Correlating whole sample EDS and Raman mappings - A case study of a Chelyabinsk meteorite fragment

Abstract

Whole sample microscopy mappings are of interest in many cases as they provide analytical information of phases varying in size by orders of magnitude and in composition across the sample. These benefits are amplified if more than one microscopic technique is used for the mappings. However, to take full advantage of correlative whole sample mappings, the data of each technique has to be carefully prepared, treated, correlated and evaluated. With this work, we want to present the key steps of our data treatment approach as well as the results on an exemplary sample, the Chelyabinsk meteorite. The most important step in our data treatment approach is to start by evaluating the spectral maps separately as far as possible (at-% quantification for EDS for example) and then generate pseudo spectral maps from this evaluation in the form of image stacks. This allows us to preserve the advantages of specialized software packages and standard work flows for every spectral mapping, whilst also unifying the data format and compressing the data sufficiently for correlation and the application of machine learning tools. We have performed whole sample mappings using SEM, EDS and Raman on a cross-section of a Chelyabinsk meteorite fragment, roughly 1.0cm × 0.8cm large. Combining these mappings into a single "super" spectral map, we were able to produce a uniquely detailed mapping of the composition of the meteorite fragment, as well as perform a quantitative analysis of the elemental composition of several crystallographic phases. The results of our compositional analysis; olivine (Fo₇₂Fa₂₈), pyroxene (≈ 97 % En₈₀Fs₂₀Wo₀ and 3 % En₅₆Fs₆Wo₃₈), feldspar (albite), troilite, FeNi (taenite and kamacite), merrillite, chromite and hydroxyapatite; agree qualitatively with other reports from literature.

Keywords: Correlative microscopy; Large area mappings; Machine learning; Spectral maps.