Using Energy Resolution to Collect Data for EDS

X-ray detectors for energy dispersive spectroscopy (EDS) have technologically improved since the introduction of the first Silicon Drift-based EDS detector (SDD) 15 years ago.

The first such EDS systems had 5mm² active area with 160-200eV energy resolution and an optimum collection rate reaching 100,000 input counts/second.

Today’s commercially available SDD-based EDS detectors have active areas up to 150mm² per device and feature multiple detectors working in tandem with collection rates of 1 million input counts/second and spectral energy resolution down to 121eV.

The detector performance is oversimplified because energy resolution as measured and reported becomes a key metric today. A more detailed evaluation and a much broader specification are crucial in actual end-user applications. These issues are addressed in this article.

Energy Resolution and the Entire Energy Spectrum

Since energy resolution has reached its theoretical limit, further improvements in energy resolution become important. Other factors or method may influence EDS data collection.

An EDS spectrum of a BN sample with surface contaminated by C and O is depicted in Figure 1, and Figure 2 shows an EDS spectrum of Be sample with surface contaminated by C and O.

Figure 1. EDS spectrum of BN with C and O surface contamination as obtained with a fully evacuated EDS X-ray detector employing an ultra-thin, polymer window (blue) and the simulated spectrum if the detector module were back-filled with inert N₂ (red).

Figure 2. EDS spectrum of Be with C and O surface contamination as obtained with a fully evacuated EDS X-ray detector employing an ultra-thin, polymer window (blue) and the simulated spectrum if the detector module were back-filled with inert N₂ (red).

Very clean and well-separated Be or B peaks and N, C, and O peaks can be observed in these spectra. For these data collections measured at Mn Kα, the EDS detector used an energy resolution of 122eV. In both spectra, the energy resolution of the C peak is 39eV.

Only half of the improvement needed to generate the spectra described above is represented by the sophisticated SDD. Attenuation of low-energy X-rays is a major challenge in low energy X-ray analysis owing to the fact that these X-rays are attenuated by various sources:

• Low energy X-rays are absorbed by the window employed for isolating the SDD crystal from the microscope vacuum (or ambient while vented). Conventional Be windows would not transmit X-rays lower than the Na Ka line (~1keV) and must therefore be eliminated as an option.
• Low energy X-rays are absorbed by the inert gas present between the window and the SDD crystal. Inert N₂ gas is used in most designs to back-fill the volume between the light-element window and the SDD crystal. It is possible to simulate the deterioration of low-energy sensitivity by inert N₂ gas through the application of the X-ray absorption curves for N₂ gas across the distance that must be travelled by the X-rays the N₂ gas. Figures 1 and 2 show these simulated spectra. The variation in low energy sensitivity is stark when compared against a detector whose volume is fully evacuated rather than back-filled with N₂.
• Low energy X-rays are absorbed by the sample itself when they are produced deeper than the average escape depth for that energy. For instance, the escape depth of Li is only few tens of nm so that it is extremely difficult to detect Li with any EDS detector even in the presence of trace amounts of surface contamination.

Therefore, the use of an ultra-thin, polymer window (<300nm) or the elimination of the window altogether is preferred to effectively achieve low energy detection.

In the case of detection below 300eV, the volume between the SDD module and the thin window should not be backfilled with N₂ gas. The ideal option is a completely windowless detector or a fully evacuated detector module.

Analysts must rely predominantly on analysis of the top 10-50nm of the sample due to the absorption of low energy X-rays by the sample itself. This implies that the sample surface has to be carefully prepared and preserved and the SEM needs to be operated at a shallow penetration depth (<5kV) to prevent dilution of the low energy analysis of the surface region with the high-energy X-rays produced from the overall bulk of the sample.

Energy Resolution and Count Rate

Although an energy resolution of 121eV that extends the Bremsstrahlung limit is remarkable, it is hardly ever observed during normal operation. This energy resolution is generally specified at less than 5000 counts per second of input count rate.

It is possible to employ a peaking time such as 6.4ms at these low rates. Shorter peaking times are prerequisites for high count rate acquisition, but lead to a greater statistical uncertainty in the energy of the acquired X-ray. This means a lower energy resolution.

The optimal peaking times and corresponding effect on energy resolution needed for spectral acquisition at an optimum dead time of 50% are presented in Figure 3. In this graph, the energy resolution varies from 125eV at 6.4ms up to 175eV at 0.2ms.

Figure 3. Energy resolution as a function of input count rates at 50% dead time (output count rate is approximately half of input count rate)

Using an SDD, most mapping applications take place at an output count rate of a few hundred thousand counts per second. The ratio of input and output count rates is 2 at a dead time of 50%. At higher dead times, this ratio will be greater. Mapping at 200,000 input counts/ second (100,000 output counts/second) leads to a resolution deterioration of approximately 8eV at a 1 microsecond time constant.

This deterioration is prevented by forced operation at a long time constant, which, however, leads to very slow acquisition rates (few thousand output counts per second) when the very high dead time limits overall throughput.

This scenario is not suitable for mapping. A de facto advantage in energy resolution can be achieved with a system that enables multiple peaking times rather than only 2 or 3 peaking times due to the possibility of automatic selection of the longest possible peaking time for any given input count rate.

Certain designs may exhibit 121eV energy resolution at <5000 input counts per second, but only reach 140-150eV energy resolution in mapping mode at output counts beyond 100,000 per second. Hence, a broader specification for energy resolution, including both high- and low-throughput requirements, is helpful to the users.

