Qualitative Multiplatform Microanalysis of Individual Heterogeneous

Sep 14, 2014 - High-resolution microscopic analysis of individual atmospheric particles can be difficult, because the filters upon which particles are...
0 downloads 8 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

Qualitative Multiplatform Microanalysis of Individual Heterogeneous Atmospheric Particles from High-Volume Air Samples Joseph M. Conny,* Sean M. Collins,† and Andrew A. Herzing Material Measurement Laboratory National Institute of Standards and Technology, 100 Bureau Drive, Stop 8372, Gaithersburg, Maryland 20899-8372, United States S Supporting Information *

ABSTRACT: High-resolution microscopic analysis of individual atmospheric particles can be difficult, because the filters upon which particles are captured are often not suitable as substrates for microscopic analysis. Described here is a multiplatform approach for microscopically assessing chemical and optical properties of individual heterogeneous urban dust particles captured on fibrous filters during high-volume air sampling. First, particles embedded in fibrous filters are transferred to polished silicon or germanium wafers with electrostatically assisted high-speed centrifugation. Particles are clustered in an array of deposit areas, which allows for easily locating the same particle with different microscopy instruments. Second, particles with light-absorbing and/or light-scattering behavior are identified for further study from bright-field and dark-field light-microscopy modes, respectively. Third, particles identified from light microscopy are compositionally mapped at high definition with field-emission scanning electron microscopy and energy-dispersive Xray spectroscopy. Fourth, compositionally mapped particles are further analyzed with focused ion-beam (FIB) tomography, whereby a series of thin slices from a particle are imaged, and the resulting image stack is used to construct a three-dimensional model of the particle. Finally, particle chemistry is assessed over two distinct regions of a thin FIB slice of a particle with energy-filtered transmission electron microscopy (TEM) and electron energy-loss spectroscopy associated with scanning transmission electron microscopy (STEM). sampling offer low resistance to air flow, because the spacing between fibers is large, relative to particle size,8 yet are highly efficient because particles often encounter multiple layers of fibers before attachment to individual fibers.9 Nevertheless, particle-embedded filters are often unsuitable for microscopy, because the particles are only indirectly exposed to the energy source (e.g., light or electrons) due to scattering by the filter matrix. X-ray microanalysis is also problematic, because of the scattering of X-rays by the filter matrix. Thus, it is often necessary to transfer particles to a more-suitable substrate. The ideal substrate for electron microscopy conducts excess electron charge from specimen to ground, provides sufficient imaging contrast with the specimen, and its X-ray spectrum does not interfere with that of the particles. Substrate smoothness may help minimize within-particle fluorescence by substrate electrons that backscatter into the particle if, for example, the particle sits in a crevice. Polished silicon and germanium crystal wafers, along with glassy carbon planchets, provide ideal smoothness. Observation of Light-Absorbing and Light-Scattering Particles with Bright-Field and Dark-Field Light Microscopy. Bright-field and dark-field modes in reflected light

T

he interaction of sunlight and long-wave radiation with atmospheric aerosols is affected by the spatial arrangement of optically diverse chemical phases within individual particles.1−3 Thus, particle heterogeneity plays a major role in the direction and magnitude of direct radiative forcing of climate by aerosols.4,5 Increasingly, high-resolution microscopy techniques are used to identify spatially resolved features in individual atmospheric particles.6 However, analytical challenges exist in assessing individual atmospheric particles microscopically,7 not the least of which is retrieving particles from high-volume air samples. High-volume sampling typically involves fibrous filters that are unsuitable for microanalysis, and, thus, selected particles must migrate to a substrate that is compatible with multiple microscopic techniques and whereby individual particles can be located when analyzed by different instruments. Here, we demonstrate how individual coarse-mode (2.5−10 μm) dust particles from fibrous filters are retrieved on a versatile substrate and microscopically assessed using a variety of microanalytical techniques. The approach involves light microscopy, high-definition qualitative electron microanalysis with scanning electron microscopy (SEM), focused ion-beam (FIB) tomography with SEM, three-dimensional (3-D) spatial modeling, and electron energy-loss imaging and spectroscopy with transmission electron microscopy (TEM). Retrieval of Particles from Fibrous Filters Used in Aerosol Samplers. Fibrous filters used in high-volume air This article not subject to U.S. Copyright. Published 2014 by the American Chemical Society

Received: June 19, 2014 Accepted: September 14, 2014 Published: September 14, 2014 9709

