HOME
THE USE OF AUTOMATED MINERALOGY TO INTERPRET THE BATCH FLOTATION PERFORMANCE OF MERENSKY REEF ORE M Becker1, P Harris2, J Wiese3, K Corin4 and D Bradshaw5 ABSTRACT Merensky Reef ores are processed by flotation to recover the valuable platinum group elements and minerals, as well as base metal sulfides and selected reagents are added to optimise this process. Batch flotation tests are one of the key tools used to evaluate and understand the different interactions taking place between the reagents and minerals. Metallurgical performance is generally evaluated from the extrapolation of chemical assays that are routinely available, but which does not allow investigation of the types of particles being affected under the different flotation conditions. With the advent of quantitative mineralogical analyses (MLA, QEMSCAN) on metallurgical samples, these previously qualitative interpretations of flotation performance can now be quantified. Therefore, the aim of this study was to evaluate the use of chemical assays to interpret the batch flotation performance through the use of automated mineralogy in the context of a larger study that investigates the effect of depressant addition on the flotation of the base metal sulfides on a Merensky Reef ore. The QEMSCAN results showed that calculations of total gangue recovered to the concentrate based on chemical assays were shown to be relatively accurate. Interpretation of chalcopyrite and pentlandite flotation performance based on chemical assays was good, although the interpretation of pyrrhotite performance was complicated by the presence of pyrite. The base metal sulfides recovered were over 70 per cent liberated. No reduction was observed in the floatability of liberated chalcopyrite and pentlandite in the two size fractions investigated, due to the addition of depressant. The inadvertent loss of pyrrhotite during flotation however, was accounted for by the effect of depressant addition on froth stability. This study has shown the usefulness of both the extrapolation of chemical assay and mineralogical analysis in evaluating the results of batch flotation tests, although the results need to be interpreted with care. Keywords: liberation, depressants, automated mineralogy, pyrrhotite
INTRODUCTION The Merensky Reef in South Africa, discovered in 1924 by Dr Hans Merensky represents one of the South Africa’s most important platinum group element (PGE) ore deposits. In this deposit, the valuable platinum group elements (PGE) are hosted by the base metal sulfide (BMS) minerals, either as discrete platinum group minerals (PGM) or in solid solution with the BMS minerals; primarily pentlandite, pyrrhotite and chalcopyrite (eg Schouwstra, Kinloch and Lee, 2000; Ballhaus and Sylvester, 2000). The valuable PGE are typically recovered by froth flotation as a bulk sulfide float. However, there is considerable local variation in the degree of association of the PGE with the BMS minerals, as well as in the relative abundance and texture of the major gangue minerals (eg orthopyroxene, plagioclase, talc) along the strike of the reef and therefore each of the mining and processing operations along the strike of the reef has over time, developed their own particular flotation reagent suite to process the ore. Typically batch flotation testing has been used as the work horse used to evaluate these different reagent suites on the Merensky Reef, and therefore it is of importance to be able to extract as much valuable information from these tests as possible. 1. Centre for Minerals Research, Department of Chemical Engineering, University of Cape Town, Private Bag, Rondebosch 7701, South Africa. Email:
[email protected] 2. Centre for Minerals Research, Department of Chemical Engineering, University of Cape Town, Private Bag, Rondebosch 7701, South Africa. Email:
[email protected] 3. Centre for Minerals Research, Department of Chemical Engineering, University of Cape Town, Private Bag, Rondebosch 7701, South Africa. Email:
[email protected] 4. Centre for Minerals Research, Department of Chemical Engineering, University of Cape Town, Private Bag, Rondebosch 7701, South Africa. Email:
[email protected] 5. Centre for Minerals Research, Department of Chemical Engineering, University of Cape Town, Private Bag, Rondebosch 7701 South Africa; also Julius Krutschnitt Mineral Research
Centre, University of Queensland, Indooroopilly Qld 4068, Australia. Email:
