. Fig. 23.16  Schematic illustration of orientation movements to optimize EDS collection from a feature of a rough, irregular surface: a initial position gives high absorption due to X-ray path through bulk of specimen; b rotation about a vertical axis brings feature to directly face EDS; c rotation about a horizontal axis places feature perpendicular to beam to minimize backscattering and remote X-ray excitation

b

Better!

Bad!

23.6\ Particle Analysis

23.6.1\ How Do X-ray Measurements

of Particles Differ From Bulk

Measurements?

The analysis of microscopic particles whose dimensions approach or are smaller than the interaction volume in a bulk target of the same composition is subject to effects similar to analyzing rough, bulk surfaces but with additional challenges. Like the rough bulk case, the curved or locally

Best, but still compromised!

tilted surface of a particle acts to modify electron backscattering, which alters X-ray production, while deviations from the ideal flat surface alter the X-ray path to the detector, which modifies X-ray absorption compared to a flat bulk target. The geometry of a particle leads to additional loss of beam electrons due to penetration through the sides and bottom of the particle, as shown in the DTSA-II Monte Carlo simulations in .  Fig. 23.17, which depict trajectories in a 1 μm-diameter aluminum particle at different beam energies. At the highest energy simulated, E0 = 30 keV, most trajectories pass through the particle with some lateral

. Fig. 23.17  DTSA-II Monte Carlo simulation of electron trajectories in a 1-μm-diameter Al sphere on a bulk C substrate at various beam energies; incident beam diameter = 50 nm. Trajectories inside the Al

particle are blue. Green shows trajectories that emerge from the Al particle which change to orange when they enter the C substrate

. Fig. 23.18  DTSA-II Monte Carlo simulation of electron trajectories in Al spheres of various diameters at E0 = 20 keV on a bulk C substrate. Trajectories inside the Al particle are blue. Green shows trajectories that

emerge from the Al particle which change to orange when they enter the C substrate

scattering. As the beam energy is lowered and the electron range decreases, penetration through the bottom and sides diminishes, and at E0 = 5 keV the interaction volume is completely contained within the 1-μm-diameter aluminum particle. The beam penetration effect also depends on the particle size, as shown in .  Fig. 23.18 for particles of various sizes at E0 = 20 keV, and on composition, as shown for particles with a range of atomic numbers in .  Fig. 23.19. Moreover, as opposed to the backscattered electrons in the high angle portion of the cosine distribution which are likely to leave the vicinity of the particle without further interaction, the beam electrons that penetrate through the sides and bottom of the particle are likely to reach the supporting substrate where they will create X-rays of the substrate material that contribute to the overall spectrum measured. This effect can be seen in .  Fig. 23.20, which shows the Al and C peaks calculated by the Monte Carlo simulation for 1-μm diameter Al spherical particles. At a beam energy of E0 = 5 keV, the electron trajectories are contained within the particle which effectively acts like a bulk target. No electrons penetrate the sides or bottom to reach the substrate so there is no C contribution to the EDS spectrum. With increasing incident

energy, electron penetration through the particle into the substrate occurs, and the C peak of the substrate increases relative to the Al-peak from the particle.

23.6.2\ Collecting Optimum Spectra

From Particles

Before meaningful qualitative and quantitative analysis of particles can be attempted, it is important to optimize the EDS spectrum that is collected. As illustrated in the trajectory plots in the Monte Carlo simulations shown in

.  Figs. 23.17, 23.18, and 23.19, and the calculated spectra shown in 23.20, because of the impact of particle size and shape (geometry) on electron interactions, the EDS spectrum of a particle will always be compromised compared to the spectrum of a material of identical composition in the ideal flat, bulk form. The analyst must be aware of the major factors that modify particle spectra and seek to minimize these effects. The particle spectrum can be optimized through careful sample preparation and by understanding of the factors that affect the strategy for beam placement.

