TOP 10 AM papers: Riley et al. 1989, Hydrobiologia 176/177, 509-524

QEMSCAN mineral map of fluvial sediments collected from the 2010 Brisbane flood

Review of Riley et al. 1989, Hydrobiologia 176/177, 509-524

In terms of early adopters of automated mineralogy technology beyond mineral processing, due credit must be given to Riley, Creelman, Warner, Greenwood-Smith and Jackson from Australia. In their groundbreaking journal paper on "The potential in fluvial geomorphology of a new mineral identification technology (QEM*SEM)" they pioneer the application of automated SEM-EDS compositional mapping to the study of depositional environments and sediment sources. I only became aware of this paper after publishing a similar study on late Pleistocene flood deposits in the Flinders Ranges, South Australia as part of my PhD thesis, and would like to take this opportunity to give due credit to a study well ahead of it's time.

Riley et al. set out explaining why mineral analysis remains an underutilized diagnostic tool in fluvial geomorphology and sedimentology. Optical microscopy on fine fluvial sediments is cumbersome, statistically questionable and of limited value when it comes to discriminating source areas based on clay mineralogy.

The objective the authors face is to establish the maximum flood level the Nepean/Warragamba Rivers in New South Wales over the recent geological past. The deeply entrenched river terraces in the lower reaches are inappropriate indicators because they represent a different hydrological regime. In the absence of slack water deposits, veneers of alluvium mixed with colluvium by bioturbation remain the only record.

The task was to perform micron-scale textural analysis and discriminate non-fluvial from fluvial material as well as to quantify the sediment contributions from different tributaries in the catchment. Challenges in fingerprinting and provenancing fluvial sediments by mineralogical analysis include sorting and differential comminution during transport, post-depositional weathering, contamination by reworking and aeolian deposition, and mixing of sediments from different source rocks. All this requires a focus on minerals resistant to weathering, and sophisticated statistical analysis of the data.

The methodology section is exceptional. The paper provides detailed information on the QEM*SEM system configuration including a schematic diagram and a technical discussion of the data acquisition and processing in the appendix. The authors developed their own application-specific mineral identification protocol (SIP) and primary mineral lists, differentiating readily identifiable mineral species from broadly related silicate groups. This was at a time when no interactive Measurement Debug module for SIP development was available as in later versions of the iDiscover software package. Unfortunately, it was also before the powerful categorizer tools were developed that provide classification of particles by mineral association, size and shape. As a result, the authors had to make do with modal mineralogy data for 4 physical size fractions. Riley et al. clearly set an example by applying principle component analysis to produce and assess independent variables. The spatial relationship between minerals from different sampling locations was investigated by multiple cluster analysis.

The results are convincing and clearly differentiating fluvial from non-fluvial deposits by QEM*SEM mineralogical data. The conclusion was that the estimate of Probable Maximum Flood discharge for the downstream dam had to be revised. The conclusion after reading this paper is that sedimentologist can clearly benefit from revisiting this pioneering paper before applying the latest image analysis capabilities that compositional mapping solutions such as QEMSCAN provide.

TOP 10 AM papers: Grant et al. 1976, Scanning Electron Microscopy/1976 (III)

1982 prototype of QEM*SEM with mini-computer to the right at CSIRO Melbourne

Review of Grant et al. 1976, Scanning Electron Microscopy/1976 (III)

In the 70's, it was not uncommon to publish outstanding geoscientific research in workshop proceedings. This groundbreaking paper on "Multicompositional particle characterization using the SEM-microprobe" by Grant, Hall, Alan Reid and Martin Zuiderwyk is published in the Proceedings of the Workshop on Techniques for Particulate Matter Studies in SEM held at the IIT Research Institute in 1976. Over three decades ago, the authors - listed in alphabetical order - demonstrated the first computer-controlled automated mineralogy system and outlined a number of principle functions in mapping particles which changed little over the years despite the revolution in computational power and software languages since.

Grant et al.'s vision, as outlined in the introduction, is to determine sizes and composition of complex particles, and to "measure rather than infer" areas and perimeters to derive shape functions to better understand physical and chemical behaviors of particles in industrial and mineral treatment processes, i.e. mineral flotation and the degree of liberation during grinding processes.

The original instrument design consists of a mini-computer controlled e-beam which is automatically moved across the sample along a user-defined pattern. An initial "fast scan" locates the particles for detailed scanning. The dwell time for x-ray acquisition at each point can be defined. In addition, an "event acceptance filter" is in place to only record changes in material composition. Composition and e-beam coordinates are saved in form of digital maps composed of line segments and points.

