Low Electron Beam Energy X-ray Microanalysis: The Adventure Continues!

The “adventure” that awaits the user of electron-excited X-ray microanalysis with energy dispersive spectrometry (EDS) when performed in the low beam energy regime consists of several challenges [1, 2]: (1) The upper limit of the low beam energy range is E0=5 keV, which is the lowest beam energy for which all elements of the Periodic Table can be effectively measured, excepting H and He which do not produce a characteristic X-ray. Even at this beam energy, certain elements must be measured with unfamiliar X-ray peaks, e.g., the Ti L-family instead of the Ti K-family and the Ba M-family instead of the Ba L-family. As the beam energy is reduced below 5 keV, progressively more elements become inaccessible to analysis even when present as major constituents (mass concentration > 0.1) due to inadequate or no excitation. (2) The fluorescence yield for low energy L-, M- and N- shell X-rays is low, reducing sensitivity. (3) Low photon energy peaks such as Ti L and Ba M suffer significant interference from K-shell X-rays of the low atomic number elements Li to F. Accurate intensity measurements of interfering peaks depend on implementing a robust peak deconvolution procedure. (4) The shallow depth of excitation at low beam energy increases the influence of a surface layer, such as a native oxide or an applied conductive coating, upon the analytical results. Figure 1 shows an example of the influence of an oxidized surface layer on AlN, which becomes apparent as the beam energy is lowered and the O and C K-peaks are observed to increase relative to the N K-peak. (5) Sample preparation to expose unmodified material thus becomes a critical issue for successful analysis. How accurate is analysis performed at low beam energy? To answer this question, a large suite of materials of known composition (including Certified Reference Materials (CRM) and materials of well-defined stoichiometry, e.g, FeS2, TiSi2, GeTe ) was analyzed following the standards-based protocol using the NIST DTSA-II software platform for spectrum processing and matrix correction [3]. The materials were prepared to minimize surface layer effects. Measured spectra were carefully inspected to exclude those with unexpected contamination, and multiple measurement locations were compared to avoid local defects. The figure of merit that has been used to assess the accuracy is the Relative Deviation from Expected Value (RDEV), which is defined as: where the ”expected value” is the certified composition for CRMs or the stoichiometric composition for other materials tested. Figure 2 shows the histogram of RDEV values (bin width = 1% RDEV) for the atomic concentrations determined for 179 analyzed constituents in 75 materials, The distribution of RDEV values is such that all measured concentrations but one fall within ±5% RDEV. 82% of the results fall in the range -2% to 2% RDEV. (Detailed results are provided in the ancillary material [4].) Comparison of AlN spectra measured at E0 = 15, 10, and 5 keV; scaled to the N K-peak. Histogram of RDEV values for standards-based analysis using DTSA-II at E0 = 5 keV; the bins are unity in width centered on each integer, e.g., the 0 bin collects results from -0.5% to +0.5% RDEV.