Short-Wave Infrared (SWIR) Hyperspectral Characterization of Alteration at the Sadiola Hill Gold Deposit, Mali, Western Africa

Introduction: Sadiola Hill is an ~8 Moz gold deposit located in western Mali within ca. 2200-2050 Ma tectonic window known as Kédougou-Kénieba Inlier (KKI) exposing Western African craton. The deposit is hosted in metasedimentary package made up of impure carbonate rocks, wackes, and arenites intruded by three distinct igneous phases. The N-S-trending Sadiola shear zone, related to the regional Senegal-Mali shear zone, and NNE-trending third-order fault splays acted as conduits for auriferous hydrothermal fluid flow. The deposit has undergone a complex polyphase alteration history. Alteration assemblages related to gold mineralization consist of biotite-carbonate-quartz-sulphide. Other assemblages include calc-silicates, chlorite, white mica, scapolite, and tourmaline (Hein and Tshibubudze, 2007; Cameron, 2010; Masurel et al., 2017). Current research is aimed at characterization of alteration at the mineral scale, as well as assessment of cooling trend(s) and alteration footprint(s) with high-resolution short-wave infrared (SWIR) hyperspectral scanning. In addition to detailed mineralogical classification, changes in fluid chemical parameters, temperature, and pH are determined with variations in white mica, namely, the position of Al-OH bond in the SWIR range and white mica crystallinity. Furthermore, hydrothermal fluid chemistry is assessed with Fe2+ content in carbonate group minerals. Protolith control on alteration expression is also investigated.

Methodology: The research is carried out on drill core samples from six diamond holes capturing best alteration examples and is solely laboratory based. Hyperspectral scanning of drill core samples and offcut blocks was acquired at Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente with Specim Hyperspectral camera at medium (0.26 mm pixel size) and high-resolution (26 µm pixel size). Data for each pixel is collected in x-, y-, and z-direction with x- and y-values of the pixel cube corresponding to length and width of a pixel within a horizontally stationed sample. Z-values represent a stack of bands in the SWIR range with 12-nm spectral resolution, amounting for the total of 288 bands. Conversion of the raw data into calibrated hyperspectral images was done with hyperspectral python (hyppy), an in-house software developed in ITC. Calibrated images were converted into wavelength maps over six different ranges capturing depth and position of first, second, and third deepest absorption features. Decision trees for mineral classification were developed in ENVI Classic 5.5 software for project-specific mineralogy as matching algorithms between recorded bands to spectra from USGS spectral library. Polished thin sections were prepared from drill core samples scanned at high-resolution for one-to-one comparison of mineral maps to petrographic observations (van Ruitenbeek et al., 2017). Conventional transmitted-light microscopy was added to the workflow to validate hyperspectral mineral maps and to put paragenetic constraints related to crosscutting relationships and replacement textures to the interpretation. Furthermore, chemical composition of each spectral class is determined with electron probe microanalyser (EPMA).

Preliminary Results: Ore-stage vein-controlled and pervasive carbonate alteration is represented by ankerite and ferroan dolomite with complex zoning and exsolution patterns. Post-ore carbonate species are characterized by fracture-fill calcite. Finally, recrystallized impure carbonate host is dolomite-calcite dominant.

Muscovite is more aluminous in igneous units in comparison to greywacke and impure carbonate reflecting availability of Al3+. In cases where Na+ is ubiquitous, e.g., tonalite unit, white mica composition changes to paragonite (Fig. 1b). Mineral-scale variations were also noted in muscovite. When muscovite replaces biotite, Al-OH feature shifts to longer wavelength consistent with incorporation of Mg2+ from biotite into muscovite crystal structure. Illite crystallinity has a common trend of increasing inward into a vein and is also protolith dependent, with low-crystallinity illites constrained to greywacke and muscovite to igneous units and impure carbonates.

Stoichiometry of carbonates and phyllosilicates has been validated with electron microprobe (Fig. 2). Ferrous drop calculated from R1600/R1310 hyperspectral band ratio correlates well with Fe2+ content in carbonates. Analyses obtained from white mica confirmed its Al-poor chemical composition. Biotite porphyroblasts are of intermediate Fe-Mg composition, while early hydrothermal biotite falls in Mg-rich field not reaching phlogopite composition. Chlorite composition is also Fe-Mg intermediate.

Overall, highly reactive folded argillaceous layers within the impure carbonate acted as chemical trap for gold mineralization during the D3 event. Biotite supplied iron for sulphidation reactions leading to pyrrhotite-arsenopyrite-chalcopyrite-sphalerite-pyrite-free gold precipitation, while available Al3+ promoted white mica crystallization. Mineralization within the greywacke is more sporadic with iron also sourced from biotite and is associated with Fe-bearing carbonate-muscovite alteration.

Future Work: Zonation of ore-stage carbonate group minerals and relation of Fe2+ content to gold mineralization will be investigated further. Crosscutting relationships and chemical composition of carbonates will be assessed with cathodoluminescence (CL). In addition, drill core samples scanned at 26-µm resolution will be submitted for complete geochemical characterization to ALS Global. Trace element associations will compliment hyperspectral interpretation of the alteration footprint and cooling trends within the system. In addition, protolith control on mineralization and alteration expression will be constrained.