Fontana, F. Fernando1,2; Tassios, Steven2,3; Stromberg, Jessica2,4; Tiddy, Caroline1,2; van der Hoek, Ben1,2; Uvarova, Yulia2,4.
1Future Industries Institute, University of South Australia, Mawson Lakes, Australia, 2Mineral Exploration Cooperative Research Centre – MinEx CRC, 3CSIRO Mineral Resources, Clayton, Australia, 4CSIRO Mineral Resources, Kensington, Australia.
Rock classification and discrimination is commonly performed by geologists through visual inspection of rocks in the field or by geological logging of drill cores. However, this visual interpretation of rock lithology is notoriously subjective and inconsistent. The utilisation of handheld and core scanning instruments for collecting standardized data coupled with rapid and consistent methods for extracting geological information (e.g. clustering algorithms and boundary detection techniques) is gaining popularity for supplementing traditional geological classification. Such techniques are objective and reduce inconsistencies, improving rock classification and geological logging outputs. Laser-Induced Breakdown Spectroscopy (LIBS) is an emission spectroscopy technique of interest for lithogeochemical analysis due to its sensitivity to light elements common in geological materials (e.g. Li, Be, Na, Mg), that are difficult to quantify using other techniques such as X-ray fluorescence (XRF). The wavelet tessellation method utilises the wavelet transform for edge detection in spatialized data and can be used for determination of boundaries in downhole geochemical data which may represent lithological contacts. Wavelet tessellation can be applied to raw LIBS data outputs (counts vs wavelength) with no requirement for data filtering or manipulation. In this study we demonstrate the application of wavelet tessellation for rapid analysis of LIBS geochemical data to produce pseudo-logs that are representative of a test-block geological stratigraphy. A test-block stratigraphy was created using 22 rock slabs glued together to produce a block ~16 cm in length. The samples were chemically (XRF) and mineralogically (XRD) characterised and selected to represent two discrete and geochemically distinct rock types (granitoid and marble). LIBS spectral data for nine major elements (Al, Ca, Fe, K, Mg, Na, Si, Ti and Mn) was generated along a continuous line at a constant speed of 10 mm/min at 5 Hz in which every group of 10 spectra were averaged into 1 spectrum, creating 460 spectra that each correspond to an interval of ~0.35 mm. Clustering algorithms were used to separate the samples into groups reflective of rock-types based on the LIBS outputs for those 9 elements. Wavelet tessellation was undertaken to produce pseudo-logs of the test-block stratigraphy. Comparison of the known test-block stratigraphy from laboratory XRF and XRD data with the wavelet tessellation pseudo-logs generated from the LIBS data shows the potential application of multi-variate wavelet tessellation analysis for rapid interpretation of stratigraphic boundaries and rock-type from LIBS geochemical data. The discrete lithological groups of marble and granitoid were effectively separated, and less obvious distinctions between Ca and Mg-rich and Ca-rich marble slabs, and mafic to alkali-rich granitoids were also successfully made. The methodology presented here highlights how a workflow using LIBS coupled with wavelet tessellation can be used for rapid lithogeochemical interpretation to produce reliable and objective lithogeochemical pseudo-logs to supplement subjective visual drill core logging. Moreover, the sensitivity of LIBS analysis for light elements provides the opportunity for the development of new lithogeochemical interpretation workflows for exploration campaigns and characterisation of alteration halos.
Fernando is a geologist interested in geochemistry and mineral exploration. Currently, he is a PhD candidate at the University of South Australia and postgraduate researcher in the MinEx CRC, where he is applying LIBS in geosciences including for downhole assays.