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716

Analytical Chemistry

Article

bonding.15 We show how π-bonding and σ-bonding between carbon atoms revealed from EELS can provide insight into whether carbonaceous inclusions are graphitic with more πbonding or organic with more σ-bonding.

microscopy with magnifications as low as 50× provide a quick and simple way to qualitatively observe the optical character of heterogeneous coarse-mode particles on a smooth reflective substrate. Illumination from above the specimen allows for the detection of light that is angularly scattered, i.e., diffracted, or reflected directly back toward the microscope objective. In bright-field mode, undiffracted light from a highly reflective substrate contrasts sharply with an absorbing specimen. In dark-field mode, light that is directly reflected by the substrate (or specimen) is blocked by an opaque stop, and only angularly scattered light from the specimen is detected. High-Resolution Spatial Mapping of Particle Elements with SEM and Energy-Dispersive X-ray Spectroscopy (EDX). To observe inclusions and adducts with electron microscopy and EDX, the electron beam must be bright but spatially and energetically narrow, which modern field-emission electron microscopes provide. For an individual particle occupying a raster field width of 5 μm and with pixel densities of 1024−2048, the electron probe at the specimen surface should have a diameter equal to or less than the pixel size (4.9 nm to 2.4 nm), while delivering a probe current capable of inducing sufficient characteristic X-ray emission at each pixel. Based on the manufacturer’s specifications, the Schottky field emitter in the scanning electron microscope used in this study had a probe resolution of 1.1 nm at 15 keV while delivering a current of as much as 20 nA to the specimen. Three-Dimensional Spatial Rendering of Particle Composition Based on Focused Ion-Beam SEM Tomography. Two-dimensional imaging and element mapping with SEM provide only limited insight into the internal composition of heterogeneous particles. The depth of particle phases may be revealed to some degree by tilting the stage and having the beam enter the particle at an oblique angle.7 However, a more comprehensive picture of particle composition is revealed with tomography involving focused ion-beam SEM (FIB-SEM).10,11 FIB-SEM tomography is the process of reconstructing a threedimensional structure or chemical composition of the specimen by serially removing slices (15−20 nm) of material with the ion beam (typically gallium) and collecting electron images and/or compositional maps of the newly exposed faces of the specimen.12,13 For example, a stack of 150 images of 20-nm slice faces from a 3-μm particle is sufficient to create a highresolution 3-D representation of the entire particle. Spatially Resolved Nanoscale Chemical Analysis with Energy-Filtered TEM (EF-TEM) and Electron Energy-Loss Spectroscopy (EELS). A 3-D compositional picture from FIBSEM tomography and EDX provides little insight, however, into the allotropic, molecular, or crystalline forms of a particle’s inclusions and adducts. Without allotropic information, assumptions must often be made regarding the complex refractive indices of materials for optical property calculations. We show how insight into the chemical forms of inclusions can be acquired from EF-TEM images and the spatially resolved spectra of energies absorbed by the particle from the primary beam (EELS). Here, the particle is first identified as optically interesting from light microscopy and then milled with FIBSEM to a thickness that is ideally less than the electron mean free path for a single inelastic scattering event. We believe this is the first report of the preparation of a section of an ambient coarse-mode atmospheric particle by FIB-SEM for analysis via EELS. EF-TEM imaging and EELS allow us to distinguish crystalline from amorphous carbon.14 Energy-loss spectra may also provide insight into optical properties regarding carbon

