Quantitative Analysis with ESPRIT

Precision and accuracy of standardless and standard-based quantification are some of the most discussed issues in quantitative EDS analysis. This application example using a stainless steel focuses on possibilities, potential pitfalls and the quality of results.

Two critical factors influence the accuracy of the quantification decisively: the correctness of element identification and the accuracy of peak deconvolution. Stainless steel may contains a number of elements in a concentration of well below 1 %. These are easily overlooked. Peak overlaps are quite common in a steel sample, especially in the mid atomic number range (24–28), where Kß lines of lighter elements overlap with the Kα line of the next heavier element. This may also lead to errors, if the heavier element has a low concentration.

Conclusions during the course of analyses:

• Automatic peak identification often leads to unsatisfactory results regarding the low concentration of elements. In this case ESPRIT's powerful deconvolution tool helps to avoid overlooking minor constituents.
• While standardless analysis already provides very good results, the following approach is suggested for highest accuracy:
  a) Initial standardless quantification
  b) Use of results for standard selection
  c) Subsequent standard-based analysis
  The unique hybrid quantification of ESPRIT can be used, which combines the true standardless analysis of elements with the standard-based analysis of elements for which reference standards are available.
• The results of ESPRIT's standardless P/B-ZAF analysis can be further optimized through the use of references. This is especially useful when analyzing rough samples.

TQuant is the light element / low energy quantification routine that is now part of ESPRIT's standardless quantification software, here it is applied to the analysis of boron nitride at a range of different acceleration voltages. Boron nitride (BN), with an element ratio of 50:50 atomic%, displays 2 peaks, the B-K peak is at 183 eV and the N-K peak  is at 392 eV. Both are in the very low energy range, which means that certain factors influencing peak intensities have to be considered:

• Efficiency of the detector
• Absorption of generated radiation within the sample
• Overvoltage (Acceleration voltage of the SEM, which should be around 2.5 to 3 times the energy of the peak of interest in keV)

In summary this means that there is no way to judge the element concentrations in a light element spectrum by the peak heights. An adapted quantification routine like TQuant is needed to correctly determine these concentrations, also at different voltages:

This application example focuses on the analytical problem of standardless quantification in the low energy range. The challenge lies in absorption effects and in difficulties in background calculation due to high absorption edges and according to statistical error. The advanced light element quantification TQuant was used to solve this challenging problem.

The sample investigated is a sintered hard ceramic material, mainly composed of titanium di-boride (TiB₂), titanium carbide (TiC), silicon carbide (SiC) and a number of minor constituents. A polished but uncoated section of this material was analyzed.

The task was to quantify the area spectra of the analysis locations shown in the figure above. The spectra were extracted from regions of same composition of a HyperMap that was acquired from the sample. The spectra of 16 pixels were combined to improve statistics, even so they contain only 5,000 to 9,000 counts, corresponding to less than 0.5 s measurement time. Five of these area spectra were quantified. The average values and deviations are listed below. "s" is the standard deviation, "Deviation" is the relative deviation, the difference between the quantification result and the expected stoichiometric value:

The findings show that TQuant provides reliable standardless quantification results, even under the adverse measurement conditions described in this example.