Heterogeneity Assessment in Individual CaCO3−CaSO4 Particles

Anal. Chem. 2001, 73, 4574-4583

Heterogeneity Assessment in Individual CaCO3-CaSO4 Particles Using Ultrathin Window Electron Probe X-ray Microanalysis Chul-Un Ro,*,† Keun-Young Oh,† Ja´nos Osa´n,‡ Johan de Hoog,§ Anna Worobiec,§ and Rene´ Van Grieken§

Department of Chemistry, Hallym University, ChunCheon, KangWonDo, 200-702, Korea, KFKI Atomic Energy Research Institute, P.O. Box 49, H-1525 Budapest, Hungary, and Department of Chemistry, University of Antwerp (UIA), Universiteitsplein 1, B-2610 Antwerpen, Belgium

In our previous studies, it has been demonstrated that both the excitation interactions between electrons and the atoms of the matrix and the matrix and geometric effects of electron-induced X-ray signals can be described by Monte Carlo simulation for low-Z elements, such as carbon, nitrogen, and oxygen, in individual atmospheric microparticles. In addition, by the application of a quantification method, which employs Monte Carlo simulation combined with successive approximations, at least semiquantitative specification of the chemical compositions could be done. This has enlarged the scope of electron probe X-ray microanalysis (EPMA) for the single particle analysis of atmospheric environmental aerosol particles. In this work, we demonstrate that the heterogeneity of individual particles, even of micrometer size, can be characterized by the application of EPMA. X-ray photons obtained with different primary electron beam energies carry information on the chemical compositions for different regions in the particles. Artificially generated heterogeneous CaCO3-CaSO4 individual particles were measured at different accelerating voltages, and it was found that the Monte Carlo calculation is a powerful technique to extract the information on the heterogeneity of the particles that is contained in the measured X-ray data. Our approach can even estimate the thickness of the surface CaSO4 species by the application of the Monte Carlo calculation. A preliminary result for carbon-coated glass particles is also presented. The complexity involved in the analysis of real world particles is briefly mentioned with a result for heterogeneous SiO2 particle. Characterization of airborne particles deepens our understanding of the source, reaction, transport, and removal of atmospheric chemical species. Since atmospheric particles are chemically and morphologically heterogeneous, and the average composition and aerodynamic diameter do not describe well the population of the particles, microanalytical methods have proven to be useful for * Corresponding author. Phone: +82 33 240 1428. Fax: +82 33 256 3421. E-mail: [email protected]. † Hallym University. ‡ KFKI Atomic Energy Research Institute. § University of Antwerp.

4574 Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

studying atmospheric particles. Electron probe X-ray microanalysis (EPMA) is capable of simultaneously detecting the chemical composition and morphology of a microscopic volume, such as a single atmospheric particle.1 In our previous study2, it was found that excitation interactions between electrons and the matrix atoms and the geometric and matrix effects on electron-induced X-ray signals for light elements in individual atmospheric microparticles can be described by Monte Carlo simulation. By the application of a quantification method, which employs Monte Carlo simulation combined with successive approximations, it was shown that at least semiquantitative specification of the chemical compositions can be performed.3 Furthermore, the chemical species, in addition to chemical composition, in individual urban particles can be determined by the application of ultrathin window EPMA.4 This recently developed EPMA technique specially allows the determination of the concentration of low-Z elements, such as carbon, nitrogen, and oxygen, as well as the elements that are observed using conventional energy-dispersive EPMA. Conventional energy-dispersive X-ray (EDX) detectors are not suitable for low-Z element analysis, mainly because their Be window, used for protecting the semiconductor detector surface from contamination, absorbs low-energy X-rays and, thus, hinders the detection of the low-Z element X-rays. The determination of low-Z elements in individual environmental particles allows the improvement of the applicability of the single particle analysis; many environmentally important atmospheric particles, for example, sulfates, nitrates, ammonium, and carbonaceous particles, contain low-Z elements, which have not been characterized using the conventional EPMA. However, it is of prime importance to have an analytical tool to distinguish chemical species in the surface region from that of the core region in individual particles of micrometer size, because the analysis would allow investigating directly and more conclusively the nature of atmospheric reactions that some airborne particles may experience. For example, sea salt can react with (1) Jambers, W.; De Bock, L.; Van Grieken, R. Analyst 1995, 120, 681-692. (2) Ro, C.-U.; Osa´n, J.; Van Grieken, R. Anal. Chem. 1999, 71, 1521-1528. (3) Szaloki, I.; Osa´n, J.; Ro, C.-U.; Van Grieken, R. Spectrochim. Acta 2000, B55, 1017-1030. (4) Ro, C.-U.; Osa´n, J.; Szaloki, I.; Oh, K.-Y.; Kim, H.; Van Grieken, R. Environ. Sci. Technol. 2000, 34, 3023-3030. 10.1021/ac010438x CCC: $20.00

