A multivariate statistical approach for mineral geographic provenance determination using laser-induced breakdown spectroscopy and electron microprobe chemical data: A case study of copper-bearing tourmalines

The geographic provenance of minerals provides key insights into a range of geologic problems, including the source of gem materials. The tourmaline supergroup is unparalleled in its ability to record and preserve extensive chemical signatures of its formational environment. To evaluate the likelihood that tourmalines of similar compositions from separate geographic localities could be differentiated, a multivariate statistical approach has been utilized on two complementary data sets. These chemical analytical data sets of copper-bearing "Para & iacute;ba" tourmaline include data sets acquired with Laser Induced Breakdown Spectroscopy (LIBS) and electron microprobe analysis (EMPA). Fifty-four samples of copper-bearing tourmalines from known source locations from Brazil (S & atilde;o Jos & eacute; de Batalha of Para & iacute;ba state and the neighboring Rio Grande do Norte state), Mozambique, and Nigeria, were analyzed using LIBS with a subset of these samples analyzed by EMP. Data sets obtained by each method were evaluated with multivariate statistics (PCA, PLSR). Although the sample set is limited, sequential PLSR modeling of the spectra clearly distinguished the four localities with high success: >95% for LIBS and >87% for EMP. The statistical analyses of the two techniques, LIBS and EMP, suggest that each technique emphasizes different elements for discrimination when considered in the context of the available data. The elements Cu, Mn, Fe, Mg, Ti, Zn, K, H, Co, and V were significant in LIBS chemometric models. Statistically significant elements in EMP models were Mn, Cu, Al, Ca, K, and F. Each technique results in a robust determination for geographic provenance of tourmalines with comparable compositions. The significant distinguishing chemical elements reflect geochemical distinctions in each host environment that are imparted on the tourmaline. Multivariate statistics applied to LIBS and EMP data provide an effective tool for provenance discrimination of Para & iacute;ba tourmalines, distinguishing Brazilian-sourced samples from African-sourced materials. These data provide new methods for separating the geographic origin of minerals with very similar composition such as demonstrated here for copper-bearing tourmalines.