characterising chalcopyrite liberation and flotation potential: examples from an iocg deposit
TRANSCRIPT
Minerals Engineering 24 (2011) 1271–1276
Contents lists available at ScienceDirect
Minerals Engineering
journal homepage: www.elsevier .com/locate /mineng
Characterising chalcopyrite liberation and flotation potential: Examplesfrom an IOCG deposit
Julie Hunt a,⇑, Ron Berry b, Dee Bradshaw b
a CODES ARC Centre of Excellence in Ore Deposits, University of Tasmania, Private Bag 79, Hobart, Tasmania 7001, Australiab Julius Kruttschnitt Mineral Research Centre, University of Queensland, Australia
a r t i c l e i n f o
Article history:Available online 10 May 2011
Keywords:ClassificationLiberation analysisMineral processingSimulationModelling
0892-6875/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.mineng.2011.04.016
⇑ Corresponding author. Tel.: +61 3 6226 2782; faxE-mail address: [email protected] (J. Hunt).
a b s t r a c t
A critical aspect of geometallurgy is quantifying mineralogical and textural relationships that affect min-eral processing (e.g., liberation and recovery) and it is vital that this information is included in the plan-ning process for both mining and mineral processing. However, to date, this has been an expensive andtime consuming venture and only minimal amounts of this type of data are available to be included in theplanning process. Our research is focused on developing new methods that will produce the requiredmineralogical and textural data rapidly and inexpensively. These include obtaining quantified texturaldata, such as the size and distribution of the valuable phase and its association with other minerals,by extracting it directly from mineral maps. In addition, simulated breakage of drill core samples wasused as a rapid way of looking at various particle sizes to determine potential liberation behaviour.The predicted liberation parameter compares favourably with results obtained from typical MLA recoveryanalysis, is spatially coherent and can be used to recognise domains of high and low liberation potentialthat are expected to affect the grade recovery curve. The flotation response was evaluated and the tech-nique validated using a small scale test being developed at the Julius Kruttschnitt Mineral Research Cen-tre, i.e. the JKMSI (mineral separability indicator).
� 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Practical geometallurgy requires a database of parameters thatpredict mineral processing performance. Obtaining this type ofinformation rapidly and inexpensively is vital so that a represen-tative number of measurements can be carried out and the resultsincluded in the planning process for both mining and mineralprocessing. Recent advances in digital photography, particularlyin image processing software, have led to a resurgence of interestin optical microscopy mineralogy as a source of rock textureinformation. Our work suggests that optical techniques have aplace in generating medium quality cost-effective microscalemineral maps with direct application to geometallurgy. In theexamples presented here we have used mineral maps producedthrough optical mineralogy and automated mineral identificationplus simulated breakage as a rapid way of looking at variousparticle sizes to determine potential liberation (and flotation)behaviour. This method does not mimic actual breakage but canprovide a way of ranking samples in terms of their relativeprocessing behaviour.
ll rights reserved.
: +61 3 6226 7662.
2. Methodology
2.1. Image collection
Ninety-six 2 m-long samples of half drill core containing coppermineralisation (as chalcopyrite) were chosen from five drill holesto give a cross-section through an iron oxide–copper–gold orebody. Each sample was crushed and a representative (riffle splitter)sample of particles in the size range from �1.18 to +0.6 mm wasselected. As this particle size is more than five times the grain sizeof the Cu minerals it allows the fundamental rock properties to bemeasured before modification by grinding. The grain mounts typi-cally contain 500 particles and 1000–5000 grains of Cu sulphide.The coarse particles were mounted on a polished thin sectionand analysed using optical microscope techniques as described inBerry (2008), Berry and McMahon (2008) and Hunt et al. (2010).Image collection for each sample was carried out using a micro-scope with a high precision stage (<1 lm error in reproducibility)to allow the direct tiling of frames and good registration of multi-ple image layers. Transmitted-light plane-polarised, transmitted-light cross-polarised and reflected-light plane-polarised imageswere collected along with a transmitted-light cross-polarised im-age with a tint plate inserted. All lighting conditions were keptconstant for all image acquisition. Exposures were set to avoidany saturated pixels. The images were collected at one third
1272 J. Hunt et al. / Minerals Engineering 24 (2011) 1271–1276
resolution to give a 0.55 megapixel image. The resolution wasagain reduced (�2) during mosaic generation to give an overallpixel size in the final images of 4.73 lm. This size allows rapidanalysis of a 3 cm2 area. The smallest object recognised at this res-olution is 10 lm across.
