Premium
Quantitative compositional analysis of sedimentary materials using thermal emission spectroscopy: 2. Application to compacted fine‐grained mineral mixtures and assessment of applicability of partial least squares methods
Author(s) -
Pan C.,
Rogers A. D.,
Thorpe M. T.
Publication year - 2015
Publication title -
journal of geophysical research: planets
Language(s) - English
Resource type - Journals
eISSN - 2169-9100
pISSN - 2169-9097
DOI - 10.1002/2015je004881
Subject(s) - mineralogy , spectroscopy , calcite , mineral , gypsum , partial least squares regression , range (aeronautics) , clay minerals , analytical chemistry (journal) , chemistry , geology , materials science , mathematics , physics , statistics , paleontology , organic chemistry , chromatography , quantum mechanics , composite material
Fine‐grained sedimentary deposits on planetary surfaces require quantitative assessment of mineral abundances in order to better understand the environments in which they formed. One way that planetary surface mineralogy is commonly assessed is through thermal emission (~6–50 µm) spectroscopy. To that end, we characterized the TIR spectral properties of compacted, very fine‐grained mineral mixtures of oligoclase, augite, calcite, montmorillonite, and gypsum. Nonnegative linear least squares minimization (NNLS) is used to assess the linearity of spectral combination. A partial least squares (PLS) method is also applied to emission spectra of fine‐grained synthetic mixtures and natural mudstones to assess its applicability to fine‐grained rocks. The NNLS modeled abundances for all five minerals investigated are within ±10% of the known abundances for 39% of the mixtures, showing the relationships between known and modeled abundance follow nonlinear curves. The poor performance of NNLS is due to photon transmission through small grains over portions of the wavelength range and multiple reflections in the volume. The PLS method was able to accurately recover the known abundances (to within ±10%) for 78–90% of synthetic mixtures and for 85% of the mudstone samples chosen for this study. The excellent agreement between known and modeled abundances is likely due to high absorption coefficients over portions of the thermal infrared (TIR) spectral range, and thus, combinations are linear over portions of the range. PLS can be used to recover abundances from very fine‐grained rocks from TIR measurements and could potentially be applied to landed or orbital TIR observations.