Identifying and Quantifying Mineral Abundance through VSWIR Microimaging Spectroscopy: A Comparison to XRD and SEM

Visible-shortwave infrared microimaging reflectance spectroscopy is a new technique to identify minerals, quantify abundances, and assess textural relationships at sub-millimetre scale without destructive sample preparation. Here we used a prototype instrument to image serpentinized igneous rocks and carbonate-rich travertine deposits to evaluate performance, relative to traditional techniques: XRD (mineralogical analysis of bulk powders with no texture preservation) and SEM/EDS (analysis of phases and textures using chemical data from polished thin sections). VSWIR microimaging spectroscopy is ideal for identifying spatially coherent rare phases, below XRD detection limits. The progress of alteration can also be inferred from spectral parameters and may correspond to phases that are amorphous in XRD. However, VSWIR microimaging spectroscopy can sometimes be challenging with analyses of very dark materials (reflectance <0.05) and mineral mixtures occurring at a spatial scales multiple factors below the pixel size. Abundances derived from linear unmixing typically agree with those from XRD and EDS within ~10%. ; Sample: A rock was collected from an travertine conglomerate, containing ophiolite clasts. The rock was collected in Oman near the Samail ophiolite (collected Jan 2012; B. Ehlmann; analyzed in Leask & Ehlmann, 2016). Data included are: 1) UCIS (Ultra-Compact Imaging Spectrometer; B. Van Gorp et al) data cube. Data near ends of sensor may be suspect [e.g. under 500 nm, over 2500 nm]. {OM12L_001_UCIS_cube_masked} 2) EDS-SEM mosaic image cube (4x downsampled to despeckle, using a nearest neighbour algorithm). Acquired Caltech October 2015. Chemical data are in atomic %. {om12L_001_SEM_4x_masked} 3) EDS-SEM data warped to UCIS cube [degree 4 convolution, using built-in ENVI warping algorithm and hand-picked ground-control points]. Three UCIS images were mosaicked together to build the base image. Note whole image difficult to match 100% perfectly; easier to line up a smaller subset. ...