© 2001 American Chemical Society Published on Web 09/06/2001

NOx to produce sodium nitrate particles in the air.5 In addition, the atmospheric reaction between soil particles and SOx receives considerable attention in the atmospheric environment society. For example, in almost every spring, usually from March to May, Asian dust, originating in central China’s arid areas, is transported into eastern China, Korea, Japan, and even the Pacific Ocean over the Yellow Sea and, sometimes, industrialized regions in China. While the Asian dust travels for a long distance, it can react with sea salts6 and also possibly with gaseous chemical species such as SO2 and NOx7-9 so that the transport of modified Asian dust to the East Asian region could result in the deposition of sulfate and nitrate, in addition to mineral dust, in this area. East Asia is a rapidly industrializing area with high consumption of energy and a large population, resulting in a high anthropogenic emission. It was also calculated that dust particles can contain more sulfate in the aerosol form during the high dust periods, by up to a factor of 2, with almost all of the sulfate being present on the surface of the dust particles.10 This implies that sulfate aerosols in the accumulation mode, which play an important climate role in that they both directly and indirectly reflect solar radiation back to space,11 are reduced by the reaction of SO2 with the mineral dust. The size of Asian dust is large enough to have much smaller radiative forcing than the accumulation range sulfate aerosols, and the sulfate on the surface of Asian dust does not change much the radiative properties of the dust particles.10 If the reaction between CaCO3 in the mineral dust and SO2 occurs, then it is expected that CaSO4 species would be in the surface layer and CaCO3, in the core region, and thus, it would be directly proven if we can characterize both regions in individual particles. However, since the analysis volume of individual microparticles is quite small (picogram range in mass), quantitative analysis of surface and core regions in individual particles has never been tried until now. In this paper, we present a new analytical methodology that can characterize the surface region as well as the core region in individual particles. EXPERIMENTAL SECTION Samples. To investigate the validity of the new methodology by comparing experimental and simulated data, we tried to synthesize artificial particles with different chemical species in the surface and core regions. For preparing reacted calcium carbonate particles, microscopicsized crystals from a pro analysis grade of solid chemical compound were used. The particles were suspended in n-hexane, pipetted onto a silicon wafer, and dried in the air. They were attached upside down to the lid of a glass Petri dish (φ 5 cm), which was filled with sulfuric acid (pro analysis). The distance between the particles and the surface of the liquid was kept very small (3-6 mm), because sulfuric acid is not very volatile. In this way, the particles were exposed to isolated acid vapors for 1-3 (5) Gard, E. E.; Kleeman, M. J.; Gross, D. S.; Hughes, L. S.; Allen, J. O.; Morrical, B. D.; Fergenson, D. P.; Dienes, T.; Galli, M. E.; Johnson, R. J.; Cass, G. R.; Prather, K. A. Science 1998, 279, 1184-1187. (6) Fan, X.-B.; Okada, K.; Nimura, N.; Kai, K.; Arao, K.; Shi, G.-Y.; Qin, Y.; Mitsuta, Y. Atmos. Environ. 1996, 30, 347-351. (7) Song, C. H.; Carmichael, G. R. Atmos. Environ. 1999, 33, 2203-2218. (8) Roth, B.; Okada, K. Atmos. Environ. 1998, 32, 1555-1569. (9) Mamane, Y.; Gottlieb, J. Atmos. Environ. 1992, 26A, 1763-1769. (10) Dentener, F. J.; Carmichael, G. R.; Zhang, Y.; Lelieveld, J.; Crutzen, P. J. J. Geophys. Res. 1996, 101, 22869-22889. (11) Kiehl, J. T.; Briegleb, B. P. Science 1993, 260, 311-314.

