Automated Mineralogy Software vs Two Geologists: First Impressions of Zeiss’ Mineralogic and Oxford Instruments’ AZtecMineral on Non-Mining Geological Materials

Session

Poster Session Three

Authors

Dr Jennifer Mitchell (1), Francesca Willcocks (1), Dr Thomas O'Hanlon (1), Dr Alexander Strachan (1)

Affiliations

1. Plymouth Electron Microscopy Centre, University of Plymouth

Keywords

Geology, automated mineralogy, Mineralogic, AZtecMineral

Abstract text

The use of automated mineralogy software is a powerful tool in the analysis of mining samples – both whole rock and grain mounts – to investigate the composition, distribution, and association of target minerals with a particular focus on ore minerals. Although there is complexity within ore mineral groups, in a typical mining sample there is often significant differences between target and gangue (commercially worthless) phases. However, the vast majority of rocks are not mining samples and there can be both a much broader range of compositions in the sample but also within a single mineral phase. X-ray element mapping through techniques such as energy dispersive spectroscopy (EDS) allows researchers to investigate small-scale variations in the distribution and abundance of elements, creating detailed mineral maps which include the subtle variations within a single grain. Although this can produce very high quality data sets in terms of spatial resolution and minor element variations, a significant amount of post-processing needs to take place to calculate factors such as modal abundance and porosity. As such, the applications of automated mineralogy software for non-mining complex geological materials is an interesting question.

The purpose of this study is to present our initial impressions of automated software for use across various geological materials, and the relative ease of analysis and processing of each when compared to large area X-ray mapping which is the standard method of analysis at Plymouth Electron Microscopy Centre, where automated analysis is relatively new. Here, we present the two main barriers to applying automated software to non-mining materials.

Both Zeiss Mineralogic and Oxford Instruments AZtecMineral were tested on igneous rocks where certain minerals may have similar compositions, a single mineral may show a range of compositions throughout the sample, and common igneous features such as compositional zoning. This automated software generates similar results, although operate in slightly different ways. Mineralogic uses a pixel-by-pixel analysis and characterises samples based on elemental composition, whilst AZtecMineral first identifies features based on the backscattered electron greyscale and then analyses each feature for composition. Here, Mineralogic was run on a Zeiss Sigma 300 LV FEG-SEM with two Oxford Instruments X-Max 65mm² EDS detectors and AZtecMineral on a JEOL 7001F FE-SEM with Oxford Instruments X-Max 50mm² EDS detector at accelerating voltage 20 keV, probe current ~3 nA, and appropriate working distances for the particular detector set-ups. 

The first main stumbling block of automated software is the effect of sample preparation. Defects in the sample surface such as scratches, depressions, and voids drastically affect the identification of features from the greyscale images used by both programs. This can be somewhat mitigated in large area element mapping provided the defect is not too extreme as the process does not rely on feature identification, but can severely hamper classification during automation where the defects are often be misidentified by the software, or in cases where a minimum grain size and/or grain boundary is set can cause grains to not be classified. Suitable sample preparation can be time consuming, particularly for plastic or fragile minerals that can easily deform or be pulled from the sample during preparation. This therefore runs the risk that – for difficult samples – more time is spent on preparation than analysis which has implications on costings and whether or not the use of automated software is beneficial over standard element mapping.

Secondly, and most importantly, it is unlikely to be possible to create a classification scheme that can be applied across multiple lithologies. Igneous and metamorphic rocks in particular record a wide variety of mineral compositions across a rock type, a single sample, and within a single mineral grain. Features such as igneous compositional zoning require the classification scheme used for automated processing to be significantly broadened, thus increasing the chances of a mineral being misclassified. Additionally, minerals that are very similar in composition such as olivine, (Mg,Fe)₂SiO₄, and certain pyroxenes, (Mg,Fe)₂SiO₃, where the only real variation is in the amount of oxygen and silica. Both oxygen and silica are often required to have a somewhat wide spread in both Mineralogic and AZtecMineral, drastically limiting the ability to separate such phases. Whilst a classification scheme for a single rock sample can be set-up in a practical time-frame, the further applications of it will be limited and future analysis will require adjustments to be made.

Through our initial trials of Mineralogic and AZtecMineral, we recognise the power that these software have for the classification of rock samples where distinct minerals are readily distinguished. However, at present it appears that large area element mapping better suits analysis and characterisation of complex non-mining geological materials such as basalts and granites. Updates to the software and additional experimentation may provide further insights into the applications of these software.