■

EXPERIMENTAL SECTION Coarse-mode particles were collected for 24 h on heat-purified quartz-fiber filters in a dichotomous (dichot) virtual impactor (MSP Model 310 Universal Air Sampler), with an airflow at 285 L min−1, during December 2004 at a sampling site near downtown Los Angeles. [Commercial products identif ied here specif y the means by which experiments were conducted. Such identif ication is not intended to imply recommendation or endorsement by the National Institute of Standards and Technology (NIST) nor that the identif ied products are necessarily the best available for the purpose.] Particles were then transferred to semiconductor wafers as described later. Light microscopy was performed with a Leitz Aristomet stereomicroscope in reflection mode and with the light source maintained at the same illumination level for all particles. The sample particle population was analyzed with an automated Aspex Personal SEM using a 25 keV beam, 1 nA of current, and a working distance of 18 mm. The probe resolution was 5−10 nm. Secondary electron (SE) imaging was used to determine particle size and aspect ratio. Particles with large aspect ratios were identified as filter fibers and, therefore, omitted from analysis. Quantitative silicon-drift EDX analysis16 was based on k-ratios17 of the measured X-ray spectra of reference materials. Element mapping of individual particles was accomplished at 15 keV with an Oxford X-MAX 80 mm2 silicon-drift EDX detector on an FEI Nova NanoLab 600 DualBeam (FIB-SEM). Milling of particles to reveal their internal composition was also accomplished with the FIB-SEM system. Platinum was deposited, via ion-beam decomposition of injected trimethyl(methylcyclopentadienyl)platinum gas, to protect particles from ion-beam damage and to provide SE contrast between particle and background. Additional details of particle milling were reported earlier.11 3-D visualization of particles from the reconstruction of FIB-SEM slice SE images by the segmentation of gray scale contrast in the images was accomplished using Avizo software. EF-TEM was performed using a Philips CM-300 TEM system operating at 300 keV and equipped with a Model GIF 200 imaging energy filter (Gatan, Inc.). Low-loss hyperspectral image data were collected by serial acquisition of energy-filtered images from 0 eV loss to 90 eV loss using a 5 eV wide selection slit and an energy step of 5 eV. Each image was acquired using an exposure time of 2 s. The image data were subsequently processed by principal component analysis (PCA), using AXSIA software.18 EELS data were acquired using a FEI Titan 80-300 TEM/ STEM operating at 300 kV and equipped with a Gatan Model Tridiem 865 high-resolution spectrometer. Core-loss spectra were acquired over a 200 nm × 200 nm region of the specimen containing both matrix and particle phases using a 2.0 nm pixel size. Each spectrum was collected for 100 ms using a probe convergence angle of 13 mrad and a spectrometer collection angle of 13 mrad. The background was subtracted from the core-loss spectra by fitting the pre-edge background to the power function, I = AE−r, where I and E are the intensity and energy loss, respectively, while A and r are curve-fitting parameters.19 9710

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716

Analytical Chemistry

Article

Background fit intervals were 15 eV for aluminum and 35 eV for carbon, titanium, oxygen, and iron.

wafers do allow for the analysis of silicates, at high beam energies (e.g., 20−30 keV) K-band X-rays from germanium may fluoresce elements in particles with an absorption edge close to but less than the critical excitation energy of germanium (e.g., zinc and copper). Particle Optical Behavior From Light Microscopy. Because of their high specular reflectivity, polished silicon and germanium wafers are ideal substrates for viewing particles with reflected-light microscopy. Light-absorbing and light-scattering phases in heterogeneous particles can often be observed with juxtaposed bright-field and dark-field views, as shown in Figure 2. While particle geometry alone may enhance weakly scattering

■

RESULTS AND DISCUSSION Particle Retrieval from Filters. In this study, particles were transferred from quartz fiber filters to polished and cleaned 5 mm × 5 mm silicon or germanium wafers using a technique involving electrostatic charging and high-speed microcentrifugation. As shown in Figure 1, during centrifuga-

Figure 1. Migration of particles from a fibrous filter to polished silicon wafer. (a) Components of the transfer apparatus used in a microcentrifuge; polytetrafluoroethylene (PTFE) disks and the silicon wafer are electrostatically charged at −30 kV. (b) Dark-field light microscopy image of coarse-mode particles in an array of deposit areas. (Inset shows areas of filter coincident with screen holes where particles were removed from the filter (see feature noted by arrow).) (c) SEM image of a deposit area from a sample prepared in a manner similar to that described by panel b; fibrous strands are from the filter material.

tion, particles migrate from the filter to the wafer through a photochemically etched stainless-steel screen. Prior to centrifugation, peripheral polytetrafluoroethylene (PTFE) disks are electrostatically charged at −30 kV along with the wafer, which helps facilitate the migration. The outcome is an array of particle deposit areas (Figure 1b). Figure 1c is an individual deposit area. Large particles are more likely to be retrieved by this technique than small particles, because of both greater mass and surface area. Inertial forces causing particles to migrate to the wafer during centrifugation are proportional to mass; thus, larger particles are more likely to migrate. Particles with larger surface area retain more electrostatic charge, and their migration to the wafer is more likely to be assisted by charging. Thus, it is unlikely that all particles can be retrieved by this technique. Nevertheless, the inset of Figure 1b shows that, within a screen hole, the filter becomes depleted of many particles during the transfer process. In SEM, the relatively low SE and backscatter electron signals from silicon allowed for sufficient image contrast between the silicon wafer and the higher-atomic-number elements in the particles. For energy-dispersive X-ray (EDX) analysis, however, the use of silicon substrates typically precludes mapping of silicates unless done at low beam energy, e.g., 5 keV, to avoid having the beam penetrate the silicon substrate below the particle, where X-rays would be emitted.7 While germanium