2.2. Automated mineral identification
Automated mineral identification was facilitated via imageanalysis that was carried out using the sophisticated object-oriented multi-spectral software Definiens Developer 7 (Definiens,2008a). Using this software the four RGB images were read directlyinto 12 greyscale bands. A multi-resolution segmentationalgorithm produced small objects based on homogeneity acrossall bands and the objects were then classified using a complexset of rules (Process Tree) to produce classified images (alsoreferred to as mineral maps; Fig. 1 top). The example deposit has
Fig. 1. Top – typical example of a mineral map; bottom – example of simulated fragmenapproximately 150, 75 and 38 lm.
simple sulphide mineralogy and this was easily analysed usingthe optical microscope system. The valuable phase in this depositis chalcopyrite and grains greater than 20 lm across were recogni-sed with better than 85% precision.
2.3. Simulated fragmentation
The chessboard segmentation algorithm available in DefiniensDeveloper 7 was used to create simulated fragments from the min-eral map of each sample. In this algorithm a square grid of fixedsize, aligned to the left and top borders, is applied to the mineralmap and the image is cut in squares along grid lines (Fig. 1 bottom;Definiens, 2008a,b). We chose to use grids that would create sim-ulated fragments with sides of approximate length 150, 75 and38 lm. These equate to simulated fragments with sides of length32, 16 and 8 pixels. Simulated fragments with no chalcopyrite
tation of a mineral map. Three sizes of simulated fragments were used with sides of
J. Hunt et al. / Minerals Engineering 24 (2011) 1271–1276 1273
and/or with a resin content (i.e., image background) greater than5% were omitted from the analysis.
2.4. Statistical analysis
Statistics used in analysis of the particulate samples and simu-lated fragments were derived from data extracted from the mineralmaps. Data extracted includes information based on the originalparticles, the simulated fragments and the individual ‘‘grains’’ ofchalcopyrite. A series of summary properties were calculated basedon this extracted data. These included width and length of chalco-pyrite; relative area of each mineral phase within the sample (i.e.,mineral map), within each simulated fragment, and within rims 5
A
B
C
Fig. 2. Examples of liberation classes with 0, >0–10, >10–20, . . . >90–<100, 100% chalcopygrades – (A) 1.99%, (B) 1.06% and (C) 0.21% Cu.
pixels wide around each chalcopyrite ‘‘grain’’; and proportion ofchalcopyrite on particle rims and on the edges of simulatedfragments.
2.5. MLA analysis
Mineral liberation analyser (MLA) extended back-scatteredelectron (XBSE) analysis (Gu, 2003) was carried out on particulatematerial from the same samples used for optical microscopeanalysis. Two sizes of particles were examined: +38–150 and+150–212 lm. The MLA data presentation software was used tocalculated free surface liberation with respect to chalcopyrite.
rite on the edges of 75 lm simulated fragments for three samples with different Cu
1274 J. Hunt et al. / Minerals Engineering 24 (2011) 1271–1276
2.6. Small scale flotation (JKMSI)
The ore was dry milled and the +38–106 lm and 150–212 lmsize fractions were prepared for use in the JKMSI test. For each test,10 g of sample was weighed and added to water and conditioned,including ultrasound, to simulate a freshly milled surface. Equiva-lent concentration to 20 g/tonne ore of sodium ethyl xanthate(SEX) and 10 g/tonne MIBC was added as frother. The samplewas transferred to the JKMSI device for the test after which thetails and concentrate samples were collected, filtered and driedand sent for assay.
