen Z, thereby forming the basis for Z contrast (compositional contrast) in SEI; i.e.,  = 0.06, 0.16, and 0.4 for C, Al, and Ag, respectively [32]. The SE coefficient (ı), defined as the ratio of the SE current passing out of the specimen to the primary electron beam current injected into the specimen, increased as the Z increased up to Z = 15 but then stayed relatively constant at Z > 15; i.e., ı = 0.05, 0.1, and 0.12 for C, Al, and Ag, respectively [32]. The CNO-rich particles appeared gray on the SEIs for the TEM grid, due to the very low SE and BSE yields of the TEM grid (with 90% C content). As the water-soluble, CNO-rich particles spread out and were flattened on the Ag and Al foils, most primary electrons penetrated through the particles and X-rays from the Ag and Al substrates became dominant; i.e., the Ag L and Al K X-ray peaks were the strongest in the spectra of Supplement Fig. 2a and b, respectively, and the Bremsstrahlung backgrounds from the substrates are evident. The CNO-rich particles on the TEM grid maintained their original spherical morphology and the influence of the TEM grid substrate on the X-ray spectrum of the particle was minimal (Supplement Fig. 2c). The X-ray spectrum of the watersoluble, CNO-rich particle on the TEM grid showed the highest P/B value, which is a critical finding for accurate determination of elemental concentrations. Some particles were often observed to be internally mixed with solid species in the core region of the particles and with the water moiety in the surface region. These particles were formed either by the coagulation of water-soluble aerosols and solid mineral particles or by the growth of the water-soluble moiety on the surface of hygroscopic inorganic mineral particles. As shown in Fig. 3a and b, the water moiety (dark shade on SEIs) spread out on the Ag and Al foils when this “core–shell” particle type was collected on the foils, and the solid mineral moiety (bright core) was clearly distinguishable. Indeed, the X-ray spectra obtained from the darkly shaded particles in Fig. 3a and b indicated that they were from water-soluble, CNO-rich moieties, whereas those obtained from the bright cores indicated that they were from CaCO3 /SiO2 and Fe oxides. The core–shell particle type collected on the TEM grid did not show any clear morphological separation between the two regions, as shown in Fig. 3c. However, a careful interpretation of its X-ray spectrum revealed its core–shell structure. The Al peak in Xray spectra (Fig. 3c and Supplement Fig. 2c) was from an aluminum stub on which TEM grid sits during SEM/EDX measurement. However, after the development of a homemade sample holder for TEM grids to minimize the Al interference, the artifact of Al peak is no longer a problem [27]. Similarly, Si peak in X-ray spectra (Fig. 3c

125

Fig. 4. Number distributions of particles collected on the three different substrates according to their sizes.

and Supplement Fig. 2c) was attributed to the artifact of TEM grid substrate as shown in Fig. 1. As the water droplet particles spread out on the hydrophilic surfaces and their original morphology can no longer be maintained, the sizes of water-soluble particles are always overestimated when they are collected on Ag and Al foils. The number distributions of the indoor aerosols collected on the three substrates according to their size are shown in Fig. 4. Since most of particles were irregularly shaped, the sizes of individual particles were calculated in terms of equivalent projected diameters from their SEIs. The equivalent diameter of a particle on SEI is calculated as that of a circular particle with the same area. The number distribution for the TEM grid had two maxima at diameters of 2 and 4 ␮m, whereas compared to four (at 2, 4, 6, and 9 ␮m) and three (at 2, 4, and 6 ␮m) maxima for the Ag and Al foils, respectively. These additional maxima observed in the number distributions for the Ag and Al foils were attributed to the size overestimation of the water-soluble particles. As the Ag foil is more hydrophilic than the Al foil, the water-soluble particles spread out more on the Ag foil. Thus, the number distribution for the Ag foil showed more maxima than that for the Al foil. 3.3. Data analysis of X-ray spectra obtained from the different collecting substrates For a decade, the low-Z particle EPMA technique has been successfully applied for the characterization of various types of airborne particle sample [33,34]. Ag foil has usually been used as the collecting substrate since it allows automatic recognition of particles and X-ray measurements using an image processing routine [1,17,35]. Recently, however, Al foil has been used

Fig. 5. X-ray spectra of a CNO-rich particle on the Al foil (a) before and (b) after applying background subtraction.