Figure 2. Light microscopy of particle deposit area on a silicon wafer with individual particles 1−3 further magnified: (a) bright-field images and (b) dark-field images. Magnification of deposit area is 50×. Scale bar for individual particles (shown in the upper right inset of panel a) is 1 μm.

particles in dark-field images, particles 1−3 in Figure 2 appear to have distinctive light-scattering phases while absorbing strongly overall. For comparison purposes, the pixel area within the particles in the dark-field images was apportioned based on the pixel gray scale values. Histograms shown in Figure S-1 in the Supporting Information did not show consistent boundaries for grouping gray-scale values, particularly for the higher values. Therefore, values were arbitrarily grouped in three equally sized bins as follows: (1) 1−85, (2) 86−170, and (3) 171−255. If we associate the third bin (171−255) with strong light scattering, then 4.5%, 2.3%, and 4.1% of the pixel area in particles 1, 2, and 3 in Figure 2, respectively, is designated as strongly lightscattering. Each deposit area in the array on the wafer contained a manageable number of particles (typically 10 wt %) and minor (170 and, thus, are designated as strongly light scattering (see Figure S-4 in the Supporting Information). The chlorine-containing particle (4), possibly sea salt, appears to be the least absorbing among the five types, with over 40% of the dark-field image pixels designated as strongly light scattering (see Figure S-5 in the Supporting Information). The element map (image 4-d in Figure 3) shows the heterogeneity of calcium and magnesium. The brake-wearlike particle (5) exhibits the least scattering among the five types, with no dark-field image pixels designated as strongly light scattering. Barium and iron are present throughout, while a carbon phase appears to be a surface adduct based on the element map (image 5-d in Figure 3) and the SE image (image 5-c in Figure 3). The iron-containing particle (2) also contains barium (albeit in a small region of the particle) and, thus, is compositionally similar to the brake-wear-like particle. However, the two particles are clearly different, based on their lightscattering behavior from the dark-field images. FIB-SEM Tomography and Spatial 3-D Modeling. Once particles of optical interest are identified from light microscopy and their heterogeneity is observed from two-dimensional (2D) element mapping, FIB-SEM is used to characterize their internal structure. Figure S-6 in the Supporting Information shows the rich compositional complexity of an interior surface of a strongly light-absorbing mineral-like particle. Sequential slicing and imaging with FIB-SEM allows us to create a 3-D representation of a particle’s shape and interior structure from a stack of tens to hundreds of images. As shown in Figure 4 for a mineral-like particle that had migrated to a germanium wafer, the particle (Figure 4a) is first embedded in a block of platinum (Figure 4b). To allow platinum to reach below the particle, it is first deposited as four wedges around the particle (Figure 4b). This is accomplished with the stage at 0° tilt so that the ion beam is oriented 52° from normal. The stage is then tilted 52° so that the ion beam is at normal to deposit a platinum cap (upper right inset in Figure 4b). Lastly, the deposition is trimmed to form the platinum block (lower right inset of Figure 4b). Embedding the particle in platinum provides for consistent SE contrast between the particle and background at a slice face (Figure 4c). The image stack acquired during sequential slicing is then input to 3-D rendering software to create a spatial model of the particle. Figure S-7 in the Supporting Information shows the 3-D rendering of the particle’s shape. The 3-D model of the particle’s internal composition based on SE imaging is shown in Figure 5. The 2-D element map (Figure 5a) suggests that the particle has two calcium and three titanium phases in an aluminum oxide matrix. Figure 5b is the

Figure 4. FIB milling of a mineral-like particle after first embedding it in a block of platinum: (a) overhead view of the particle, (b) embedding the particle in platinum (Pt wedges, Pt cap, trimmed block); and (c) milled slice showing SE image contrast between the particle and platinum. The arrow shown in panel c indicates a void in the Pt block.