[email protected]
XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2797
M BECKER et al
The Centre for Minerals Research at the University of Cape Town has been involved in an ongoing research programme on the different flotation reagents used in the processing of the Merensky Reef ore for numerous years (eg Wiese, Harris and Bradshaw, 2005). The focus of the research has been on understanding the effects of the chemical interactions between different collectors, activators and polysaccharide depressants during the flotation of the BMS. Based on the extrapolation of chemical assays which are routinely available, the effects on the different BMS and gangue minerals are evaluated. This method however has several short comings primarily that one cannot investigate the effect of the reagents on the different particle types recovered (eg liberated versus locked BMS). With the advent of quantitative mineralogical analysis such as MLA or QEMSCAN (eg Fandrich et al, 2007; Gottlieb, 2008), these problems can be overcome and already numerous PGM based case studies of the use of this type of instrumentation exist in the literature (eg Nel, Valenta and Naude, 2005). These instruments are of particular use for the analysis of plant based PGM samples, especially for the analysis of tailings samples due to the ease with which these instruments are able to locate the valuable PGMs based on the degree of their brightness of back scattered electrons (Brown and Dinham, 2006; Fandrich et al, 2007). The objective of this paper is to compare the use of chemical assays with results obtained from quantitative mineralogical analysis (QEMSCAN). This will not only allow validation of the methodology used at the CMR for the interpretation of the results on chemical analysis, but also an investigation of the effect of depressant addition on the flotation of the different BMS particle types. Additional outcomes of the work will include an evaluation of the statistical and practical validity of using automated mineralogical techniques on batch flotation samples where very small masses of material are generally recovered.
EXPERIMENTAL DETAILS Merensky Reef ore was sourced from the Impala platinum mine as run of mine ore and has been used in an extensive programme investigating the floatability this ore (eg Wiese, Harris and Bradshaw, 2005). The bulk sample was crushed, blended, riffled and split using a rotary splitter into 1 kg subsamples at the Centre for Minerals Research (CMR) at the University of Cape Town prior to batch flotation tests. Batch flotation tests were conducted using the standard procedure as outlined in Wiese, Harris and Bradshaw (2005) at a grind of 60 per cent passing 75 μm. Four successive timed concentrates were collected. Batch flotation tests were conducted at two conditions; with and without a modified guar gum depressant (300 g/t guar). Further details of the reagents used in batch flotation tests are given in Becker et al (2009). Feed and concentrate samples were subsequently submitted for analysis at Mintek in Johannesburg on QEMSCAN. A common feed sample was analysed in four size fractions, namely +5/-20, +20/-38, +38/-75, and +75 μm using the Bulk Mineralogical Analysis (BMA) routine in QEMSCAN (Goodall, Scales and Butcher, 2005; Gottlieb, 2008). Concentrates from quintuplicate batch flotation tests were combined and wet screened to the same size fractions. Only samples from the +20/-38 and +38/-75 μm size fraction were selected for further QEMSCAN analysis. The choice of size fractions for mineralogical analysis was based on the selection of particles in the optimum size range for true flotation with minimal contribution from entrainment (Savassi, 1998). Concentrate samples were run using the BMA and specific mineral search (SMS) routine on the same instrument. Since the majority of the sized concentrate samples were less than 10 g, care was taken to ensure sufficient sulfide particles were analysed. Typically between 5000 and 10 000 sulfide particles were analysed, although for some of the sized concentrates from the tests without depressant only ~ 2000 sulfide particles were found. The complete modal mineralogy of the feed and concentrate samples are given in Becker et al (2009). Chemical assays of all concentrates were obtained using AAS and a Leco Sulfur Analyser at UCT. No platinum group element analyses were performed in this study.