126

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

as a collecting substrate in our group because manual measurements provide detailed morphological information, which cannot be obtained using automatic X-ray data measurements [36,37]. As previously stated, characteristic X-rays from Ag and Al foils can overlap with X-rays from particles, e.g., Ag M-lines with C Kline, Ag L-lines with K and Cl K-lines, and substrate Al K-line with particle Al K-line. Herein, we briefly summarize the correction procedures for these interferences. The Ag M X-ray peak for pure Ag foil was fitted using the AXIL curve fitting program [20] and the ratio of intensities between the Ag L␣ and M X-ray lines for pure Ag foils was obtained. In the calculation of the C content of the particles, the apparent C K␣ intensity, if any, was thus corrected for the influence of the Ag M-line by applying the intensity ratio between the Ag L␣ and M X-ray lines for pure Ag foils, although this first approximation will always overcorrect the influence of the Ag substrate on the intensity of the C X-ray peak. The K and Cl K X-ray lines completely overlap with the Ag L-lines. The overlaps of the Ag L␤ and L␥ X-rays on the K and Cl K-lines, respectively, were corrected for by using the intensity ratios of Ag L␤ /L␣ and L␥ /L␣ , respectively, for pure Ag foils. This correction procedure is described in more detail elsewhere [35]. Airborne Al2 O3 particles are very rare and substrate Al interference can be a problem only for aluminosilicate particles. To correct for the substrate Al interference on the aluminosilicate particles, the intensity ratio of Al/Si, which is 0.55 ± 0.06, was obtained from about 2000 aluminosilicate particles collected on Ag foils in our previous studies [3,38]. Consequently, for particles collected on Al substrate, the part of Al intensity greater than “Si intensity × 0.55” was attributed to Al foil. The Bremsstrahlung background of the Al substrate is dominant in the X-ray spectra of the water-soluble, CNO-rich particles (Fig. 3b and Supplement Fig. 2b). Indeed, the AXIL curve fitting program failed to correctly fit this anomalous background by any means. Since the Bremsstrahlung background in the X-ray spectra of the CNO-rich particles was almost exclusively from the Al foil (see Supplement Fig. 3), the background in the CNO-rich particle spectra was subtracted using the spectrum of the bare Al foil as the first approximation. Fig. 5 shows the X-ray spectra of a CNO-rich particle on the Al foil before and after applying this background subtraction procedure. As shown in Fig. 5b, the background subtraction worked well by successfully removing the anomalous background from the Al foil. Furthermore, the new spectrum could be effectively fitted using the AXIL program. This background subtraction method was applied before the elemental concentrations of the CNO-rich particles collected on the Al foils were determined using the quantification procedure based on the Monte Carlo calculation. A total of 715 particles, approximately 240 for each substrate sample, were analyzed manually. Based on the morphology and chemical composition of the particles collected on the three substrates, they were classified into 11 different chemical types. The basic classification rules are summarized briefly. First, a particle was regarded as being composed of just one chemical species if this species constituted at least 90% in atomic fraction. Second, the chemical species were specified even for particles comprising an internal mixture of two or more chemical species. Third, elements with less than 1.0 at.% were ignored since EPMA has high detection limits and the elements at trace levels could not be reliably investigated [23]. The characteristics, in terms of sources, reactivities, seasonal variations, and chemical compositions, of the indoor aerosol samples, as well as the classification results, will be discussed in detail in a separate article. Herein, our focus is on the degree of similarity between the classification results obtained from the three collecting substrates as these substrates were mounted parallel in the sampler and the final classification results have to be similar. As shown in Fig. 6, the relative abundances of the eleven chemical groups were very similar to each

Fig. 6. Overall relative abundances of chemical species encountered at significant levels in samples collected on the three different substrates. Particles that could not be assigned to any of the ten different chemical groups due to poorly interpretable X-ray spectra were classified into the NEIC group (i.e., not enough information contained in their spectra).