Figure 5. Internal composition from FIB tomography of the minerallike particle in Figure 4: (a) 2-D element map with location of EDX detector and (b−d) 3-D spatial models based on SE imaging in the x− y plane (panel b), x−z plane (panel c), and y−z plane (panel d). Phases 1, 4, and 5 are titanium, and phases 2, 3a, and 3b are calcium. Arrows in panel a point to numbered phases in panels b−d. Arrows in panels b−d indicate the location of one of the inclusions that is not indicated in the 2-D element map shown in panel a.

model overhead (x−y) view corresponding to the map view shown in Figure 5a; Figures 5c and 5d are the model side views (x−z and y−z). The model nominally exhibits 12 inclusion phases (more than twice the number determined from 2-D EDX mapping). Maps were not acquired at individual FIB-SEM slices; therefore, we do not know the elemental composition of all of the inclusions revealed by FIB tomography. Nevertheless, 9713

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716

Analytical Chemistry

Article

aluminum is a strong absorber of C Kα X-rays (mass absorption coefficient = 27 234 cm2 g−1 (see ref 21)), so it is plausible that one or more of the inclusions revealed by tomography but absent from the 2-D element map (red phases in Figures 5b− d) are carbonaceous, e.g., elemental carbon. The 3-D model elucidates the actual sizes, shapes, and depths of the phases identified from 2-D element mapping. In Figure 5c, we see that titanium phase 1 is substantially larger than that indicated by the map. This is primarily due to the shape of the titanium phase, as viewed in Figure 5b, rather than the absorption of X-rays by aluminum in the matrix before the Xrays reach the detector (note the detector position in Figure 5a), since the mass absorption coefficient of the Ti Kα X-rays by aluminum is relatively small (264 cm2 g−1). Similarly, in Figure 5b, we see that calcium phase 2 is larger than that from the element map, perhaps because of X-ray absorption by an overlaying phase (indicated by the arrow in Figures 5b−d). Calcium phase 3 in the element map is actually two separate phases: 3a and 3b (Figure 5d). In contrast, only one of the two titanium phases (features “4” and “5” in the map shown in Figure 5a) is discernible in the model (labeled as “4” in Figures 5b−d). Internal Composition from Electron Energy Loss in TEM. In TEM analysis, primary electrons that lose significant energy as they pass through a specimen can be energy-selected to show compositional variation at the nanoscale. However, specimen thickness should be on the order of the mean free path for a single scattering event to avoid additional peaks in the energy-loss spectrum due to multiple inelastic scattering. For example, with a beam energy of 200 keV, the total (plasmon and inner shell ionization) inelastic mean free path for amorphous carbon is 160 nm.14 Thus, with sizes in micrometers, coarse-mode atmospheric particles are typically unsuitable for TEM. In this study, ion-beam milling with FIBSEM was used to prepare a thin section of a particle for TEM. Figure 6 shows a SE image of the face of a particle section prepared with the FIB along with images from TEM and scanning TEM (STEM) analyses of the particle section. Here, the mineral-like particle shown in Figure S-6 in the Supporting Information was selected, with the section taken at position 2 in Figure S-6c. FIB preparation of the particle section is shown in Figure S-8 in the Supporting Information. The thinnest part of the section in Figure 6a is nominally ∼320 nm. We can compare the slice thickness to the total inelastic mean free path (λ), using the following measurement-based approximation:14,22

Figure 6. EF-TEM and STEM EELS imaging of a mineral-like particle (see Figure S-6 in the Supporting Information): (a) surface of a thin slice of the particle showing the region of EF-TEM imaging; (b) lowloss EF-TEM PCA image of region in panel a (EF-TEM was performed with the image frame rotated ∼25° counterclockwise); (c) dark-field STEM image of the EF-TEM region in panel a, indicating the smaller region of core electron energy-loss imaging with STEM; (d) region in panel c where EELS spectra were taken for the inclusion and matrix (Figure 7); and (e) energy-loss map from STEM of region in panel c.