RESULTS AND DISCUSSION Comparison of chemical and mineralogical analyses Prior to any analysis or interpretation of the mineralogical results, the routinely accepted methodology comparing the actual chemistry with the calculated chemistry from QEMSCAN analyses was used to ensure the integrity of the mineralogical analyses. Comparison of the actual chemical assays (Cu, Ni, S) with the calculated assays based on the QEMSCAN results was relatively good for the majority of the concentrate samples (Figure 1). Some minor discrepancies were noted for sulfur in some of XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2798
THE USE OF AUTOMATED MINERALOGY TO INTERPRET THE BATCH FLOTATION PERFORMANCE OF MERENSKY REEF ORE
FIG 1 - Comparison of the calculated QEMSCAN assay versus actual chemical assay from the QEMSCAN analyses of all concentrate samples from batch flotation tests of Merensky Reef ore. Results are shown for tests with and without depressant addition (guar), for both the +38/-75 and +20/-38 μm size fractions.
the final concentrates from the flotation tests (Concentrates 3 and 4). This is considered a function of the chemical assay, given that the slow floating pyrrhotite is reactive towards oxidation and the formation of associated oxidation products (Mycroft, Nesbitt and Pratt, 1995; Belzile et al, 2004). Following this initial step, the mineralogical analyses could be used to validate the interpretations from chemical analysis employed at the CMR to analyse the batch flotation performance of Merensky Reef ore. The first step used to analyse and interpret the batch flotation performance is to calculate the mass of gangue recovered during flotation by subtracting the mass of sulfides recovered during flotation from the total concentrate mass for each size fraction. It is well-known that the predominant base metal sulfide minerals in the Merensky Reef are pyrrhotite, pentlandite and chalcopyrite (eg Liebenberg, 1970; Schouwstra, Kinloch and Lee, 2000) and which was confirmed by the modal analysis of these concentrates (Becker et al, 2009). Using sulfur contents of 34.9 weight per cent for chalcopyrite (CuFeS2), 33.2 weight per cent for pentlandite ((Fe,Ni)9S8) and 39.3 weight per cent for pyrrhotite (Fe1-xS), it was assumed that the average sulfur content of all the base metal sulfides recovered during flotation is 36.5 weight per cent sulfur. On this basis, the total mass of sulfides recovered can be calculated and subtracted from the concentrate mass to obtain the gangue mass. Comparison of the gangue mass calculated from chemical assays with the actual gangue mass obtained from mineralogical analysis, as shown in Figure 2, shows good correlation for the +38/-75 and +20/-38 μm size fractions of the concentrates. The results also show that the effect of depressant
FIG 2 - Comparison of the cumulative mass of gangue recovered in batch flotation tests of Merensky Reef ore as measured using QEMSCAN (QEM) and calculated from chemical assays (chem). Results are shown for tests with and without depressant addition (guar), for the +38/-75 and +20/-38 μm size fractions. XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2799
M BECKER et al
FIG 3 - Cumulative mass recovery of pentlandite, pyrrhotite and chalcopyrite versus water from batch flotation tests of Merensky Reef ore calculated from chemical assays. Results are shown for tests with and without depressant addition (guar) for the two size classes combined (+20/-38 and +38/-75 μm).
addition is substantial since the mass of gangue recovered in these two size fractions is almost negligible (<5 g gangue recovered relative to a 5 kg feed sample). The next step used to validate the method employed at the CMR to analyse the batch flotation performance of Merensky ore based on chemical assays is based on the recovery of the individual sulfides. The results from the different tests were originally compared as a per cent recovery calculated from the QEMSCAN analyses of a common feed sample. However, since anomalous size by size recoveries were obtained it suggested that the quantitative mineralogical analysis of batch flotation samples where very small masses are analysed is not simple. It also showed that the analysis of a sub-sample from a 1 kg feed sample was not statistically representative of the feed from the quintuplicate flotation tests for the two conditions investigated. Therefore further comparisons of the mineralogical data were made on a mass recovery basis, a measurement which is not calculated from the feed analysis. Using nickel and copper assays which are routinely available, the recovery of pentlandite and chalcopyrite was calculated based on the nickel and copper contents of these minerals. It was also assumed that the nickel recovered is entirely hosted by pentlandite and that the recovery of nonsulfide nickel is negligible. Pyrrhotite recovery was calculated on the assumption that all sulfide sulfur not deported in pentlandite and chalcopyrite, could be accounted for by pyrrhotite. Figures 3 and 4 show the cumulative mass recovery of chalcopyrite, pentlandite and pyrrhotite calculated from
FIG 4 - Cumulative mass recovery of pentlandite, pyrrhotite, chalcopyrite and pyrite versus water from batch flotation tests of Merensky ore measured with QEMSCAN. Results are shown for tests with and without depressant addition (300 g/t guar) for the two size classes combined (+20/-38 and +38/-75 μm). XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2800
THE USE OF AUTOMATED MINERALOGY TO INTERPRET THE BATCH FLOTATION PERFORMANCE OF MERENSKY REEF ORE
FIG 5 - Copper and nickel grade recovery from batch flotation tests of Merensky Reef ore conducted with (300 g/t guar) and without depressant addition. Grade-recovery curves are shown for unsized concentrates. The standard error from replicate batch flotation tests is shown.