other, demonstrating that the single-particle characterization of real airborne particles can be reliably performed using either Ag foil, Al foil, or TEM grid as the collecting substrate. However, as discussed above, the appropriate data analysis for each substrate should be applied to ensure the reliability. Of the particles collected on the Ag and Al foils, 4.8% and 0.9%, respectively, were not assigned to any particle type due to insufficient information contained in their X-ray spectra (the “NEIC” group, i.e., not enough information contained in their spectra, in Fig. 6), whereas all the particles collected on the TEM grid were clearly identified owing to high P/B ratios of the spectra.

4. Conclusions This work investigated the influence of three different collecting substrate materials, Ag and Al foils and TEM grid, on the singleparticle characterization of a real atmospheric aerosol sample by the application of the low-Z particle EPMA technique. The morphologies and the quality of X-ray spectra for crystalline mineral particles were very similar among the three substrates. However, water-soluble, CNO-rich aerosols showed different morphologies among the three substrates, mainly due to the differences in the substrates’ hygroscopic properties. The quality of the X-ray spectra of the CNO-rich particles was optimal when collected on the TEM grid. To reliably assess the characteristic X-rays of the CNO-rich particles collected on the Ag and Al foils, appropriate data analysis was necessary. Especially, the X-ray spectra of the CNO-rich particles collected on the Al foil required a new background subtraction procedure. The overall relative abundances of the chemical species, obtained from the three collecting substrates, were in good agreement with each other and the single-particle characterization of the real aerosol sample was feasible on the different substrates. However, the TEM grid substrate was the most appropriate for single-particle analysis of the water-soluble CNO-rich particles as: (i) it retains the original morphology and size of the particles, (ii) it allows high contrast in the backscattered electron image (BSEI) mode, and (iii) it provides a high peak-to-background ratio (P/B) with small and correctable interferences in the X-ray spectra.

S. Maskey et al. / Analytica Chimica Acta 658 (2010) 120–127

Acknowledgements This work was supported by a grant from the Korea Science and Engineering Foundation (KOSEF, grant no. R0A-2007-000-20030-0) and funded by the Korea government (MOST). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.aca.2009.11.006. References [1] S. Kang, H. Hwang, Y. Park, H. Kim, C.-U. Ro, Environ. Sci. Technol. 42 (2008) 9051–9057. [2] H. Hwang, H. Kim, C.-U. Ro, Atmos. Environ. 42 (2008) 8738–8746. [3] H. Hwang, C.-U. Ro, Atmos. Environ. 40 (2006) 3869–3880. [4] H. Kim, H. Hwang, C.-U. Ro, Spectrochim. Acta B61 (2006) 393–399. [5] C.-U. Ro, J. Osán, R. Van Grieken, Anal. Chem. 71 (1999) 1521–1528. [6] J. Osán, I. Szalóki, C.-U. Ro, R. Van Grieken, Mikrochim. Acta 132 (2000) 349–355. [7] K. Wittmaack, Atmos. Environ. 36 (2002) 3963–3971. [8] R.H.M. Godoi, S. Potgieter-Vermaak, J. de Hoog, R. Kaegi, R. Van Grieken, Spectrochim. Acta B61 (2006) 375–388. [9] I. Szalóki, J. Osán, A. Worobiec, J. de Hoog, R. Van Grieken, X-ray Spectrom. 30 (2001) 143–155. [10] J.C. Seaman, Environ. Sci. Technol. 34 (2000) 187–191. [11] A. Worobiec, J. de Hoog, J. Osán, I. Szalóki, C.-U. Ro, R. Van Grieken, Spectrochim. Acta B58 (2003) 479–496. [12] P.-F. Huang, B. Turpin, Atmos. Environ. 30 (1996) 4137–4148. [13] J.C. Birkbeck, N.L. Kuehler, D.L. Williams, W.E. Moddeman, Surf. Interface Anal. 27 (1999) 273–282. [14] M. Choël, K. Deboubt, J. Osán, P. Flament, R. Van Grieken, Anal. Chem. 77 (2005) 5686–5692. [15] M. Choël, K. Deboubt, P. Flament, Microsc. Res. Technol. 70 (2007) 996–1002. [16] H. Hwang, C.-U. Ro, J. Geophys. Res. 110 (2005) D23201.

