New Sample Preparation Developments to Minimize Mineral Segregation in Process Mineralogy

Abstract

Sample preparation is the most critical step in a microscopy-based characterization. Ideal products should be the most representative and the least biased. Under this condition, the imaged surface displays a random dispersion of particles, with morphological, densitary and compositional deviations virtually absent. In practice, every single preparation procedure is a superposition of several steps with different degrees of imperfectness. Different kinds of artifacts generated by epoxy manipulation and polishing procedures are easily minimized; however, artifacts generated by densitary segregation are far more difficult to avoid and even to detect, being frequently negligenced. This effect has been noteworthily observed in standard QEMSCAN sample preparations. Instead of behaving as a particle deagglomerant, the graphite filler floats, segregating from the sinking sample. A typical symptom is seen in the data reconciliation, where elements component of sulfides and oxides (S, Cu, Ni, Fe, etc.) are mineralogically overestimate, while those component of silicates (Si, Mg, K, Al, etc.) are mineralogically underestimated. As long as the gravity force is always present and geological materials are usually heterogeneous in all scales, the effect of mineral segregation can only be minimized. Two types of samples were considered for this purpose: fractioned and bulk. In fractioned samples, the proposed solution is a dynamic epoxy curing, where a slow and continuous lateral rolling is applied to covered moulds, until complete hardening is achieved. This relatively simple device minimizes at once the effects of particle agglomeration and mineral segregation, although the moulds can retain air enough to create bubble artifacts, requiring additional steps of sealing with epoxy. Nonetheless, the cost-benefit is worth indeed. In bulk samples, the proposed solution is a centrifuge-forced epoxy curing. The once hardened sample preparation is then cut in the vertical axis and potted again in a conventional circular mold. This procedure ensures the same probability of exposition for all range of sizes and densities in the new surface, and has been applied successfully in Vale for the last years. The products generated by both dynamic procedures where inspected in optical and electron microscopy. Both procedures were designed to produce the same 30 mm-diameter QEMSCAN/MLA-standard sample preparation. However, the purpose, measurement type, data interpretation and mineralogical information are totally diverse. In fractioned sample preparations, only preliminary tests were made but are noteworthy the homogeneity, the lack of clusters and the absence of touching particles. This characteristics are the most desirable for PMA analysis in QEMSCAN. Additionally, there is no addition of graphite or other fillers, reducing contamination risk, time and costs. Bulk sample preparations, inversely, have been used for long time in optical microscopy applications, and more recently were successfully adapted for BMA field scan analysis, generating statistically robust analysis with visualisation of rock textures. The lack of mineral segregation is attested by an optimum data reconciliation.