Classification of soil aggregates using SEM-EDX hyperspectral data analysis

Soil is a complex system characterized by peculiar chemical, physical and biological properties. Soil characteristics are the results of the attributes of micrometric and submicrometric domains. For this reason, the use of sensitive techniques with high spatial resolution is mandatory. Scanning electron microscopy (SEM) is a very powerful analytical technique for the analysis of morphology and microstructure of the soil and its components. When SEM is coupled with energy dispersive (EDX) or wavelength dispersive (WDX) x-ray detectors, the chemical analysis of the sample can also be performed. The combination of microstructural and chemical data can give information about elemental associations within minerals or soil aggregates, allowing their detailed characterization. Chemical databases are not fully used and scientists often superimpose different chemical maps in order to find correlations among element on the base of different colour scales. However, by elaborating data this way, most of the objective hyperspectral database information is lost and the obtained results rely on the scientist’s subjective choice. To overcome this problem, in the present work “Datamuncher_gamma”, a software recently developed for the analysis of SEM-EDX hyperspectral data , is presented and applied to study soil samples. Datamuncher_gamma allows to obtain elemental maps from hyperspectral datasets and to compare the characteristic fluorescence lines of all the elements found in the sample. The visual recognition of particular correlations then allows to identify particular mineral phases and soil features. Specifically, datamuncher_gamma was applied for the study of chromium polluted soil aggregates. SEMEDX analysis were conducted on soil thin sections with a Zeiss Σigma 300 VP FEG-SEM working at 15 kV and equipped with an Oxford EDX C-MaxN SDD. Different element correlations with characteristic ratios were recognised. For example, Si vs Al scatterplots were useful for the identification of different aluminosilicates and a more precise characterization was obtained by comparing the Si signal with the signal of K, Ca and Mg. Chromium was found mostly associated with aggregates having high C/O ratio. Among them, four different types of Cr-aggregates were recognised on the base of Fe/Cr scatterplot. Such hyperspectral approach using SEM-EDX data could be used to investigate many soil processes regarding metal(loid) pollutants or micronutrients at the microscale.