Automated Mineralogy

A machine learning solution to geological sample characterisation

Introduction

Automated mineralogy is 'a range of analytical solutions that are used to quantitatively evaluate the mineralogy of samples'. Essentially, automated mineralogy is a software that can be used in conjunction with Scanning Electron Microscopes and associated detectors ( Energy Dispersive Spectroscopy and Backscattered Electron Imaging) that takes multiple different outputs provided by the microscope of a heterogeneous sample, which can then be used to categorise sections of the sample to produce a full characterisation of the sample in respect to:

bulk mineralogy
mineral association
liberation data
free surface area
porosity
element concentration
grain size.

The software is mostly used in geoscience, although, it can be used in disciplines such as: forensic sciences, archaeology and material sciences.

Samples that can be analysed using automated mineralogy

To run automated mineralogy on a sample it needs to prepared so that it is suitable for analysis, and most of this can be done in house at PEMC. If the sample is large it often needs cutting down before being mounted into a resin block and polished down to 1 µm. If the sample consists of grains e.g. sand, crush rock, then this will be made into a slurry of resin and mounted into a resin block before also being polished. Thin sections are also possible to analyse using automated mineralogy, however, this sample preparation technique is not currently available at PEMC. Find out more on sample preparation facilities at PEMC .

How it works

The software works by using a multitude of outputs from SEM, most notably using EDS data and BSE images, which are used in conjunction to separate out phases, minerals or grains, depending on the sample. This is then used to characterise the sample using a classification scheme that can be one that has already been created, or a unique user input one if the sample has a specific characterisation scheme needed.

Different automated mineralogy software characterise samples in different methods. We have two types of software in house. Zeiss’s Mineralogic (available on the Zeiss Sigma 300 LV FE SEM ) and Oxford Instrument’s AZtec Mineral (available on the JEOL 7001 FE SEM ).

The different software work better for different samples depending on what kind of analysis is needed.

With both software available to us at PEMC, it is possible to gather a host of information about a sample through automated mineralogy including:

bulk mineral analysis
particle mineral analysis
specific mineral search
trace mineral search
porosity measurements
grain size analysis
elemental assay
liberation analysis.

Zeiss Mineralogic vs Oxford Instruments AZtec Mineral

Zeiss’s Mineralogic uses pixel by pixel analysis and characterises samples via their chemical composition. Each pixel is assigned a mineral/element depending on its chemical composition which is compared to the characterisation scheme. Greyscale on BSE tends to get used for specific mineral searches, if looking for grains of a specific greyscale, and then is characterised via chemical methods.

Works for:

whole samples
crushed/grain samples
specific mineral searches
elemental assays.

Oxford Instruments' software firstly separates our different grains and crystals by analysing the BSE image of the sample, splitting images into distinctive features depending on their greyscale. Each feature is then analysed by EDS to gain an average composition of the feature, which is then assigned a mineral/element from the characterisation scheme.

Works for: