This data set contains mineral chemical analyses of chromite, orthopyroxene and plagioclase of five chromitite layers and their immediate host rocks from drill core ZK135 of the northwestern Bushveld Complex. The sampled interval of ZK135 covers the transition of the Lower and Middle Group chromitites (LG6, LG6a, MG1, MG2, MG2 II). Detailed geochemical profiles are presented by data sheets and graphically to reveal small-scale variations in mineral chemistry.Mineral chemical analyses were conducted on drill core material from borehole ZK135 from the Thaba mine, operated by Cronimet Chrome Mining SA (Pty) Ltd. Briefly, each layer is hosted in pyroxenite and comprises a main chromitite layer and in some chromitites additional stringers above or below the main layer. Chromitites are composed of chromite, which are cemented by intercumulus plagioclase and orthopyroxene oikocrysts. A detailed study of the orthopyroxene oikocrysts in MG1 was published by Kaufmann et al. (2018). Additionally, euhedral to subhedral cumulus orthopyroxene is present in the LG6a, MG1 and MG2 layers. Pyroxenitic partings, small layers of pyroxenite within the chromitite, occur throughout the LG6 to MG2 layers. The main mineral phases chromite, orthopyroxene and plagioclase were analyzed by electron microprobe in each layer and the adjacent pyroxenitic host rocks.Major-element compositions of chromite, orthopyroxene and plagioclase were analyzed by electron microprobe at the Museum für Naturkunde Berlin with a JEOL JXA-8500F EMP equipped with a field emission cathode and five wavelength-dispersive spectrometers. For more information please consult the data description file and Kaufmann et al. (2019) to which these data are supplementary material.
The GEOROC database includes helpful compilations of mineral compositions aggregated from measurements reported in decades worth of publications, but it can be challenging to consistently filter mislabeled, inaccurate, or incomplete mineral compositions. MIST (Mineral Identification by Stoichiometry) is a stoichiometry-based computational algorithm that identifies geochemical observations with normalized elemental ratios matching natural minerals. The stoichiometric filters that were manually coded in MIST for over 240 mineral species are based on reported mineral formulas and well-documented examples of mineral chemistry reported in RRUFF and associated databases, typically including a ~5-10% tolerance in stoichiometric ratios based on measurement errors, vacancies, and substitutions. The MIST model can therefore efficiently filter the GEOROC mineral compilation files to recognize compositions whose normalized oxides match the labeled mineral stoichiometry. Furthermore, the MIST output includes results of intermediate data manipulation steps, a detailed stoichiometric formula for each input composition, and consistently calculated mineral endmembers such as Fo, En, Ws, and Fs. MIST is agnostic to the instrument used to collect oxide data. Because MIST uses normalized oxides, it cannot distinguish between some mineral species, where applicable, they are reported as a group (e.g., gypsum/bassanite/anhydrite). MIST can only recognize minerals encoded in the algorithm, so other real but less common minerals will not be recognized. The full list of minerals MIST can recognize, along with more details of the algorithm and results pages, are published in Siebach et al. (https://doi.org/10.1016/j.cageo.2025.106021).
This dataset includes fifteen of the Compiled Mineral files published by GEOROC in 12-2024 including the MIST results (whether or not a species was confirmed by MIST). Prior to running the data through MIST, all files were filtered to only include mineral compositions that included major oxides (e.g., silicate mineral compositions where SiO2 > 0 wt%). Furthermore, all variations of reported Fe were collapsed into a single column representing FeOT.
Metadata is preserved from the original compiled GEOROC files, so users may add additional filters as appropriate for different purposes. Results have not been filtered for reported sum of total oxides, but doing so can help identify particular mineral species (e.g., separate gypsum from bassanite). An additional file preserves the full reference information for each mineral compilation.
We suggest using the compositions that MIST identifies as stoichiometrically consistent with a mineral species as a standardized filter on the GEOROC datasets prior to utilizing the data in machine learning models or similar applications. These may also be helpful any time a user would like standardized formulas or mineral endmember information for these mineral compilations.