The original system is setup to accept secondary and backscattered electron signals, absorbed specimen current, as well as energy-dispersive x-ray counts from either EDS detectors or microprobe. The measurement mode outlined is a proto-type for what is to become the Particle Mineral Analysis (PMA) in the QEM*SEM and later QEMSCAN solution. It uses either rapid BSE or SE signals to locate the edges of particles which are in turned scanned in detail using x-ray signals. The software includes algorithms linking particles extending across multiple frames, similar to the "field stitch" pre-processor in iDiscover. In addition, mixed signals between particle and mounting medium, referred to as "boundaries" are resolved, a first step in the development of Species Identification Program (SIP) boundary phase definitions and the award-winning "Boundary Phase" pre-processor by Paul Gottlieb in the iDiscover software package. Touching particles are discussed beyond shape parameters using secondary electron images to discriminate same phases in the discussion with reviewers at the end of the paper.

The boundary coordinates are saved allowing for a visual display of the data and basic image analysis functions, including the particle-by-particle calculation of area, perimeter, centroid and even the option to report phase contributions to the perimeter. Particular consideration is given the stereological challenges of reporting unbiased particle size, shape and composition from 2-D data. Interestingly, the authors point out to future investigations in mounting particles onto surfaces and using a second set of detectors providing biaxial views to better estimate particle sizes in three dimensions.

It can be fairly said that the authors laid the foundation for automated mineralogy and future software developments. It is a testimony to their visional capacity that they discus ways to move forward in 3-D particle analysis and even applications of their algorithms beyond rock particles, such as the analysis of pore space or images of macro-scale objects.

TOP 10 AM papers: Jackson et al. 1984, Proc. Australas. Inst. Min. Metall. 289, 93-97

QEMSCAN® Sample Blocks

Review of Jackson et al. 1984, Proc. Australas. Inst. Min. Metall. 289, 93-97

The quality of information provided by automated mineralogy solutions, as with any other analytical technique, depends first of all on the quality of the measured sample. It depends on proper sample collection, subsampling procedures, sample preparation and presentation. It is therefore most appropriate to start this review of the Top 10 Automated Mineralogy papers with this classic, although difficult to access, technical note by Jackson, Reid and Wittenberg on "Rapid production of high quality polished sections for automated image analysis of minerals". The 1984 seminal paper on sample preparation describes in great detail a method developed to mount particles for automated mineralogy analysis which remains, with minor adjustments, the standard protocol applied in SEM-EDS laboratories across the world to this day.

The paper starts with laying out the requirements for the production of a representative section mount, so that the mineralogy in the cross-section is representative of the overall parent sample mineralogy. The fundamental requirements are to provide a random, even 3-D distribution of particles, without segregation of particles by mass, density or grain size, or introducing preferential orientation. In addition, a high degree of surface integrity in the sectioned plain needs to be achieved in order to minimise bias introduced by pitting, plucking, grain shattering, or the preferential removal of less competent minerals.

Jackson et al. describe seven steps of sample preparation, including screen sieving, rotary riffling, mechanical dilution, epoxy mounting, grinding, polishing and conductive coating.

1) Screen sieving and/or cyclosizing in an ultrasonic ethanol bath to remove loose aggregation, oversized foreign matter, and provide optimal narrow size ranges for measurement.

2) Random subsampling of the parent material to a few grams is achieved by a rotary riffler.

3) Mechanical dilution of the particle samples is discussed to prevent segregation, minimizes particle-to-particle contact, and to assists random particle orientation. Jackson et al. suggest mixing the sample with crushed graphite of similar size range and surface angularity as an inert filler.

4) Even and random 3-D distribution is achieved by mechanical shaking the mixture in cylindrical plastic moulds. Subsequently, the dry mixture is cast by covering it in resin and hardener and stirring the sample. The epoxy-sample slurry can be evacuated to remove air bubbles.

5) Grinding the hardened block consists of two stages; the first to cut away surface epoxy and particle layers of preferred orientation well into the mixture of particles and filler, the second to remove damaged damage areas.

6) Final polishing with a sequence of diamond paste cloth laps is applied to improve the surface finish. Between grinding and polishing stages, the sample block is cleaned in an ultrasonic bath using a detergent solution or alcohol.

7) Finally, the sectioned sample surface is sputter-coated with a 20-30 nm carbon film making it electro-conductive.

The fundamental sample preparation protocol for automated SEM-EDS analysis laid out by Jackson et al. remains largely unchanged over the past 25 years and is widely applied by the leading service providers in the mining industry.