. Fig. 23.19  DTSA-II Monte Carlo simulation of electron trajectories at E0 = 20 keV in 1 μm-diameter spheres of various elements (Al, Ti, Cu, Ag, Hf, Au) on a bulk C substrate. Trajectories inside the particles are

blue. Green shows trajectories that emerge from the particle which change to orange when they enter the C substrate

Particle Sample Preparation: Bulk Substrate

ily adhere. If a carbon tape preparation is used, the analyst

must measure a blank spectrum of the tape since the polymer base of the tape frequently contains elements such as oxygen in addition to carbon. If carbon is of analytical interest, other low atomic number substrates are available, including high purity boron and beryllium (but beware of the health hazard of handling beryllium, especially in the form of beryllium oxide which may be released by surface abrasion). Other pure element substrates such as aluminum and silicon can be used if that element(s) is not of interest, but because of the high

degree of excitation of the Al K-L2,3 and Si K-L2,3 characteristic X-rays under typical operating conditions, a significant

fraction of the EDS deadtime will be taken up by the substrate X-rays, diminishing the analytical information collected per unit time on the particle of interest, as well as contributing coincidence peaks that might be misinterpreted as minor or trace constituents, for example, 2 Al K-L2,3 (2.974 keV) is close

to the energy of Ag L3-M4,5 (2.981 keV) and 2 Si K-L2,3 (3.480 keV) is close to the energy of Sn L3-M4,5 (3.440 keV)

When depositing particles on a substrate, the area density should be minimized to avoid situations where electrons scattered off the particle being analyzed can strike nearby particles and excite X-rays, which will then contribute an artifact (“cross talk”). While this remote excitation is likely to be at the equivalent of a minor or trace level constituent, it is critical to understand such contributions when low level constituents are of interest.

23.6 · Particle Analysis

a130000

120000

110000

100000

200 400 600 800 1000 1200 1400 1600 1700 2000 2200 2400 Photon energy (eV)

200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 Photon energy (eV)

Noisy[MC simulation of a 1.000 mm

diameter sphere of Al on C] 30keV

E0 = 30 keV

. Fig. 23.20  EDS spectra calculated with the DTSA-II Monte Carlo simulation for a 1-μm-diameter Al particle on a C substrate at various beam energies: a 5 keV, b 10 keV, c 20 keV, and d 30 keV. Note the

absence of the C peak from the substrate at E0 = 5 keV, its appearance at E0 = 10 keV and the increase relative to the Al peak as the beam energy increases

Counts

K411_sphere-on-tape_20kV10nA

K411_sphere-on-film_20kV10nA

K411

E0 = 20 keV

C – tape substrate

C film (20 nm) on Cu grid

Particle Sample Preparation: Minimizing

Substrate Contributions With a Thin Foil Substrate

As the particle dimensions decrease below 1 μm, electron penetration through the particle increases dramatically, raising the substrate contribution to the X-ray spectrum. Even if the characteristic peak(s) of the substrate material is not of interest and can be ignored, the deadtime due to the substrate spectrum will eventually dominate the overall spectrum measurement, even if the substrate consists of carbon which is relatively poorly excited. Moreover, the contribution of the substrate to the composite spectrum occurs not only from the characteristic peak(s), for example, C K, but also from the continuum background (bremsstrahlung) which affects all photon energies. Increased background from the substrate has the effect of lowering the peak-to-background for all characteristic peaks from the elements of the particle, which

23 degrades all aspects of quantitative analysis but especially impacts the limit of detection, raising the minimum concentration that can be reliably measured.

The substrate contribution to the composite spectrum can be minimized by reducing the mass of the substrate. Fine particles, especially those with nanometer dimensions, can be dispersed by various methods, including air-jetting or deposition from fluid drops, onto thin carbon films, typically 20 nm in thickness, which are supported on a grid (copper, nickel, carbon, etc.), as shown in the sequence of images in .  Fig. 23.22. To further stabilize the particle deposit, it is typical practice to apply a thin (<10 nm) carbon coating to provide conductivity and also to provide some mechanical constraint. The contribution to the spectrum from the thin carbon support film plus the final coating is much reduced compared to the situation on a bulk or thick tape carbon substrate, as shown in the comparison of spectra from K411 particles of similar size (~1 μm in diameter) shown in .  Fig. 23.21. Beam electrons that are scattered laterally from particles can excite the material of the grid, as shown by the presence of copper in the K411 particle spectrum of .  Fig. 23.21, but if this system radiation is problematic, this unwanted spectral contribution can be controlled by choosing alternative grid materials, such as carbon.