days. We also tried to make nitric acid vapor react with the calcium carbonate particles, but the reaction occurred too rapidly, so that we obtained totally reacted calcium nitrate particles (we observed only X-ray peaks from calcium nitrate for those particles), even when the particles were under nitric acid vapor for only a few minutes (nitric acid is more volatile). Carbon-coated glass particles were prepared from 20-µmdiameter calibrated uniform soda-lime glass microspheres of SPI 2716 (Structure Probe Inc., West Chester, PA). Their certified mean diameter is 19.9 ( 1.4 µm, and they have a specific gravity of 2.4-2.5 g cm-3. Although the spheres are not certified standards for chemical composition, it was found that the differences in major composition are very small between different particles selected. The advantage of the use of these particles is their welldefined spherical shape and the fact that they are already conductive without carbon coating. The particles were transferred to silicon wafer substrates using a n-hexane suspension. The particles were coated with carbon layers consecutively four times, using the standard process for scanning electron microscopes. EPMA measurements were performed after each coating. EPMA Measurements. The measurements for heterogeneous particles on the silicon wafer were carried out using a JEOL 733 electron probe microanalyzer equipped with an Oxford energydispersive X-ray detector. The Si(Li) detector is equipped with an Oxford atmospheric ultrathin window of 0.2 µm. The resolution of the detector is 133 eV for Mn KR X-rays. The spectra were recorded by a Canberra S100 multichannel analyzer under control of homemade software. Measurements on individual particles were carried out manually in the point analysis mode. We acquired four X-ray spectra for each particle using four different primary electron energies, namely 5, 10, 15, and 20 kV. Five CaCO3-CaSO4 particles of different sizes and several carbon-coated glass particles were analyzed in this way. The beam current was 1.0 nA for all of the measurements. To obtain statistically enough counts in the X-ray spectra, a measuring time of 50 s was used. In addition, the measurement was carried out at around -193 °C stage temperature using the cold stage of the electron microprobe cooled by liquid nitrogen. By making measurements at low temperature, contamination is minimized, and beam damage to the sample is significantly reduced. The size and shape of each individual particle was measured and estimated from a high-magnification secondary electron image (magnification >10000×). These estimated geometrical data were set as input parameters for the Monte Carlo calculation. The net X-ray intensities for the elements were obtained by nonlinear least-squares fitting of the collected spectra using the AXIL program.12 Monte Carlo Simulations. In our previous study, it was found that excitation interactions between electrons and the atoms of the matrix and the matrix and geometric effects of electroninduced X-ray signals for light elements in individual atmospheric microparticles can be described adequately by Monte Carlo simulation.2 The Monte Carlo simulation is based on the ANSI standard C code CASINO program, developed by Hovington et al.13 The original program can deal with a spherical inclusion embedded in a substrate, but the surface of the whole sample must be flat. Our modification of the program allows the simulation (12) Vekemans, B.; Janssens, K.; Vincze, L.; Adams, F.; Van Espen, P. X-ray Spectrom. 1994, 23, 278-285. (13) Hovington, P.; Drouin, D.; Gauvin, R. Scanning 1997, 19, 1-14.