F=

11ρ0.3

⎡ ln⎢ ⎣

(

2

2

α + β + 2θE2 + δ 2 2

2

α +β +

2θc2 +

δ

2

)

θc2 ⎤

θE2 ⎥ ⎦

(1)

Here, E0 is the beam energy (in keV), ρ is the material density (in g cm−1), α is the incident convergence angle determined at 13 mrad, β is the collection semiangle determined at 13 mrad, δ2 = |α2 − β2|, and θc = 20 mrad. The angle θE is defined as follows:14,22 θE =

5.5ρ0.3

(

E 0F 1 +

E0 511 keV

)

(1 +

E0 1022 keV 2 E0 511 keV

)

(3)

As discussed below, the slice likely includes amorphous carbon and iron oxide, likely Fe2O3. At 300 keV, λ = 192 nm for amorphous carbon with a density of 1.95 g cm−1 and λ = 150 nm for Fe2O3 with a density of 5.24 g cm−1. Thus, the nominal slice thickness in Figure 6a is ∼1.7−2.1 times larger than the mean free path, and we cannot rule out multiple inelastic scattering. EF-TEM and EELS Imaging. In conventional TEM, EFTEM imaging is accomplished by broadly flooding the specimen with primary electrons and filtering the transmitted electrons within an energy slit width at pixel positions on the image plane. In this process, an energy-loss spectrum is limited by the size of the energy slit width. In scanning TEM (STEM), an energy-loss spectrum is collected at each pixel of the scanned electron beam. Compared with EF-TEM, resolution in the energy-loss spectrum from STEM is greater, and moredefinitive chemical information may be derived. Because of limited spectral resolution, element-specific information is often not accessible with EF-TEM. However, PCA may be used to identify statistically correlated features related to composition, with a spectrum derived for each significant principal component. Figure 6b shows the overlay of the principal component images derived from the raw low-loss EF-TEM image data (see Experimental Section) for a fragment of the rectangular region outlined in Figure 6a. The sharp color contrast between the left and right sides of Figure 6b is due to the abrupt increase in the section thickness (shown in Figure S-

200FE0

λ≈

1+

(2)

F in eqs 1 and 2 is a relativistic factor, defined as follows:23 9714

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716

Analytical Chemistry

Article

L3-edge and its larger size, relative to the higher-energy Fe L2 peak, suggests that the iron phase consists of hematite.24 Iron, titanium, and aluminum were also detected via SEM-EDX (see Figures S-6e and S-6f in the Supporting Information). The C K-edge and ELNES between 285 eV and 340 eV in Figure 7b and the O K-edge centered at 532 eV14 and ELNES in Figure 7d provide insight into the chemistry of the matrix. We observe three C K ionization features. First, there is a modest shoulder at 290.5 eV (also in the inclusion spectrum) that appears to correspond to the 1s electron excitation to an unoccupied π-orbital (*π). Second, the main C K peak at 302.5 eV corresponds to the 1s electron excitation to an unoccupied σ-orbital (σ*). The corresponding peak in the inclusion spectrum has the same or perhaps slightly higher energy maximum (303−306 eV). Third, a broad peak exists between 320 and 340 eV into the region associated with extended energy-loss fine structure. For graphitic-like carbon, the π* feature is typically a resolved narrow peak indicating the prevalence of ordered and planar sp2-hybridized bonding as in graphite. For amorphous carbon, the π* feature is typically a shoulder indicating less sp2hybridized bonding.15,25 The energy separation between the π* and σ* provides an indication of average bond length, and thus the presence of single, double, and/or triple bonds.26 The σ*−π* separation for the matrix is 12 eV, which resembles the energy separation for amorphous carbon;15 this has been reported for doubly bonded trans-1,3-butadiene.26 The broadness of the σ* peak is similar to that observed with near-edge X-ray absorption fine structure, indicating a randomness of bonding states and, thus, a lack of molecular order.27 Results here suggest that the matrix contains amorphous carbon with nonordered single and double bonds. The position of the π*-transition (302.5 eV) is perplexing, because if the matrix is largely carbonaceous, then the position of π* should be several eV less at 284−285 eV,28 although a transition as high as 287.5 eV was reported for cyclohexane.27 The σ* peak position does not change with sample thickness,15 nor would we expect the π* peak to change with thickness. The π* peak position does change substantially, however, if the material is carbonate. Mineral carbonates typically exhibit a large narrow π* peak at 290.2 eV and a broad σ* peak at 301.3 eV,24 in agreement here with peak positions for the matrix as well as the inclusion (Figure 7b). Thus, it appears that the matrix contains carbonate, in view also of oxygen present in the matrix (Figure 7c). The EDX map of a slice of the particle (see Figures S-6e and S-6f in the Supporting Information) showed inclusions of aluminum, calcium, titanium, and iron where carbonate may be present. Nevertheless, much of carbon in the particle slice (Figure S-6e in the Supporting Information) is not associated with the inclusions. Moreover, we do not find significant evidence of metals in the matrix EELS spectra (see Figure 7). While carbonate may be present in the particle, it is likely that the matrix region in Figure 6d contains substantially more amorphous carbon. A key indicator is the π*/σ* intensity ratio. The ratio is reportedly 0.56−0.62 for mineral carbon,15 but, in fact, it may be >1 for pure carbonates.24 The π*/σ* intensity ratio for the matrix in our study is 0.48, below the reported range for partially sp2-ordered carbon black (0.51−0.60), but within the range reported for amorphous carbon in soot aggregates (0.44−0.60).15