Energy Resolution and Post-Processing Algorithms

The actual spacing of the X-ray lines produced by the elements within the sample itself is another concern associated with improved energy resolution. While the value of an improved EDS detector is demonstrated by the spectra in Figures 1 and 2, elements Be, B, C, N, and O lead to single X-ray lines within the spectrum that are isolated by 150-250eV.

The EDS spectrum of galena sample composed principally of Pb and S is shown in Figure 4, where the S K-line (2.307keV) is separated from the principal Pb M-line (2.346 keV) by only 39eV. In addition, multiple Pb M-lines can also be observed. Two different EDS detectors with an energy resolution of 138eV and 122eV, respectively, were used for the sample analysis.

Figure 4. Energy-dispersive X-ray spectrum of galena (primarily Pb and S) for: (a) an SDD at 138eV (circa 2002), (b) an SDD at 122eV (circa 2012), and (c) a MagnaRay WDS spectrometer.

A Thermo Scientific™ MagnaRay™ wavelength dispersive spectrometer (WDS) (Figure 5) was finally used for the sample analysis.

The 122eV detector spectrum shows improvement compared with that of the 138eV detector. However, there is significant overlapping of the Pb and S peaks in spite of the 16eV improvement in energy resolution and the elemental X-ray lines are not clearly discerned. It is necessary to have an energy resolution below 40eV to attain serious separation of the peaks.

The S K-line and the Pb M-lines can be accurately resolved into separate peaks by a WDS, thanks to its superior energy resolution. Even if the best EDS detector is used, overlapping peaks will be there if closely spaced X-ray lines are present continuously, thus preventing direct analysis of unknown specimen elements and posing challenges to qualitative and quantitative analysis and effective element mapping. However, these challenges can be addressed with the appropriate application of post-processing algorithms.

The element maps of the galena sample plotted with "gross" X-ray counts (i.e., counts that were not corrected after collection) are presented in Figure 5a. The WDS element map for Pb and for F is also given for reference. The elements determined as present by the EDS spectrum are Pb, O, F, S, Ca, Mn, Cu, As and Sb. Within the map, three regions can be observed:

• Phase 1 at middle left and middle right comprised mainly of Ca and F
• Phase 2 at center composed mainly of Cu and S
• Phase 3 - a matrix material of Pb and S

The mapped backgrounds are quite high for a number of elements such as O, Mn, Cu, and As, indicating a consistent, low-level distribution of these elements across the sample or consistent misidentification of the elements owing to overlapping of peaks within the spectrum.

Phase 2 also contains "ghost-like" regions of Ca and Sb, which might be dismissed as artifacts. However, this is a risky assumption. Moreover, there is no matching between the EDS map for Pb and the WDS map for the Pb M-α line, indicating the presence of errors within the EDS mapping results due to peak overlaps.

The same element maps generated utilizing the exact same raw data are presented in Figure 5b. However, they are "quantitative elemental mapping" or "Quant maps" involving the application of peak deconvolution, a background subtraction algorithm, and a matrix correction to the acquired spectra in each pixel. The direct comparison of the element maps in Figures 5a and 5b clearly shows the impact of these Quant maps.

Figure 5. (a) X-ray element maps of galena (PbS) sample exhibiting “gross” or uncorrected X-ray counts, (b) same X-ray element maps of (a) but with corrections for peak deconvolution and background subtraction

The overall image noise around the concentrations of several elements, especially O, Mn and Cu, falls to near zero or zero other than in the actual phase wherein it exists. The As counts go to zero everywhere, showing that the As that originally determined by the As K-line at 10.532keV is an overlap with the Pb L-lines at 10.549eV.

The EDS identification of Pb in the central Cu-S region (Phase 2) is removed, bringing the wt% map of Pb in direct alignment with the WDS map of Pb. S is still present in the central region and the matrix region around it, thus helping to determine a and S containing region for Phase 2 surrounded by a Pb-S, or galena, matrix material in Phase 3.

Regarding the Ca and Sb "ghost-like" distributions in Phase 2, the Ca distribution observed in the Cu-S region has vanished totally. This leads to the isolation of the Ca and F to Phase 1. This region is identified as fluorite (CaF₂).

The presence of Sb, which is absent in Phase 1, in the Cu-S Phase 2 region is confirmed by the Quant map. This discards any assumption that the Sb located in this region is an artifact. The phase type is changed from copper sulfide to chal-costibite (CuSbS₂) due to the presence of Sb in Phase 2.

This mapping exercise determined six major errors in a single sample when analyzed using basic mapping techniques and an advanced EDS detector with 122eV resolution. Although EDS detectors have undergone many significant advances in terms of energy resolution over the past decade, it is clear that until energy resolution approaches the level of WDS, the role of efficient post-processing algorithms for enhancing element mapping will be more significant than any further enhancements in energy resolution.

Conclusion

Energy resolution is one of the key factors in acquiring high-quality EDS data. Windowless or thin window technologies and the removal of any inert gas between the SDD crystal and thin window are crucial to low energy detection. Low capacitance SDDs and fast electronics are important for mapping applications to mitigate the energy resolution degradation, which takes place at output counts rates of a few hundred thousand per second.

While aesthetically pleasing spectra are generated with an extreme EDS energy resolution, the most robust lever short of full WDS element mapping) for generating world-class, accurate EDS elemental maps is provided by powerful post-processing algorithms.

Although EDS detectors have witnessed significant developments over the past 15 years, the energy resolution specification now reaches the limits of physics. Other advances in EDS now focus on the improvement of EDS data collection and analysis.

This information has been sourced, reviewed and adapted from materials provided by Thermo Fisher Scientific – Materials & Structural Analysis.

For more information on this source, please visit Thermo Fisher Scientific – Materials & Structural Analysis.