DISCRIMINATING POTASSIUM FELDSPAR FROM LCT AND NYF PEGMATITES BY LASER-INDUCED BREAKDOWN SPECTROSCOPY (LIBS)

Granitic pegmatites provide rare metals which are critical for many modern technologies. Although exhibiting a wide diversity in composition, two families of granitic pegmatites are widely recognized – LCT pegmatites characterized by enrichment in Li, Cs, Rb, Be and Ta & NYF pegmatites characterized by enrichment in Nb, Ti, Y, REE, Zr, U, Th & F. Although both pegmatite types are considered to be derived from highly evolved and volatile-rich melts, the former are typically associated with peraluminous granites whereas the latter are generally associated with metaluminous to alkaline/peralkaline granites Although biotite tends to dominate over muscovite in NYF pegmatites, these two pegmatite types sometimes can be difficult to distinguish in the field. Handheld laser-induced breakdown spectroscopy (LIBS) offers a potential solution.

LIBS is a straightforward and versatile form of atomic emission spectroscopy that focuses a short-duration, rapidly-pulsed laser beam onto a sample to create a plasma containing its constituent elements and then uses spectral analysis of the emitted light to detect the elements present. LIBS analysis can be performed in the laboratory or the ambient environment for rapid in-situ, on-site analysis using commercial handheld analyzers. A broadband LIBS spectrum contains information on all elements in a geomaterial above their intrinsic detection limits for that matrix type and can be considered a diagnostic geochemical fingerprint.

As a foundation for this LIBS study, compositional analysis by EMP and LA-ICP-MS was undertaken for K-feldspar from 28 known sources (13 LCT & 15 NYF). Chemometric analysis of this data demonstrated that the two pegmatite types could be reliably distinguished on the basis of major and trace element composition. Then, LIBS spectra were acquired for these 28 samples using a SciAps Z-903 handheld analyzer. The broadband LIBS spectra (190-975 nm) were pre-processed through normalization and baseline correction and then used to train and validate a random forest classifier. Finally, LIBS spectra for 46 additional ‘blind’ samples were tested against the trained algorithm with >96% classification success. These results suggest that integration of an appropriate spectral library onto a handheld analyzer holds promise for real-time pegmatite type discrimination in the field.