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

4575

of electron trajectories in spherical, hemispherical, and hexahedral particles sitting on a flat surface. The simulations of X-ray spectra and Bremsstrahlung background are also implemented. The parametrization of Kirkpatrick and Weidmann14 is used for the background calculation. To deal with heterogeneous particles, further modification was done in the Monte Carlo calculation program; this new code can simulate electron trajectories and electron-induced X-rays emission and absorption for the spherical, hexahedral, and hemispherical microparticles with different chemical species in surface and core regions. RESULTS AND DISCUSSION Effect of Primary Electron Beam Energy on Characteristic X-ray Detection. For heterogeneous particles that have different chemical species in surface and core regions, the use of different energies of the primary electron beam is expected to provide useful information on the heterogeneity of the particles, mainly because of the different excitation volumes according to the energies of primary electron beam. The excitation volume of the elements is decreased with the decrease of the primary electron beam energy. If the excitation volume of the low-energy electron beam is comparable with the thickness of the surface region in heterogeneous particles, the obtained X-ray spectra will contain chemical information mainly on the surface region. In contrast, primary electron beams with higher energy would provide relatively more information on the core region. For example, the calculated Kenaya-Okayama electron penetration range15 for CaCO3 is 0.4 µm for a 5 kV electron accelerating voltage, whereas the penetration range at a 20 kV electron accelerating voltage is 3.9 µm. And thus, for a particle of micrometer size, primary electron beams with different energies probe into different regions of the particle. Figure 1 shows an example of electron trajectories for different primary electron beam energies. The model particle is a heterogeneous CaCO3-CaSO4 particle with a 2 µm gross diameter, but the thickness of the CaSO4 surface layer is 0.25 µm. It is clear that primary electron beams with different energies provide different electron probing regions. At a 5 kV electron accelerating energy, the CaSO4 surface region is mostly probed, whereas at 20 kV, many electrons get through the particle, and the signal from the surface region is relatively suppressed. However, there are two additional factors that should be considered: the X-ray generation volume and the absorption of generated X-rays in the particle. In general, the generation volume of X-rays from light elements is more similar to the penetration volume of the electron beam because of their low ionization energies. And the generation volume of X-rays with higher energy gets smaller than its penetration volume. Using Anderson-Hasler equations,15 the X-ray generating ranges at a 5 kV electron accelerating voltage are calculated as 0.3 µm for carbon and oxygen and 0.1 µm for calcium in CaCO3. At a 20 kV electron accelerating voltage, they are 3.5 µm for carbon and oxygen and 3.2 µm for calcium. Carbon and oxygen KR photons are generated in a volume similar to that of the penetration, whereas other characteristic lines, such as those of calcium and sulfur, are generated in a relatively shallower region. Therefore, with increasing primary electron beam energy, (14) Kirkpatrick, P.; Weidmann, L. Phys. Rev. 1945, 67, 321-339. (15) Goldstein, J. I.; Newbury, D. E.; Echlin, P.; Joy, D. C.; Romig, A. D., Jr.; Lyman, C. E.; Fiori, C.; Lifshin, E. Scanning Electron Microscopy and X-ray Microanalysis, 2nd ed.; Plenum Press: New York, 1992.

4576

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

low-energy X-rays are generated at deeper regions than higher energy X-rays. Additionally, the absorption of characteristic X-rays, especially of light elements, plays an important role, even in the particle of micrometer size. Since low-energy photons are much more strongly absorbed than higher-energy photons while they travel through the sample (e.g., the attenuation lengths of oxygen and calcium KR photons are 0.3 and 29.3 µm, respectively, in CaCO3), the characteristic radiation of light elements coming from the shallow region is mainly detected. The penetration and generation volumes of the elements are dependent on the primary electron beam energy, whereas the absorption of generated X-rays is dependent on the geometry of the particle. Application of Monte Carlo Calculation to Experimental Data. Since the detection of characteristic X-rays is a nonlinear function of several factors and since, also, the heterogeneity and the geometric effect of the particles are involved, the modified Monte Carlo simulation was applied to explain measured data for the heterogeneous CaCO3-CaSO4 particles. For a comparison of measurement data with simulation results, five heterogeneous CaCO3-CaSO4 particles, in different size ranges, were measured at four different electron accelerating voltages, namely 5, 10, 15, and 20 kV. As a representative example, X-ray spectra for a heterogeneous CaCO3-CaSO4 spherical particle of 1.5 µm diameter collected on Si wafer substrate are shown in Figure 2B (secondary electron image of the particle is given in Figure 2A). The measured characteristic X-ray intensities for the elements in the particle and the substrate vary with the variation of primary electron beam energies; the intensities of calcium and silicon increase as the beam energy increases. The silicon X-ray intensity increases more drastically than for calcium. Sulfur shows a difference from calcium in the sense that the minimum intensity is obtained at 5 kV excitation voltage, and yet, the intensities are similar at 10, 15, and 20 kV electron accelerating voltages. For oxygen, the X-ray intensity is maximum at 10 kV electron accelerating voltage and mininum at 20 kV. The intensities of carbon X-rays are higher at 10 and 15 kV, and lower at 5 and 20 kV. From the observation of different (characteristic X-ray intensity variation according to the variation of electron beam energies) trends for the elements, these X-ray spectra certainly contain information on chemical species and heterogeneity of the particle. In Figure 3, we show simulated spectra calculated by our modified Monte Carlo program. The model system is a heterogeneous CaCO3-CaSO4 spherical particle of 1.5 µm diameter, and the thickness of the surface region is 0.18 µm. The similarity between the simulated and experimental spectra is remarkably obvious. The Monte Carlo calculation almost perfectly simulates the X-ray intensity variations for the elements according to the variation of the primary electron beam energies. In Table 1, measured and calculated X-ray intensities normalized to the maximum value among the ones for the four different accelerating voltages are shown to demonstrate how closely the Monte Carlo calculation simulates the variation of intensities. There are just four pairs of X-ray intensity values that are different between measured and calculated ones (carbon and sulfur at 5 kV, calcium at 10 kV, and carbon at 15 kV; they are shown in bold in Table 1). Even though some deviations are found, overall trends are extremely well-followed.

