Platinum group elements (PGEs) are increasingly being used in a variety of environmentally-related technologies such as catalysts and catalytic converters which have strong expected growth to meet environmental and technological challenges this century. The platinum industry is actively seeking to progress its commitment to sustainability principles by reducing the negative impacts of their mining and mineral processing operations. Technical innovation to improve future plant designs, as well as the development of management policies, guidelines and protocols for efficient operation of process plants has therefore become a strategic priority for the South African platinum industry. The industry has also made an effort to understand the environmental impacts of its products from mine to metal, using life cycle methods. However, very limited research has been done to investigate what environmental value could be created if strategic and design decisions in minerals processing were life cycle based, particularly in the context of PGMs. Seminal work by Stewart (1999) investigating the environmental life cycle consideration for design-related decision making in the minerals industry has not led to significant adoption. Forbes et al. (2000) analysed metal processing using LCA and were able to identify opportunities for improved environmental performance. They however did not explore how it would be incorporated into the decision making cycle. Therefore, the main objective of this research is to determine whether life cycle assessment could help inform design decision making in the minerals industry. In the years 2002-2008 several PGM-producing companies commissioned new S02 scrubbing technologies to meet the regulations that had been set to prevent the release of excessive amounts of sulphur dioxide from smelters in the Rustenburg area, a mining town located in the North West Province of South Africa. Using these cleanup process retrofits as case studies, this dissertation aims to determine whether the introduction of LCA as an environmental analysis tool would have provided additional value to the decision makers. The case study approach that was chosen compared and assessed the performance of S02 abatement technologies and the effect of efficiencies chosen on environmental performance by using life cycle assessment modelling. By doing the life cycle assessment on the different options that the companies had, it was possible to evaluate the indirect environmental impacts that could have been overlooked during the design decision making process. In addition , experts who were involved in the design processes of the S02 abatement retrofits were interviewed to establish: i) how the design decisions were made and ii) whether the life cycle based insights into technology performance would have been of use in the design work. The goal of the life cycle assessment was to identify whether there were design decisions that induced environmental burden shifting when platinum smelters in the Rustenburg area added S02 abatement technologies to their processes, which could have been avoided had the LCA perspective been taken into account. The assessments considered two key variables, namely extent of recovery and technology choice. The study shows that the energy requirements increase exponentially with increasing recovery for both technology options. This is a result of the increased pumping energy requirements which are directly related to the increasing quantities of solvent that have to be pumped. The environmental impacts that were analysed during the Life Cycle Impact Assessment (LCIA) phase were: abiotic resource depletion, fossil fuel depletion, acidification, global warming , human toxicity and water depletion. The background processes for soda ash and lime production dominated abiotic resource depletion, fossil fuel depletion , global warming potential and human toxicity impacts that were associated with the concentrated dual alkali process. The foreground system had significant effects on the acidification potential, water depletion as well as global warming potential. For the scrubber with acid plant, the transportation of the acid that was produced and the sulphuric acid production used in the system expansion were observed to have a major impact on most of the impact categories, with the exception being the acidification potential and water depletion in which the foreground system was the principal contributor to the impact categories. The magnitude of impacts increased with increasing recovery in the concentrated dual alkali case with the only exception being the acidification potential. For the scrubber with acid plant option, the magnitude of impacts decreased with increasing recovery with the exception being the amount of water used and the global warming potential. Overall, the LCA revealed that the scrubber with acid plant choice mostly has significantly lower environmental impacts. The results of the LCA were then presented to design experts as way of gaining insights as to whether or not carrying out a life cycle assessment during the design phase would be capable of informing and influencing design decision making in mining companies. The interviews were also used as a platform to gain a better understanding of how design decision making works in the mining sector. The expert interviews revealed that the decision making process is not an individual job but rather requires input from different project teams. Before a decision is made they would all need to agree unanimously on a specific technology option which they would all deem beneficial after carrying out a cost benefit analysis. With regards to decisions on retrofitting , the major drivers were identified to be the case in which design specifications were not being met, surfacing of new regulations or availability of improved technology. With regards to choosing between technologies that satisfied the same purposes the interviewees felt that the main determinant would be the quality of gas to be treated. Once this had been established, the company would then decide on whether or not they wanted to opt for a high OPEX or high CAPEX process depending on their financial stability. Most of the interviewees felt that companies did not do much to incorporate environmental concerns into their design apart from doing the prescribed Environmental Impact Assessment (EIA). Therefore once presented with the LCA results from the study they felt that such an assessment would be really useful especially if it were to be incorporated during the early stages of the design. By so doing the environmental aspects would gain more weighting in the decision making matrix that is used. The major concern that was brought up was that in as much as the LCA quantified the impacts associated with the different options, and a comparison was made between the different key variables, it would be difficult to make decisions without also including a rational and consistent normalisation process. This would help decision makers see the relevance of the impacts presented to them, by relating them to some form of reference system. It is concluded that in the case analysed, LCA would have generated useful further insights to the design team on the technology and design variable choices. Additionally, there would be some interest from design decision-makers to include such insights into design projects if this could be done without introducing significant extra work or delays.