chemical assays (Figure 3) and from the mineralogical analysis (Figure 4). For chalcopyrite, the mass – water recovery curves between Figures 3 and 4 are very similar for one another. This is also true for pentlandite, although there is a slight underestimation of pentlandite recovery from chemical assays due to the assumption that all nickel recovered during flotation is hosted by pentlandite. The correlation between the calculated and actual pyrrhotite mass recovery however, is not as good due to the presence of minor pyrite (FeS2) in these concentrates (Figures 3 and 4). Since pyrrhotite and pyrite are both iron sulfides, there is no simple methodology which can be used to decouple the performance of these two minerals. This is exacerbated by the effect that pyrrhotite is a monosulfide, whereas pyrite is disulfide.
Effect of depressant addition on sulfide recovery Copper and nickel grade recovery curves from batch flotation tests with and without depressant addition are shown in Figure 5. It is apparent that the addition of guar depressant results in a decrease in both copper and nickel recovery (~5 per cent loss in Cu, 3 per cent loss in Ni), although there is a big improvement in concentrate grade (~2 per cent increase in Cu, 4 per cent increase in Ni). Since there is some loss in recovery of the sulfides with depressant addition, it is of interest to establish which of the sulfide particle types are affected by depressant addition. According to the results shown in Figure 4, there is an apparent enhancement in pentlandite flotation with the addition of guar depressant for the +20/-38 and +38/-75 μm size fractions combined. At first, this was interpreted as possible evidence of slime cleaning by the depressant, but then the representativeness of the results were reconsidered and the feed characteristics of the respective flotation tests re-examined. This showed a slightly higher average Ni feed grade in the tests with depressant addition (0.175 ± 0.016 wt per cent Ni), relative to those with no depressant addition (0.168 ± 0.12 wt per cent Ni). Therefore, the apparent enhancement in pentlandite flotation with guar addition was interpreted to be an artifact and not a significant feature, providing just one example how careful interpretation is needed when dealing with such small concentrate masses. No significant loss of chalcopyrite recovery was observed due to the addition of the guar depressant (Figure 4). Similarly, no loss of pyrite recovery was observed due to the addition of guar depressant. It is also of interest to note the similarity in flotation kinetics between chalcopyrite and pyrite in the Merensky Reef ore flotation. For pyrrhotite however, there does appear to be a significant loss due to the addition of depressant for ore. Therefore, it is of interest to explore the liberation characteristics of the sulfides and the effect of depressant addition on the different sulfide particle types. The liberation of the different sulfides is shown for reference in Table 1. Due to the limitations of the study however, Table 1 should only be treated as a guide to the degree of liberation and not as a quantitative measure. The results show that pentlandite is generally the most liberated sulfide (84 87 per cent liberated). Pyrrhotite and chalcopyrite liberation is generally fairly similar (69 - 83 per XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2801
M BECKER et al
TABLE 1 Mass per cent liberated pentlandite, pyrrhotite and chalcopyrite in concentrates from batch flotation tests with and without depressant (300 g/t guar) addition. Results are shown for only the +38/-75 and +20/-38 μm size fractions. Mineral
No depressant
300 g/t guar
+38/-75 μm
+20/-38 μm
+38/-75 μm
+20/-38 μm
Pentlandite
84
87
85
87
Pyrrhotite
75
83
69
78
Chalcopyrite
75
82
75
80