Figure 1. Simulation of electron trajectories in a heterogeneous CaCO3-CaSO4 spherical particle at (A) 5, (B) 10, and (C) 20 kV accelerating voltages (assumed overall size of the particle is 2.0 µm in diameter, and surface thickness is 0.25 µm.).

In Figure 4, X-ray intensities calculated by varying electron accelerating voltages are given for the heterogeneous CaCO3-

CaSO4 spherical particle of 1.5 µm diameter and 0.18 µm surface thickness in order to understand the effect of the excitation energy Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

4577

Figure 2. (A) Secondary electron image for an artificially generated heterogeneous particle (spherical, 1.5 µm diameter). (B) X-ray spectra obtained at 5, 10, 15, and 20 kV accelerating voltages from the artificially generated heterogeneous particle (see A).

on the characteristic X-ray photon detection in more detail. In the plot, the calculated intensities for the elements are given after being normalized to the maximum value among the ones for the different accelerating voltages. The detected calcium X-ray intensities increase as the electron excitation energy increases until 13 kV, and then they are relatively constant in the range of 13-20 kV. Sulfur intensities increase from 5 to 11 kV excitation energies, and they somewhat decrease as the electron accelerating voltage gets higher than 11 kV. Silicon X-ray photons are enhanced almost linearly after 9 kV electron accelerating voltage with the increase of the excitation energy. Since these elements have much larger attenuation lengths (29.3 µm for calcium in CaCO3, 17.3 µm for sulfur in CaSO4‚2H2O, and 6.2 µm for Si in SiO2) than the size of the particle, the generated X-rays from these elements are detected without being significantly absorbed in the particle. The X-ray 4578

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

intensities for these elements are mostly dependent both on the beam energy and on the depth where the elements exist. The calcium species is well inside the particle, sulfur is in the surface, and silicon is in the collecting substrate under the particle. As the excitation energy gets higher, the electron excitation volume also gets larger, and more X-rays for these elements are generated. Since silicon is well under the particle, its X-rays are not generated until 9 kV, where the electron beam can start to hit the wafer with passing through the particle. Above the 9 kV, more and more electrons can excite silicon as the beam energy increases, resulting in the increase of its X-rays intensities. The intensities of the calcium X-rays increase with the increase of the excitation energy up to 13 kV, mainly because of increased excitation volume. The reason its intensities are relatively constant above 13 kV energy is that the generation volumes of the X-rays are

Figure 3. Simulated X-ray spectra using Monte Carlo calculation for a spherical heterogeneous CaCO3-CaSO4 particle at different accelerating voltages (overall size of the particle is 1.5 µm in diameter, and surface thickness is 0.18 µm).

