Quantification of electron-exited X-ray spectra following the standards-based “k-ratio” (unknown/standard intensity) protocol with corrections for “matrix effects” (electron energy loss and backscattering, X-ray absorption, and secondary X-ray fluorescence) is a well-established method with a record of rigorous testing and extensive experience. Two recent studies by Gopon et al. working in the Fe–Si system and Llovet et al. working in the Ni–Si system have renewed interest in studying the accuracy of measurements made using L-shell X-ray peaks. Both have reported unexpectedly large deviations in analytical accuracy when analyzing intermetallic compounds when using the low photon energy Fe or Ni L-shell X-ray peaks with pure element standards and wavelength-dispersive X-ray spectrometry. This study confirms those observations on the Ni-based intermetallic compounds using energy-dispersive X-ray spectrometry and extends the study of analysis with low photon energy L-shell peaks to a wide range of elements, Ti to Se. Within this range of elements, anomalies in analytical accuracy have been found for Fe, Co, and Ge in addition to Ni. For these elements, the use of compound standards instead of pure elements usually resulted in significantly improved analytical accuracy. However, compound standards do not always provide satisfactory accuracy as is demonstrated for L-shell peak analysis in the Fe–S system: FeS and FeS2 unexpectedly do not provide good accuracy when used as mutual standards.

This, the second in a series of articles present a new framework for considering the computation of uncertainty in electron excited X-ray microanalysis measurements, will discuss matrix correction. The framework presented in the first article will be applied to the matrix correction model called “Pouchou and Pichoir's Simplified Model” or simply “XPP.” This uncertainty calculation will consider the influence of beam energy, take-off angle, mass absorption coefficient, surface roughness, and other parameters. Since uncertainty calculations and measurement optimization are so intimately related, it also provides a starting point for optimizing accuracy through choice of measurement design.

2018 marked the 50th anniversary of the introduction of energy dispersive X-ray spectrometry (EDS) with semiconductor detectors to electron-excited X-ray microanalysis. Initially useful for qualitative analysis, EDS has developed into a fully quantitative analytical tool that can match wavelength dispersive spectrometry for accuracy in the determination of major (mass concentration C > 0.1) and minor (0.01 ≤ C ≤ 0.1) constituents, and useful accuracy can extend well into the trace (0.001 < C < 0.01) constituent range even when severe peak interference occurs. Accurate analysis is possible for low atomic number elements (B, C, N, O, and F), and at low beam energy, which can optimize lateral and depth spatial resolution. By recording a full EDS spectrum at each picture element of a scan, comprehensive quantitative compositional mapping can also be performed.

Scanning electron microscopy with energy-dispersive spectrometry has been applied to the analysis of various materials at low-incident beam energies, E0≤5 keV, using peak fitting and following the measured standards/matrix corrections protocol embedded in the National Institute of Standards and Technology Desktop Spectrum Analyzer-II analytical software engine. Low beam energy analysis provides improved spatial resolution laterally and in-depth. The lower beam energy restricts the atomic shells that can be ionized, reducing the number of X-ray peak families available to the analyst. At E0=5 keV, all elements of the periodic table except H and He can be measured. As the beam energy is reduced below 5 keV, elements become inaccessible due to lack of excitation of useful characteristic X-ray peaks. The shallow sampling depth of low beam energy microanalysis makes the technique more sensitive to surface compositional modification due to formation of oxides and other reaction layers. Accurate and precise analysis is possible with the use of appropriate standards and by accumulating high count spectra of unknowns and standards (>1 million counts integrated from 0.1 keV to E0).

The accuracy and precision of X-ray intensity measurements with a silicon drift detector (SDD) are compared with the same measurements performed on a wavelength dispersive spectrometer (WDS) for a variety of elements in a variety of materials. In cases of major (>0.10 mass fraction) and minor (>0.01 mass fraction) elements, the SDD is demonstrated to perform as well or better than the WDS. This is demonstrated both for simple cases in which the spectral peaks do not interfere (SRM-481, SRM-482, and SRM-479a), and for more difficult cases in which the spectral peaks have significant interferences (the Ba L/Ti K lines in a series of Ba/Ti glasses and minerals). We demonstrate that even in the case of significant interference high count SDD spectra are capable of accurately measuring Ti in glasses with Ba:Ti mass fraction ratios from 2.7:1 to 23.8:1. The results suggest that for many measurements wavelength spectrometry can be replaced with an SDD with improved accuracy and precision.

This article is the fourth in a series describing the setup and use of NIST DTSA-II. The previous articles may be viewed in the Microscopy Today archives at www.microscopy-today.com. Part I of this article described how to use DTSA-II to quantify an energy dispersive X-ray spectrum from a material of unknown composition. That presentation covered each step and provided background material but lacked a concrete example. This follow-up article provides two such examples.