cent liberated). Since the majority of the sulfides recovered during flotation are liberated, the focus here will be on the effect of depressant addition on the liberated sulfides. Comparison of the mass – water recovery curves for liberated chalcopyrite in Figure 6 shows no significant difference with or without depressant addition. Similarly, the addition of guar depressant has no significant effect on the flotation of liberated pentlandite (Figure 6). Although the results are only shown in Figure 6 for the +38/-75 μm size fractions, similar trends occur in the +20/-38 μm fraction. For pyrrhotite however, there is a significant difference in mass recovery with depressant addition (Figure 6). This represents a ~34 per cent loss in pyrrhotite mass recovery in the +38/-75 μm size fraction, and a ~27 per cent loss in pyrrhotite mass recovery in the +20/-38 μm size fraction. It has already been noted from the final grade recovery curves from these batch flotation tests that the addition of depressant has some effect on the recovery of nickel (pentlandite) and copper (chalcopyrite). However, the results from the two size fractions investigated show no loss or recovery of liberated pentlandite or chalcopyrite with depressant addition. This indicates that the loss of pentlandite and chalcopyrite occurs within size fractions other those investigated and is most likely in the +75 μm size fraction. This is supported by the size by size chemical analysis of these samples. This reduction in copper and nickel recovery is attributed to the loss of unliberated sulfides hosted by gangue minerals in the coarser size fractions. Due to the fact that the results have been interpreted with respect to the water recovery which is an indication of froth stability during flotation, the loss of pyrrhotite recovery with depressant addition is clearly a froth effect. The slow flotation kinetics of pyrrhotite (eg Figures 3 and 4; Miller et al, 2005) as well as its propensity for oxidation and formation of hydrophilic oxidation products (eg Mycroft, Nesbitt and Pratt, 1995; Legrand, Bancroft and Nesbitt, 2005a) provide the reason for the fact that pyrrhotite and not chalcopyrite or pentlandite is affected by froth effects. In the absence of depressant addition, naturally floatable minerals such as talc tend to have a froth stabilising effect (eg Shortridge, 2002) resulting in increased water recovery. In order to reduce the recovery of naturally floating gangue (composite orthopyroxene – talc particles; Jasieniak and Smart, 2009;
FIG 6 - Cumulative mass recovery of liberated sulfides (chalcopyrite, pyrrhotite and pentlandite) versus water from batch flotation tests of Merensky Reef ore. Results are shown for tests with and without depressant addition (300 g/t guar), for the +38/-75 μm size fraction. XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2802
THE USE OF AUTOMATED MINERALOGY TO INTERPRET THE BATCH FLOTATION PERFORMANCE OF MERENSKY REEF ORE
Becker et al, 2009) in Merensky ore flotation, depressants are added. This causes a reduction in the amount of talc reporting to and stabilising the froth (ie lower water recovery), and which consequently results in lower pyrrhotite recoveries.
CONCLUSIONS The extrapolation of chemical assays into calculated total gangue and base metal sulfide recoveries is an appropriate first order measure which can be used to interpret the batch flotation performance of Merensky Reef ores. Good correlation existed in the comparison between the chemical and actual mineralogical results in terms of the calculation of total gangue, chalcopyrite and pentlandite recovery. The calculation of pyrrhotite recovery was complicated by the presence of pyrite. The base metal sulfides recovered during batch flotation of the Merensky Reef ores were over 70 per cent liberated. The addition of guar depressant had no significant effect on the recovery of liberated chalcopyrite and pentlandite in the +38/-75 and +20/-38 μm size fractions. This indicates that the total reduction of copper and nickel recovery with depressant addition is likely related to the loss of composite unliberated sulfides in the coarser size fractions (+75 μm). The loss of liberated pyrrhotite with depressant addition is related to froth effects and the slow floating nature of pyrrhotite relative to the other base metal sulfides. This difference in froth effects due to depressant addition is linked to the presence of naturally floating gangue comprising composite orthopyroxene and talc particles in Merensky ore. Automated mineralogical analyses can be applied to the analyses of metallurgical samples from batch flotation tests where very small concentrate masses are recovered. However, special care needs to be taken with respect to the representativeness of the samples presented to the instrument, and in the interpretation of the results. These analyses provide valuable insight into the types of BMS particles that are or are not affected by depressant addition.
