Position-tagged spectrometry: a new approach for EDS spectrum imaging

Abstract

Graphical display of the spatial distribution of elements in a specimen has been recognized as a powerful technique since the earliest days of electron-beam x-ray microanalysis. With recent advances in computing power and mass storage, it has become practical to save complete spectra at each pixel rather than simple window counts, providing great flexibility for analytical post-processing. The term "spectrum imaging" has been coined to describe such a data structure.Hunt and Williams give a concise summary of the advantages of spectrum imaging, including the ability to map an element which was not known to be present when the data were acquired, and the ability to integrate spectra from image regions of arbitrary shape to improve statistical precision. Quantitative analysis may also be applied at each pixel to produce a concentration map, which is more informative than the raw or perhaps background-subtracted window counts usually displayed. Sensitive discrimination of phases is possible using various fitting techniques on the pixel spectra, matching them against each other or against standards.

Tutorial Review: X-ray Mapping in Electron-Beam Instruments

“…They streamed the x, y, and energy data to magnetic tape in order of arrival time and later sorted them into a spectrum image. Application of this idea to X-ray mapping in electron beam instruments was developed by Mott et al~Mott et al, 1995;Mott & Friel, 1999!. In this continuous scanning method, dwell times of only a few microseconds are used, and acquisition at multiple frames per second is possible. At a pixel resolution of 256 ϫ 256, maps can be collected at several frames per second.…”

Automated Analysis of SEM X-Ray Spectral Images: A Powerful New Microanalysis Tool

“…More recently, with increases in computer CPU speed, memory, and storage space, X-ray spectrum-imaging systems have become available. A spectral image is a two-dimensional array of points in the microstructure with a complete X-ray spectrum from each point~Legge and Hammond, 1979;Jeanguillaume and Colliex, 1989;Mott et al, 1995;Anderson, 1998!. Spectral images potentially overcome the shortcomings of point analyses~e.g., how to pick the points! by being an imaging technique that allows the analyst to qualitatively analyze and perhaps quantify the spectra.…”

“…The problems with spectrum imaging up until recently were that there were no tools for extracting the chemically relevant information from the massive amount of data~.65 million individual data points for a 256 ϫ 256-pixel ϫ 1024-channel spectral image!. The tools that have been available for analyzing spectral images amount to ones that map after the fact and then allow summing of spectra that have been thresholded from maps~Mott et al., 1995;Mott and Friel, 1999!. Additionally, it can take a considerable amount of time to collect spectral images with individual spectra containing sufficient counts to perform even a qualitative, let alone quantitative, analysis of a given spectrum.…”

Diluvian Clustering: A Fast, Effective Algorithm for Clustering Compositional and Other Data

“…The most popular techniques for analyzing spectrum images and other large compositional data sets include visualization tools like X-ray spectrum images (Mott et al, 1995), statistical tools like Sandia’s AXSIA (software application) (Kotula et al, 2003) (commercially available as Thermo Scientific’s COMPASS tool) and the expectation–maximization algorithm (Dempster et al, 1977), hierarchical clustering tools (Wilson et al, 2012), nonhierarchical clustering tools like k -means (MacQueen, 1967) and supervised learning models like support vector machines (Cortes & Vapnik, 1995). Each of these techniques has strengths and weaknesses and have been used with varying degrees of success by this author and others.…”