127

[17] C.-U. Ro, K.-Y. Oh, H. Kim, Y.P. Kim, C.B. Lee, K.H. Kim, C.H. Kang, J. Osán, J. de Hoog, A. Worobiec, R. Van Grieken, Environ. Sci. Technol. 35 (2001) 4487–4494. [18] C.L. Gregory, H.A. Nullens, R.H. Gijbels, P.J. Van Espen, Anal. Chem. 70 (1998) 2551–2559. [19] A. Laskin, J.P. Cowin, Anal. Chem. 73 (2001) 1023–1029. [20] B. Vekemans, K. Janssens, L. Vincze, F. Adams, P. Van Espen, X-ray Spectrom. 23 (1994) 278–285. [21] I. Szalóki, J. Osán, C.-U. Ro, R. Van Grieken, Spectrochim. Acta B55 (2000) 1017–1030. [22] C.-U. Ro, J. Osán, I. Szalóki, J. de Hoog, A. Worobiec, R. Van Grieken, Anal. Chem. 75 (2003) 851–859. [23] C.-U. Ro, J. Osán, I. Szalóki, K.-Y. Oh, H. Kim, R. Van Grieken, Environ. Sci. Technol. 34 (2000) 3023–3030. [24] C.-U. Ro, H. Kim, R. Van Grieken, Anal. Chem. 76 (2004) 1322–1327. [25] A. Laskin, J.P. Cowin, M.J. Iedema, J. Electron Spectrosc. Relat. Phenom. 150 (2006) 260–274. [26] A. Laskin, M.J. Iedema, J.P. Cowin, Aerosol Sci. Technol. 37 (2003) 246–260. [27] M. Choël, H. Hwang, C.-U. Ro, European Conference on X-ray Spectrometry, Cavtat, Dubrovnik, Croatia, 16–20 June, 2008. [28] J.H. Lin, W.H. Chen, S.H. Su, J. Power Sources 184 (2008) 38–43. [29] M.N. Kozicki, P. Maroufkhani, M. Mitkova, Superlattices Microstruct. 34 (2003) 467–473. [30] X. Sheng, Y.P. Ting, S.O. Pehkonwn, Colloid Interface Sci. 310 (2007) 661–669. [31] X. Xiao, Y.T. Cheng, J. Mater. Res. 23 (2008) 2174–2178. [32] J. Goldstein, D. Newbury, D. Joy, C. Lyman, P. Echlin, E. Lifshin, L. Sawyer, J. Michael, Scanning Electron Microscopy and X-ray Microanalysis, third ed., Kluwer Academic/Plenum Publishers, New York, 2003. [33] C.-U. Ro, H. Kim, K.Y. Oh, S.K. Yea, C.B. Lee, M. Jang, R. Van Grieken, Environ. Sci. Technol. 36 (2002), pp. 4770–4476. [34] A. Worobiec, I. Szalóki, J. Osán, W. Maenhaut, E.A. Stefaniak, R. Van Grieken, Atmos. Environ. 41 (2007) 9217–9230. [35] C.-U. Ro, K.-Y. Oh, H. Kim, Y. Chun, J. Osán, J. de Hoog, R. Van Grieken, Atmos. Environ. 35 (2001) 4995–5005. [36] H. Geng, H. Jung, Y. Park, H. Hwang, H. Kim, Y. Kim, Y. Sunwoo, C.-U. Ro, Atmos. Environ. 43 (2009) 3364–3373. [37] S.J. Kang, H. Hwang, S. Kang, Y. Park, H. Kim, C.-U. Ro, Atmos. Environ. 43 (2009) 3445–3453. [38] C.-U. Ro, H. Hwang, H. Kim, Y. Chun, R. Van Grieken, Environ. Sci. Technol. 39 (2005) 1409–1419.