3. Results
Data extracted for chalcopyrite in the particles include area andborder length for all chalcopyrite found in the mineral map of eachsample. Note, in the mineral maps it is not possible to distinguishtrue grain boundaries, thus, the area and border length do not referto grains/crystals of chalcopyrite but to discrete patches of chalco-pyrite which may contain a number of crystals.
3.1. Liberation potential
One approach to liberation potential is to use simulated frag-ments to give an estimate that allows the division of the ore depos-it into liberation domains (e.g. low, medium, high liberationpotential). A possible measure of liberation behaviour is the per-centage of valuable phase, in this case chalcopyrite, that occurson the edges of simulated fragments, i.e. is exposed. The simulatedfragments were divided into liberation classes with 0, >0–10, >10–20 . . ., >90–<100, 100% chalcopyrite on the edges. Three exampleswith different copper grades (1.99%, 1.06% and 0.21% Cu) for75 lm simulated fragments are shown in Fig. 2. For the samplewith a copper grade of 1.99% simulated fragments with >0–10%and >10–20% chalcopyrite on the edges are dominant along witha significant number with 100% chalcopyrite on the border indicat-ing a bimodal distribution (Fig. 2A). In this sample 63% of the sim-ulated fragments have P20% chalcopyrite on their edges. For thesample with a copper grade of 0.21% most of the simulated frag-ments have >0–10% chalcopyrite on the edges. In this case only15% of the fragments have P20% chalcopyrite on their edges. The
Fig. 3. Example of down hole spatial variation of liberat
simulated fragmentation approach predicts a large difference inliberation behaviour for these three samples.
Other things being equal, i.e. reagents, hydrodynamic condi-tions for flotation etc., it is likely that a 75 lm fragment with sayP30% chalcopyrite on the edges will float. The amount of chalco-pyrite in each of these simulated fragments can be determinedfrom the mineral maps and used to calculate the liberation poten-tial. Fig. 3 shows an example of this approach; for this particulardrill hole interval only a few samples have a texture that suggestpoor liberation potential.
3.2. Comparison to typical MLA method
Liberation potential estimates determined using the simulatedfragmentation approach were compared to results for the samesamples obtained using typical MLA methods. As can be seen inFig. 4 there is reasonable correlation with the MLA method, espe-cially for the +38–150 lm particles.
3.3. Comparison to small scale flotation
Liberation potential estimates determined using the simulatedfragmentation approach were also compared to results determinedusing a small scale flotation device. Table 1 shows the results fromthe JKMSI and that the Ore A, the one with the greatest liberationpotential, achieved the highest recovery. Notably the grades fromboth size fractions were equivalent and the recoveries were simi-lar, indicating that this was a relatively coarse grained ore. For OresB and C, the grades and recoveries were much lower for the coarsesize fraction than the finer fraction indicating that in addition tothe lower grades the texture was more disseminated. The lowestliberation potential was for Ore C which correlated to the lowestflotation recovery. The ranking of liberation potential for the threesamples is the same as that determined using simulated fragmen-tation (Fig. 5).
3.4. Comparison to Cu grade
A different way to show this ‘‘liberation index’’ is to plot the lib-eration potential against Cu grade (Fig. 6). For samples with a cop-per grade >�0.7% between 55% and 85% of chalcopyrite is predicted
ion potential based on 75 lm simulated fragments.
Fig. 4. Comparison of liberation potential of chalcopyrite as determined bysimulated fragmentation and MLA analysis. Chalcopyrite with P30% of boundaryexposed.
Fig. 5. Comparison of liberation potential determined by simulated fragmentationand MSI analysis.
Fig. 6. Liberation potential predicted by simulated fragmentation versus copperassay for each 2 m-long drill core sample.