Artifacts are the nemesis of trace element analysis in electron-excited energy dispersive X-ray spectrometry. Peaks that result from nonideal behavior in the detector or sample can fool even an experienced microanalyst into believing that they have trace amounts of an element that is not present. Many artifacts, such as the Si escape peak, absorption edges, and coincidence peaks, can be traced to the detector. Others, such as secondary fluorescence peaks and scatter peaks, can be traced to the sample. We have identified a new sample-dependent artifact that we attribute to Compton scattering of energetic X-rays generated in a small feature and subsequently scattered from a low atomic number matrix. It seems likely that this artifact has not previously been reported because it only occurs under specific conditions and represents a relatively small signal. However, with the advent of silicon drift detectors and their utility for trace element analysis, we anticipate that more people will observe it and possibly misidentify it. Though small, the artifact is not inconsequential. Under some conditions, it is possible to mistakenly identify the Compton scatter artifact as approximately 1% of an element that is not present.

Quantifying an X-ray spectrum is the process of converting a measured spectrum into an estimate of the composition of the material from which the spectrum was collected. In a certain sense, interpreting X-ray spectra is very simple. A spectrum from even the most complex material can be thought of as a sum of spectra from the constituent elements (see Figure 1). To first approximation, if you know the spectrum for the constituent pure elements, you can estimate the spectrum for the complex material. Regardless of the software vendor or the name of the algorithm, this is the basis for quantification of all X-ray spectra.

X-ray elemental mapping and X-ray spectrum imaging are powerful microanalytical tools. However, their scope is often limited spatially by the raster area of a scanning electron microscope or microprobe. Limited sampling size becomes a significant issue when large area (>10 cm2), heterogeneous materials such as concrete samples or others must be examined. In such specimens, macro-scale structures, inclusions, and concentration gradients are often of interest, yet microbeam methods are insufficient or at least inefficient for analyzing them. Such requirements largely exclude the samples of interest presented in this article from electron probe microanalysis. Micro X-ray fluorescence–X-ray spectrum imaging (μXRF-XSI) provides a solution to the problem of macro-scale X-ray imaging through an X-ray excitation source, which can be used to analyze a variety of large specimens without many of the limitations found in electron-excitation sources. Using a mid-sized beam coupled with an X-ray excitation source has a number of advantages, such as the ability to work at atmospheric pressure and lower limits of detection owing to the absence of electron-induced bremsstrahlung. μXRF-XSI also acts as a complement, where applicable, to electron microbeam X-ray output, highlighting areas of interest for follow-up microanalysis at a finer length scale.

A high quality X-ray spectrum image of a 3.3 μm diameter sphere of K411 glass resting on a copper substrate was collected at 25 keV. The same sample configuration was modeled using the NISTMonte Monte Carlo simulation of electron and X-ray transport as is integrated into the quantitative X-ray microanalysis software package DTSA-II. The distribution of measured and simulated X-ray intensity compare favorably for all the major lines present in the spectra. The simulation is further examined to investigate the influence of angle-of-incidence, sample thickness, and sample diameter on the generated and measured X-ray intensity. The distribution of generated X-rays is seen to deviate significantly from a naive model which assumes that the distribution of generated X-rays is similar to bulk within the volume they share in common. It is demonstrated that the angle at which the electron beam strikes the sample has nonnegligible consequences. It is also demonstrated that within the volume that the bulk and particle share in common that electrons, which have exited and later reentered the particle volume, generate a significant fraction of the X-rays. Any general model of X-ray generation in particles must take into account the lateral spread of the scattered electron beam.

Bulk silicon-germanium (SiGe) alloys and two SiGe thick films (4 and 5 μm) on Si wafers were tested with the electron probe microanalyzer (EPMA) using wavelength dispersive spectrometers (WDS) for heterogeneity and composition for use as reference materials needed by the microelectronics industry. One alloy with a nominal composition of Si0.86Ge0.14 and the two thick films with nominal compositions of Si0.90Ge0.10 and Si0.75Ge0.25 on Si, evaluated for micro- and macroheterogeneity, will make good microanalysis reference materials with an overall expanded heterogeneity uncertainty of 1.1% relative or less for Ge. The bulk Ge composition in the Si0.86Ge0.14 alloy was determined to be 30.228% mass fraction Ge with an expanded uncertainty of the mean of 0.195% mass fraction. The thick films were quantified with WDS-EPMA using both the Si0.86Ge0.14 alloy and element wafers as reference materials. The Ge concentration was determined to be 22.80% mass fraction with an expanded uncertainty of the mean of 0.12% mass fraction for the Si0.90Ge0.10 wafer and 43.66% mass fraction for the Si0.75Ge0.25 wafer with an expanded uncertainty of the mean of 0.25% mass fraction. The two thick SiGe films will be issued as National Institute of Standards and Technology Reference Materials (RM 8905).

Spectrum simulation is a useful practical and pedagogical tool. Particularly with complex samples or trace constituents, a simulation can help to understand the limits of the technique and the instrument parameters for the optimal measurement. DTSA-II, software for electron probe microanalysis, provides both easy to use and flexible tools for simulating common and less common sample geometries and materials. Analytical models based on ϕ(ρz) curves provide quick simulations of simple samples. Monte Carlo models based on electron and X-ray transport provide more sophisticated models of arbitrarily complex samples. DTSA-II provides a broad range of simulation tools in a framework with many different interchangeable physical models. In addition, DTSA-II provides tools for visualizing, comparing, manipulating, and quantifying simulated and measured spectra.