8 in the Supporting Information), thus revealing the effect of multiple inelastic scattering. As Figure S-8 in the Supporting Information shows, the thickness is approximately doubled, going from left to right in the PCA image of Figure 6b. PCA may be qualitatively valid for showing inclusion features if the EF-TEM image field covers either a consistently thick section or thin section. However, it is clear in this case that if the sample thickness varies by a factor of 2 within the image field, the EF-TEM PCA may be misleading. Figures 6c−e show the STEM analysis of the particle section. Figure 6c is the STEM high-annular-angle dark-field (STEMHAADF) image of the outlined region in Figure 6a. Figure 6d is the small outlined region in Figure 6c for EELS. Figure 6e is the energy-loss map based on the core−electron energy-loss spectrum at each pixel. The EELS map isolates the C K, O K, and Fe L2,3 ionization edges of the energy-loss spectrum. It is clear from the energy-loss map that the inclusion is an iron oxide embedded in a carbon-containing matrix. Energy-Loss Spectra. EELS ionization edges occur at slightly higher energies than electron binding energies and typically exhibit an abrupt increase in intensity. Spectral features in the vicinity of the ionization edges are known as energy-loss near-edge structure (ELNES). Figure 7 shows ionization edges

Figure 7. EELS ionization edges and ELNES: (a) Al L1-edge, (b) C Kedge, (c) Ti L2,3-edge, (d) O K-edge, and (e) Fe L2,3-edge in the inclusion and matrix shown in Figure 6d. Solid line represents matrix data; dashed line represents inclusion data. In panel b, features “1” and “2” indicate π*- and σ*-transitions in the C K-edge, respectively; feature “3” indicates ELNES.

and associated ELNES in the overall (pixel-integrated) energyloss spectra for the matrix (solid line) and inclusion (dashed line) regions delineated in Figure 6d. Figures 7a and 7c reveal Al L1 and Ti L3 edges, centered at 118 and 461 eV, respectively,14 indicating that trace amounts of these elements are present in the inclusion. (For trace species, background subtraction may result in negative intensities.) More abundant in the inclusion is iron (Figure 7e), with the Fe L2,3-edges centered at 721 and 708 eV.14 The shape of the Fe 9715

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716

Analytical Chemistry

■

Article

(12) Holzer, L.; Muench, B.; Wegmann, M.; Gasser, P.; Flatt, R. J. J. Am. Ceram. Soc. 2006, 89 (8), 2577−2585. (13) Uchic, M. D.; Holzer, L.; Inkson, B. J.; Principe, E. L.; Munroe, P. MRS Bull. 2007, 32, 408−416. (14) Egerton, R. F. Electron Energy-Loss Spectroscopy in the Electron Microscope, 3rd Edition; Springer: New York, 2011. (15) Katrinak, K. A.; Rez, P.; Buseck, P. R. Environ. Sci. Technol. 1992, 26, 1967−1976. (16) Newbury, D. E. Spectroscopy 2009, 24 (7), 32. (17) Goldstein, J. I.; Newbury, D. E.; Echlin, P.; Joy, D. C.; Lyman, C. E.; Lifshin, E.; Sawyer, L.; Michael, J. R. Scanning Electron Microscopy and X-ray Microanalysis, 3rd Edition; Kluwer Academic Publishers and Plenum Press: New York, 2003. (18) Kotula, P. G.; Keenan, M. R.; Michael, J. R. Microsc. Microanal. 2003, 9 (1), 1−17. (19) Williams, D. B.; Carter, C. B. Microanalysis with Ionization-Loss Electrons. In Transmission Electron Microscopy; Plenum Press: New York, 1996; Vol. 4, pp 669−685. (20) Iijima, A.; Sato, K.; Yano, K.; Tago, H.; Kato, M.; Kimura, H.; Furuta, N. Atmos. Environ. 2007, 41, 4908−4919. (21) Heinrich, K. F. J. In Proceedings of the 11th International Conference on X-ray Optics and Microanalysis, Brown, J. D., Packwood, R. H., Eds.; University of Western Ontario: London, Ontario, Canada, 1986; p 67. (22) Iakoubovskii, K.; Mitsuishi, K.; Nakayama, Y.; Furuya, K. Phys. Rev. B 2008, 77, 104102-1−104102-7. (23) Malis, T.; Cheng, S. C.; Egerton, R. F. J. Electron Microsc. Tech. 1988, 8, 193−200. (24) Garvie, L. A. J.; Craven, A. J.; Brydson, R. Am. Mineral. 1994, 79, 411−425. (25) Egerton, R. F.; Whelan, M. J. J. Electron Spectrosc. Relat. Phenom. 1974, 3, 232−236. (26) Hitchcock, A. P.; Beaulieu, S.; Steel, T.; Stohr, J.; Sette, F. J. Chem. Phys. 1984, 80 (9), 3927−3935. (27) Comelli, G.; Stohr, J.; Robinson, C. J.; Jark, W. Phys. Rev. B 1988, 38 (11), 7511−7519. (28) Kincaid, B. M.; Meixner, A. E.; Platzman, P. M. Phys. Rev. Lett. 1978, 40 (19), 1296−1299.