ACKNOWLEDGEMENTS Sincere appreciation goes to Impala Platinum for funding the QEMSCAN analyses and to Mintek for running the analyses. Members of the UCT Reagent Research Facility are also acknowledged for their continued support of this ongoing research: Anglo Platinum, Impala Platinum and Lonmin Platinum.
REFERENCES Ballhaus, C and Sylvester P J, 2000. Noble metal enrichment processes in the Merensky Reef, Bushveld Complex, Journal of Petrology, 41:545 - 561. Becker, M, Harris, P J, Wiese, J G and Bradshaw, D J, 2009. Mineralogical characterisation of naturally floatable gangue in Merensky reef ore flotation. International Journal of Mineral Processing, 93, 246 - 255. Belzile, N, Chen, Y-W, Cai, M-F and Li, Y, 2004. A review of pyrrhotite oxidation, Journal of Geochemical Exploration, 84:65 - 76. Brown, M and Dinham, P, 2006. Benchmark quality investigation on automated mineralogy: A discussion of some preliminary data, in Proceedings MEI Automated Mineralogy 07 (Brisbane). Fandrich, R, Gu, Y, Burrows, D and Moeller, K, 2007. Modern SEM-based mineral liberation analysis, International Journal of Mineral Processing, 84:310 - 320. Goodall, W R, Scales, P J and Butcher, A R, 2005. The use of QEMSCAN and diagnostic leaching in the characterisation of visible gold in complex ores, Minerals Engineering, 18:877 - 886. Gottlieb, P, 2008. The revolutionary impact of automated mineralogy on mining and mineral processing, in Proceedings 24th International Mineral Processing Congress, pp 165 - 174 (Science Press: Beijing). Jasieniak, M and Smart, R S C, 2009. Collectorless flotation of pyroxene in Merensky ore: Residual layer identification using statistical ToF-SIMS analysis, International Journal of Mineral Processing, 92:169 - 176. Legrand, D L, Bancroft, G M and Nesbitt, H W, 2005. Oxidation/alteration of pentlandite and pyrrhotite surfaces at pH 9.3: Part I. Assignment of XPS spectra and chemical trends, American Mineralogist, 90:1042 - 1054. Liebenberg, L, 1970. The sulphides in the layered sequence of the Bushveld Igneous Complex, in Proceedings Symposium on the Bushveld Igneous Complex and other layered intrusions, pp 108 - 207 (GSSA: Johannesburg).
XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2803
M BECKER et al
Miller, J A, Li, C, Davidtz, J C and Vos, F, 2005. A review of pyrrhotite flotation chemistry in the processing of PGM ores, Minerals Engineering, 18:855 - 865. Mycroft, J R, Nesbitt, H W and Pratt, A R, 1995. X-ray photoelectron and Auger electron spectroscopy of air-oxidized pyrrhotite: Distribution of oxidized species with depth, Geochimica et Cosmochimica Acta, 59:721 - 733. Nel, E, Valenta, M and Naude, M, 2005. Influence of open circuit regrind milling on UG2 ore composition and mineralogy at Impala’s UG2 concentrator, Minerals Engineering, 18:785 - 790. Savassi, O N, 1998. Direct estimation of the degree of entrainment and the froth recovery of attached particles in industrial flotation cells, PhD thesis (unpublished), University of Queensland. Schouwstra, R P, Kinloch, E D and Lee, C A, 2000. A short geological review of the Bushveld Complex, Platinum Metals Review, 44:33 - 39. Shortridge, P G, 2002. The influence of polymeric charge and structure, molecular weight and ionic conditions on depressant ability to reduce the natural floatability of talc, MSc thesis (unpublished), University of Cape Town. Wiese, J G, Harris, P J and Bradshaw, D J, 2005. The influence of the reagent suite on the flotation of ores from the Merensky Reef, Minerals Engineering, 18:189 - 198.
XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010 PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
2804