J. Hunt et al. / Minerals Engineering 24 (2011) 1271–1276 1275
as having a high liberation potential for samples from all five drillholes. For lower grade samples the liberation potential is highlyvariable. This variation is expected to make a large difference tothe recovery of chalcopyrite in these rocks and could be used to de-cide which lower grade intervals can be included as ore and whichare likely to be uneconomic to mine and/or process.
4. Conclusions
The liberation potential of lower grade samples is highly vari-able and it is in this area that significant improvements in the pre-diction of recovery may be possible. If low grade samples withpredicted good recovery have spatial continuity they may be amineable resource.
New research suggests that analysis of mineral maps is a viableway of rapidly and inexpensively producing data that can be usedto rank samples with respect to liberation potential and hence
Table 1Results of MSI analysis plus data for 2 m samples and 75 lm simulated fragments.
Sample Size fraction(lm)
Mass recovery(%)
Cu assay (%)concentrate
Cu recovery (%)calc)
C 38-106 1.86 1.36 13.75B 38-106 1.96 11.2 20.66A 38-106 3.19 21.8 37.14
C 106-212 1.25 0.62 5.75B 106-212 1.09 7.61 9.10A 106-212 2.83 21.5 33.95
recovery. This would allow sufficient volume of data to be collectedto allow incorporation into mine and mineral process planning(e.g. in a block model) and scheduling. The trends obtained havebeen confirmed using a small scale, rapid flotation test (JKMSI).The validation of these measurements is continuing in additionto the development of techniques with less computationaloverhead.
Acknowledgements
This research is part of a major collaborative geometallurgicalproject being undertaken at CODES and SES (University of Tasma-nia), JKMRC, BRC and CMLR (Sustainable Minerals Institute, Uni-versity of Queensland) and Parker Centre CRC (CSIRO).
(tails Std dev on mass(%)
2 m interval Cuassay (%)
75 lm SFs-Liberationpotential (%)
0.21 0.21 250.30 1.06 560.51 1.99 74
0.10 0.21 250.05 1.06 560.36 1.99 74
1276 J. Hunt et al. / Minerals Engineering 24 (2011) 1271–1276
The authors acknowledges financial support and permission topublish from industry sponsors of the AMIRA International P843and P843A GEMIII Projects – Anglo Gold Ashanti, Anglo American,ALS, Barrick, BHP Billiton, Boliden, CAE Mining (Datamine), Code-lco, Geotek, Gold Fields, Golder Associates, ioGlobal, Metso Miner-als, Minera San Cristobal, Newcrest, Newmont, OZ Minerals,Penoles, Quantitative Geoscience, Rio Tinto, Teck, Vale and XstrataCopper (MIM).
Financial support is also being provided by the Australian gov-ernment through the CODES ARC Centre of Excellence in OreDeposits and CRC ORE. The analytical facilities at the Central Sci-ence Laboratory, University of Tasmania, were used for MLA mea-surements in this project.
We thank two anonymous reviewers and the editorial staff fortheir input and comments.
References
Berry, R.F., 2008. Automated mineral identification by optical microscopy. In: Hunt,J.A., (Ed.), Technical Report 1 (AMIRA Project P843 Geometallurgical Mappingand Mine Modelling), pp. 7.1–7.11.
Berry, R.F., McMahon, C., 2008. Automated mineral identification by opticalmicroscopy: Ernest Henry, Aqqaluk. In: Hunt, J.A. (Ed.), Technical Report 2(AMIRA Project P843 Geometallurgical Mapping and Mine Modelling), pp. 6.1–6.13.
Definiens, 2008a. Definiens website: <www.definiens.com>.Definiens, 2008b. Software Manual, pp. 15–16.Hunt, J., Berry, R., Walters, S., 2010. Using mineral maps to rank potential processing
behaviour. International Mineral Processing Congress (IMPC 2010), Brisbane,Congress Proceedings, pp. 2899–2905.
Gu, Y., 2003. Automated scanning electron microscope based mineralliberation analysis – an introduction to JKMRC/FEI Mineral LiberationAnalyser. Journal of Minerals & Materials Characterization & Engineering2 (1), 33–41.