CONCLUSIONS The analysis presented here of individual heterogeneous atmospheric particles using a variety of microscopic techniques (bright/dark-field light microscopy, SEM-EDX, FIB tomography, EF-TEM, STEM EELS) is facilitated by the migration of particles from fibrous filters to polished silicon or germanium wafers in an array of deposit areas, each containing a manageable number of particles whereby the location of each particle of interest can be registered. Because of their high specular reflectivity, polished wafers provide an ideal substrate for light microscopy, particularly for observing particle light scattering in dark-field mode. In addition to their smoothness, the relatively few characteristic X-ray peaks in the EDX spectra of polished silicon and germanium wafers make them wellsuited for conventional SEM-EDX, FIB tomography, and the preparation of particle sections for S/TEM.

■

ASSOCIATED CONTENT

S Supporting Information *

This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Address †

Current affiliation: Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, U.K. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Ian M. Anderson, formerly of NIST, for assistance with TEM. In addition, we thank Gary A. Norris (USEPA) and Constantinos Sioutas (Department of Civil and Environmental Engineering, University of Southern California) for assistance with sample collection.

■

REFERENCES

(1) Fuller, K. A.; Malm, W. C.; Kreidenweis, S. M. J. Geophys. Res.− Atmos. 1999, 104 (D13), 15941−15954. (2) Kalashnikova, O. V.; Sokolik, I. N. J. Quantitative Spectrosc. Rad. Transfer 2004, 87, 137−166. (3) Adachi, K.; Chung, S. H.; Buseck, P. R. Shapes of soot particles and implications for their effects on climate. J. Geophys. Res. 2010, 115 (D15), DOI: 10.1029/2009JD012868. (4) IPCC Climate Change 2007: The Physical Science Basis; Fourth Assessment Report of the Intergovernment Panel on Climate Change (IPCC); Cambridge University Press: New York, 2007. (5) Jacobson, M. Z. Nature 2001, 409 (6821), 695−697. (6) Posfai, M.; Buseck, P. R. Annu. Rev. Earth Planet. Sci. 2010, 38, 17−43. (7) Conny, J. M.; Norris, G. A. Environ. Sci. Technol. 2011, 45, 7380− 7386. (8) Brown, R. C. Air Filtration: An Integrated Approach to the Theory and Applications of Fibrous Filters; Pergamon Press: Oxford, England, 1993. (9) Liu, Z. G.; Wang, P. K. Aerosol Sci. Technol. 1997, 26 (4), 313− 325. (10) Young, R. J.; Moore, M. V. Dual-Beam (FIB-SEM) Systems. In Introduction to Focused Ion Beams: Instrumentation, Theory, Techniques, and Practice; Giannuzzi, L. A., Stevie, F. A., Eds.; Springer Science +Business Media LLC: New York, 2005; pp 247−268. (11) Conny, J. M. Environ. Sci. Technol. 2013, 47 (15), 8575−8581. 9716

dx.doi.org/10.1021/ac5022612 | Anal. Chem. 2014, 86, 9709−9716