Table 1. X-ray Intensities of Elements Normalized to the Maximum among the Ones for the Four Different Accelerating Voltagesa 5 kV

10 kV

15 kV

20 kV

Ca C O S Si

0.03 0.72 0.81 0.29 0.00

Measured 0.51 0.98 1.00 0.87 0.04

1.00 1.00 0.75 1.00 0.36

0.91 0.64 0.39 0.76 1.00

Ca C O S Si

0.02 0.54 0.79 0.50 0.00

Simulated 0.76 1.00 1.00 1.00 0.05

1.00 0.68 0.65 0.89 0.47

0.95 0.54 0.50 0.71 1.00

a Measured intensities are for a spherical CaCO -CaSO particle 3 4 of 1.5 µm diameter; simulated intensities are calculated for one with 0.18 µm CaSO4 surface thickness.

comparable to or larger than the size of the particle. Sulfur is in the surface region, and thus, more characteristic X-rays are generated as the excitation volume gets larger (e.g., in the range of 5-11 kV) and then its intensities somewhat decrease with the increase of the energy, because the electrons probe more and more of the core region at energies higher than 11 kV. Regarding carbon and oxygen, their intensities are mainly dependent on the absorption effect. The attenuation lengths for carbon and oxygen is quite small, for example, 0.5 µm for oxygen in CaSO4‚2H2O and 0.6 µm and 0.3 µm for carbon and oxygen, respectively, in CaCO3. As the electron beam energy increases, the excitation and generation volumes for carbon and oxygen increase; however, their generated X-ray photons should pass through the particle to be detected. That is why carbon and oxygen intensities decrease in the same way above a 9 kV primary electron energy. Below 9 kV energy, their intensities increase with the increase of the excitation energy and, thus, of the generation volume. Carbon

X-ray intensities more rapidly increase than oxygen because carbon species exists in the core region. Estimation of Surface Thickness. There are two fundamental questions to be addressed if this technique is to provide quantitative information for unknown particles with, for example, CaCO3 and CaSO4: (1) Are the particles really heterogeneous? Because sulfur X-rays are detected, there must be sulfate in the particles, and yet, CaCO3 and CaSO4 species may exist homogeneously. (2) If the particles are heterogeneous, then what is the thickness of the CaSO4 surface region? To answer the first question, the Monte Carlo calculation was applied by assuming that the particles are homogeneous. From our previous studies, it was demonstrated that our quantification methodology using the Monte Carlo calculation produces quite reliable elemental concentrations (with an accuracy of better than 10% in atomic fraction) for homogeneous particles.3,4,16 If the particles are homogeneous, then the elemental concentrations should be constant within the accuracy level, under the variation of the electron accelerating voltage. However, as shown in Figure 5, the obtained apparent elemental concentrations for carbon and sulfur significantly vary with varying electron beam energy. The variations are quite high; the variation in carbon concentration is up to 30%, and the one in sulfur is up to 60%. Oxygen and calcium concentrations are relatively constant, reflecting their small compositional differences in atomic fraction between CaCO3 and CaSO4 species. Another interesting point is that the variations of the elemental concentrations show two trends according to the variation of electron beam energy. The apparent carbon concentration increases as the electron beam energy increases, whereas the sulfur concentration decreases as the beam energy increases. The significant variations in the elemental concentrations imply that the particles are heterogeneous and also the trends in elemental concentration variations for carbon and sulfur imply that (16) Osa´n, J.; Szaloki, I.; Ro, C.-U.; Van Grieken, R. Mikrochim. Acta 2000, 132, 349-355.

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

4579

Figure 4. Calculated intensity variation of elements in a heterogeneous CaCO3-CaSO4 particle for varying accelerating voltages (calculated intensities are given after being normalized to the maximum value for the different accelerating voltages).

Figure 5. Apparent atomic fraction variation of elements calculated by assuming the particle to be homogeneous in a heterogeneous CaCO3CaSO4 particle and for varying accelerating voltages.

sulfur is in the surface region and carbon is inside the particles, because electrons with lower energy probe the shallower region of particle. This result strongly supports the heterogeneity of the particles. While assuming the heterogeneity of the particles, the Monte Carlo calculation was applied to determine the thickness of CaSO4 surface region. The Monte Carlo calculation can provide X-ray intensity for each element when the assumed thickness of the 4580

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

CaSO4 surface region is given as an input parameter. In Figure 6, ratios of simulated-to-measured intensities with the variation of the assumed CaSO4 surface thickness for a spherical CaCO3CaSO4 particles of 1.5 µm diameter are shown for a 15 kV accelerating voltage. For oxygen, the ratios of simulated-tomeasured intensities are relatively constant with the variation of the assumed CaSO4 thickness mainly because of their small compositional differences between the two chemical species.

