The Challenges of Standardless Quantification of Light Elements by EDS-XRF

Quantification of light elements (Z < 11) by energy dispersive X-ray fluorescence (EDS-XRF) is attractive for many materials workflows because it provides broad elemental coverage with minimal sample preparation and fully non-destructive analysis. In practice, however, standardless fundamental parameters (FP) quantification of light elements, and especially oxygen, remains one of the most challenging regimes of XRF microanalysis. The difficulty arises from a combination of strong low-energy X-ray attenuation, spectral congestion near the C-N-O Kα region, and limitations in the physical data and models underlying matrix corrections. Understanding how these effects manifest in real measurements is essential for defining realistic expectations and for identifying acquisition and modeling strategies that mitigate bias. Low-energy characteristic X-rays suffer severe absorption along the excitation and detection path. For most EDS detectors, the entrance window represents a dominant loss mechanism for C, N, and O Kα photons. Use of a silicon nitride detector window can provide substantially higher transmission at sub-keV energies than conventional beryllium or polymer windows commonly employed in EDS-based XRF systems. This improvement yields higher detected intensities and better counting statistics for light-element lines, directly benefiting peak detection and spectral fitting. Windowless detector configurations can further increase low-energy transmission, but they introduce additional practical and stability constraints that limit their widespread implementation. Nevertheless, increased transmission alone does not resolve the more fundamental difficulties associated with spectral overlaps and FP modeling. Spectral congestion is a major obstacle for the EDS measurement of light elements. When using a Rh anode X-ray source, several Rh M lines occur in the same energy region as light-element K lines. In addition, many materials exhibit low-energy L lines from medium-Z elements that further complicate deconvolution. A clear example is provided by calcite (CaCO3) in Fig. 1, where the O Kα line is strongly overlapped by the Rh Mγ and M2N4 lines. Although O peaks are readily detectable with the silicon nitride window (used to record this spectrum), extracted net intensities depend sensitively on how overlapping contributions are modeled. The use of filters placed after the X-ray tube can attenuate the Rh M line intensities, reducing this interference, but also alters the excitation spectrum in ways that must be consistently incorporated into FP calculations. C presents a similar problem where Rh M lines and Ca L lines overlap the C Kα region, producing apparent excess intensity if not properly accounted for. Even with improved low-energy transmission, spectral fitting must separate multiple partially overlapping peaks with very different intensities. The lower resolution of the EDS technique compared to other spectroscopic techniques (WDS, and SXES) makes this deconvolution challenging. In addition, low-energy X-rays are subject to chemical shifts making the constraints on the fitting more difficult. Beyond spectral issues, FP matrix corrections become less accurate below ∼1 keV because of high uncertainties in atomic parameter values [1]. Absorption corrections dominate at these energies, and uncertainties in mass attenuation coefficients (MACs) propagate into calculated concentrations. In addition to primary fluorescence (X-ray tube spectrum exciting characteristic lines) and secondary fluorescence (interelement excitation by characteristic lines), electron-mediated enhancement mechanisms can contribute to light-element X-ray production in a non-negligible way [2]. Absorption of primary photons produces photoelectrons with sufficient energy to ionize K shells of light elements in their vicinity. In addition, Auger emission, which dominates de-excitation for light elements, can generate additional electrons capable of inducing ionization. These photoelectron and Auger-electron processes depend on electron ranges, stopping powers, and local composition, making their incorporation into FP expressions computationally intensive. Omission of these effects can lead to inaccurate elemental quantification results, particularly in oxygen-rich matrices such as oxides and glasses. When all these effects are taken into account, good quantification numbers can be obtained as illustrated on a CaCO3 sample (Table 1). Measurements were performed on an EDAX Orbis II Super micro-XRF system equipped with a silicon nitride detector window, using a Rh X-ray source operating at 30 kV and tube current of 1 mA, and with an acquisition live time of 30 s. While major cations are quantified reproducibly, small variations in the recorded number of counts for the light elements can lead to major elemental quantification changes. When oxygen is the only light element of interest, an alternative approach is to compute oxygen by stoichiometry from measured cations and to include this oxygen explicitly inside the matrix correction iterations rather than as post-processing [3]. In this method, the matrix correction is iterated with oxygen fixed by stoichiometry based on known valences and phase chemistry so that absorption and enhancement terms are evaluated taking all the elements present into consideration. The results presented here will demonstrate both the capabilities and the limits of modern standardless EDS-XRF for light-element analysis. Silicon nitride detector windows provide a genuine improvement in low-energy sensitivity and enable practical detection of C, O and F. However, spectral overlap and fundamental uncertainties in low-energy FP modeling remain the dominant constraints on accuracy. By combining appropriate acquisition strategies with stoichiometric constraints when applicable, more reliable quantitative results can be achieved. At the same time, the data underscore the need for continued improvements in low-energy physical databases and modeling of electron excitation processes to further advance standardless light-element quantification in EDS-XRF. EDS-XRF standardless quantification of a CaCO3 sample at 30kV and 30 seconds live time. Fitting of a calcite spectrum using Gaussian functions. The spectrum was acquired using a Rh-anode X-ray tube operated at 50 kV and with a live time of 90 s. The C Kα and O Kα X-ray lines are heavily overlapped with Rh M lines and Ca L lines.