Quantitative trace element imaging using PIXE and the nuclear microprobe

Abstract

ABSTRACT:The X-ray spectra of pure elements, excited using MeV energy beam of protons from the nuclear microprobe, have known spectra signatures. This makes X-ray spectra for more complex mixtures amenable to decomposition into contributions from the component elements. By devising this procedure as a matrix operation that transforms directly from spectrum vector to elemental concentration vector, the decomposition can be performed very efficiently enabling the real-time projection of the component element signals. In the case of a raster-scanned beam, with data that contain position information for each X-ray event, this approach enables the real-time projection of component element spatial distribution images. This paper describes the matrix transform approach called dynamic analysis (DA), which enables on-line real-time imaging of major and trace elements using proton-induced X-ray emission (PIXE). The method also provides off-line iterative yield corrections to these images to compensate for changing sample composition across an image area. The resulting images are quantitative in two respects: (1) they resolve the pure element components and strongly reject interferences from other elements and (2) they can be directly interrogated for sample composition at each pixel, over areas, or along lines across the image area, with accuracy comparable to microanalytical point analysis methods. The paper describes the DA method, presents tests, and discusses its application to quantitative major and trace element imaging in geology.

Maia X-ray fluorescence imaging: Capturing detail in complex natural samples

“…The full non-linear fit is performed on the SXRF spectrum at the top beam energy for the XANES scan, in which the shape of the background and the inelastic scattering distributions are determined. Then a series of DA matrices are generated as the beam energy is reduced according to the method of Ryan et al [12,14] with the elastic and inelastic peaks shifted to track the change in beam energy. During processing of the event stream, the lookup table operation first selects the correct DA matrix for the beam energy and then selects the matrix column based on event X-ray energy to increment a 2D result array indexed by XY position repeated over beam energy.…”

“…In matrix form, these equations can be rearranged into the form of element intensities given by a matrix transformation of the spectrum of interest [11]. Using a fundamental parameters approach to SXRF analysis, the element intensities can be related to concentration, which results in a matrix transformation directly from input spectrum to element concentration vector [12]. In the Maia case, the model includes the effects of the Mo mask on individual detector solid-angles and the changing spectral relative intensities across the array, as an extension of the layered sample model in the GeoPIXE software [5,13,14].…”

“…However, this approach also can be used to map each event in turn onto its element concentration contributions [17]. Some of these contributions are negative, which is how overlap and background subtraction are accomplished [12]. In the Maia detector processor, these contributions are accumulated for all events in the current image pixel in what reduces to a simple lookup-accumulate operation implemented in the HYMOD FPGA [14].…”

PIXE Real-Time Quantitative Image Projection Applied to Synchrotron XRF Imaging using the X-ray Fluorescence Microprobe

“…A method called Dynamic Analysis (DA), developed at the CSIRO, builds a matrix transform to unfold the contributions of overlapping elements, rapidly performing the effect of a linear least squares fit to each pixel spectrum [1]. Projection of companion variance images aid in tracking uncertainty contributions.…”

“…Fundamental parameter yields calculated for these components are then combined to estimate improved concentration images. This procedure is repeated to converge on self-consistent elemental concentrations for each pixel; initial concentration errors of >50% can be reduced to <3% in ~3 iterations (as compared with electron microprobe point analyses; [1]). …”