Figure 6. Variation of ratios between simulated and measured X-ray intensities for varying assumed CaSO4 surface thicknesses at 15 kV acceleration voltage for a heterogeneous CaCO3-CaSO4 particle. Table 2. Estimated Thickness of CaSO4 Surface Regions for Heterogeneous Particles particle no. 1

2 hexahedral 0.74 × 2.7

3

4

geometry size, µm

spherical 1.5 (diam)

5 kV 10 kV 15 kV 20 kV

Estimated Thickness of Surface Region from Data with Different Excitation Energies, nm 130-150 240-260 180-200 40-60 180-200 150-180 200-220 50-70 160-200 150-180 180-210 140-160 170-220 160-180 180-220 90-120

However, the ratios for carbon and sulfur between the simulated and measured intensities are strongly dependent on the assumed thickness of the surface CaSO4 region. Furthermore, the ratios for sulfur decrease as the thickness of CaSO4 region decreases, whereas the ratios for carbon and also calcium, even though the trend for calcium is not so obvious as it is for carbon, increase as the thickness of CaSO4 region decreases. We think this result is really strong evidence for the assumed heterogeneity of the particle. For the heterogeneous CaCO3-CaSO4 particles, the sulfur species is in the surface region, carbon is in the core region, and the calcium concentration in the CaCO3 core region is higher than that in CaSO4 surface region (40 vs 29.4% w/w). Therefore, if the assumed CaSO4 surface thickness for the Monte Carlo calculation is thicker than the real one, then the calculated intensities are larger than the measured ones for sulfur, whereas they are smaller for carbon and calcium. From the result in Figure 6, the good match between the simulated and measured data is in the range of 160-200 nm thickness of the surface region. In Table 2, the estimated values of surface thickness, which are obtained in this way, are listed for the five heterogeneous CaCO3-CaSO4 particles. The values are relatively constant for the different particles and excitation energies. However, there are still some variations in the estimated values. This may be due to several possible reasons. Some sulfuric acids may be attached to the surface of the particle without reacting with CaCO3, the CaSO4 species may contain some crystal waters in it like gypsum, CaSO4‚2H2O, and also the

hexahedral 3.3 × 4.4

hexahedral 3.9 × 3.9

5 hexahedral 3.0 × 4.0 70-90 160-190 160-190 120-150

Table 3. Estimated Carbon Thickness for Carbon-Coated Glass Particles accel voltage (kV)

1

no. of carbon layer depositions 2 3

5 10 15 20

30-40 30-40 25-35 20-30

Est Carbon Thickness, nm 60-70 100-110 65-75 90-100 60-70 90-100 50-60 80-90

4

130-140 140-150 145-155 135-145

boundary between the surface and core regions may not be clearly defined and mixed CaCO3 and CaSO4 chemical species might exist in the region between the surface and core regions. To evaluate this technique more systemically, a more controlled system, that is, soda-lime glass particles (SPI 2716) coated with carbon by evaporation, were investigated. The composition of the particles was calculated from the spectra of 13 particles without carbon coating collected at 10 kV. The average composition were found to be the following (in wt %): 50.0 ( 1.6% O, 9.4 ( 1.1% Na, 3.0 ( 0.1% Mg, 0.3 ( 0.09% Al, 33.3 ( 1.4% Si, and 4.0 ( 0.3% Ca. The estimated carbon thicknesses for different numbers of carbon layers obtained by the proposed method are tabulated in Table 3. The obtained preliminary results show good agreement between the values obtained at different accelerating voltages, thus supporting the applicability of the proposed method. Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

4581

Figure 7. X-ray spectra obtained at 5, 10, and 20 kV accelerating voltages for a silicon oxide particle with minor aluminum and carbon contents. Table 4. Measured X-ray Intensities and Calculated Atomic Concentrations for a Silicon Dioxide Particle with Carbon and Aluminum, at Different Electron Accelerating Voltages accelerating voltage 5 kV

10 kV

20 kV

element

X-ray intensity

at. concn, %

X-ray intensity

at. concn, %

X-ray intensity

at. concn, %

carbon oxygen aluminum silicon

290 3721 135 1352

7.0 62.9 3.9 26.2

293 3538 193 2896

15.1 60.7 1.5 22.7

284 3532 214 5851

17.3 57.7 1.0 24.1

Further research is needed to evaluate the accuracy and reproducibility of this approach using well-defined heterogeneous particles and also to optimize the iteration procedure for automatic evaluation for the surface layer thickness of heterogeneous particles. Application of the Technique to Atmospheric Environmental Particle: One Example. The method based on the Monte Carlo calculation was applied for the analysis of individual particles collected during a so-called “Asian dust” event on a roof of Hallym University in ChunCheon, Korea. A 7-stage May cascade impactor was used to collect particles on a silver foil. We chose one illustrative particle from the 92 particles analyzed, just to demonstrate the potential applicability of this approach. It is a silicon oxide particle with minor aluminum and carbon contents. The reason for choosing this example is that the characteristic aluminum X-ray has a quite similar energy to that of silicon. And thus, the detected X-rays for aluminum and silicon are primarily dependent on the excitation volumes due to the primary electron energies, which make intuitively understanding the nature of this analytical approach easier. In Figure 7, X-ray spectra for a hexahedral particle of 1.7 × 2.7 µm size are shown at 5-, 10-, and 20 kV electron accelerating energies. With the increase of the electron beam energies, the silicon X-ray intensities significantly increase, whereas the aluminum X-ray intensities do not change as much as for silicon. However, it is somewhat difficult to get the information on the heterogeneity of the particle, just based on their observed intensities. In Table 4, measured intensities and 4582 Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

calculated atomic concentrations are given. The atomic concentrations are calculated by assuming that the particle is homogeneous. In this case, the apparent silicon concentration is relatively constant under the different electron beam energies, which implies that the silicon species exist homogeneously in the whole particle. Atomic concentrations of aluminum significantly decrease (e.g., 3.9% at 5 kV and 1.0% at 20 kV) as the energy of the electron beam increases from 5 to 20 kV, even though the aluminum intensities increase by 59% (135 at 5 kV and 214 at 20 kV). This means that the aluminum species exists more in the surface region. Carbon seems to be more in the core region (see atomic concentrations of carbon in Table 4). For oxygen, the atomic concentration slightly decreases as the increase of the beam energy, reflecting that the carbon species is not bonded with oxygen. However, for real atmospheric particles, another complexity is involved in the analysis: we do not accurately know the elemental concentrations of aluminum and carbon, as well as the structure of the heterogeneity. Further research is needed to find a way to extract information both on chemical compositions and the heterogeneity of the atmospheric particles from their X-ray spectral data. CONCLUSIONS This EPMA study demonstrated that the heterogeneity of single microparticles can be characterized by varying the energy of the primary electron beam. The Monte Carlo calculation procedure was proven to be a quite reliable tool for the analysis

of the X-ray data to obtain information on the heterogeneity of the particle. There have been works that tried to analyze thin film samples, simpler than the particle sample system, using the different primary electron beam energies in EPMA.17,18 The idea is that different excitation energies probe into different regions of a thin film sample, so that X-ray data can provide information on the depth profile of the sample. In thin-film EPMA, data analysis has been a key issue, because measured X-ray data contain convoluted information on the chemical composition and depth profiles. It is difficult to obtain clearly deconvoluted information from the X-ray data to determine both the chemical composition and the depth profile of the sample. This situation is also true for (17) Karduck, P. Mikrochim. Acta [Suppl.] 1998, 15, 109-123. (18) Richter, S.; Lesch, N.; Karduck, P. Mikrochim. Acta [Suppl.] 1998, 15, 125131.

the particle system. The development of a method to obtain both the chemical composition and the heterogeneity information together from the X-ray data is essential for the practical application of this new methodology. ACKNOWLEDGMENT This work was financed by Korea Research Foundation Grant (KRF-2000-041-D00303). The support of the Belgian State Office for Scientific, Technical, and Cultural Affairs through a research project (in the framework of the program “Sustainable Management of the North Sea”; contract MN/DD/10), is also highly appreciated. Received for review April 16, 2001. Accepted July 16, 2001. AC010438X

Analytical Chemistry, Vol. 73, No. 19, October 1, 2001

4583