Chalcopyrite is not only the most abundant of the copper sulphides, but also the most stable, making it recalcitrant to hydrometallurgical treatment processes especially in atmospheric leaching. Hence, pyrometallurgical processes are traditionally used to treat chalcopyrite concentrates. However, ore grades are falling and concentration processes are becoming increasingly costly, prompting need to revisit hydrometallurgical treatment processes (especially heap leaching), which are otherwise regarded as relatively economic and environmentally friendly. Key hydrometallurgical processes for chalcopyrite treatment are ferric sulphate, chloride and ammoniacal systems. The ferric sulphate system does not work well under atmospheric conditions, except in combination with thermophilic microorganisms, whereas the chloride system has only recently been evaluated more seriously for heap leach processes. The ammonia system remains relatively unexplored and most studies date back more than 40 years, but the system has considerable potential for further development. Ammonia systems can be effectively used to leach copper from chalcopyrite in the presence of an oxidant. The ammoniacal leaching system is heavily reliant on a good surface mass transfer system, hence it being widely studied in high pressure systems where oxygen was accepted to be the oxidant. Leach reactors were designed to use agitation systems which promote the abrasion of an iron based deposit layer thought to passivate the mineral surface. Most research on the ammonia leaching systems has previously been carried out in controlled or bulk leaching studies and only a few used electrochemical studies. A disconnect exits between the two approaches, resulting in different proposed fundamental reaction mechanisms and kinetic understanding. A fundamental electrochemical and controlled leach study of the oxidative leaching of chalcopyrite in ammoniacal solutions has been undertaken. The study covered the following aspects: a description of the mixed potentials, chemistry and kinetics of the anodic reaction, the cathodic reduction of the oxidants, the formation and effect of surface deposits and lastly a look at how results from electrochemical studies compare to those from the leaching of a similar mineral sample under similar solution conditions. A detailed study of the mixed potentials on a more or less pure chalcopyrite electrode has shown the redox reactions on the surface of the mineral to be controlled by the oxidation of chalcopyrite and reduction of copper(II). The presence of oxygen has been found to have no significant effect on mixed potentials in ammoniacal solutions in the presence of initial copper(II). Constant potential and potentiodynamic studies on the anodic reaction have shown the rate of the anodic reaction to increase with an increase in potential in a standard 1M ammonia/ammonium sulphate solution (which buffers at pH 9.6) in exponential fashion supporting conventional Butler- Volmer behaviour with a anodic transfer coefficient of 0.42 and a rate constant k* CuFeS2 of 0.0431 cms-1. Increasing total ammonia increased the rate of reaction only at low concentrations; at higher concentrations increasing total ammonia had no effect on the anodic reaction. An increase of pH at fixed total ammonia concentration showed an increase in reaction rate, but the effect cannot clearly be discerned from the concomitant shift in relative proportion of free NH3 and NH4 +. Coulometric studies have shown the oxidation reaction to proceed via the formation of a thiosulphate intermediate and this to be a 7-8 electron transfer reaction. A surface deposit layer consisting of iron, oxygen and small quantities of sulphur was formed and the sulphur component of this product layer was seen to be gradually depleted during leaching. Anodic currents were found to gradually decrease with time and this was linked to the growth of the surface deposit layer. However, the surface deposit layer did not passivate the anodic reaction; instead, it was proposed that the surface deposit layer adsorbed copper ions and displayed “ohmic” behaviour. The formation of the surface deposit layer was found to apparently promote the cathodic reduction of copper(II). While reduction of copper(II) was shown to be the primary reduction reaction, the presence of oxygen was seen to promote this reduction reaction through the regeneration of copper(II) in experiments that ran for longer time periods. An apparent accumulation of copper(I) on the mineral surface was seen to adversely affect the rate of the cathodic reaction and thus the overall rate of dissolution. The nature and morphology of the surface layer was found to be significantly influenced by the choice of cation in solution, which was thought to influence primarily the complexation/precipitation of ferric species forming near the surface. The degree of agitation during leach studies influences the rate of leaching due to the fragmentation of surface deposits, which are seen to slow the anodic reaction. A kinetic model has been developed for the anodic and cathodic reactions. This thesis presents significant new findings regarding the role of the copper(I)/copper(II) redox couple on the oxidative leaching of chalcopyrite. It also highlights the potentially limiting role of the cathodic reactions which have frequently been overshadowed by the focus on chalcopyrite oxidation reactions. Furthermore, the growth of a surface inhibiting layer which cannot be removed in heap leach systems due to the lack of mechanical agitation can now potentially be addressed by looking into the complexation and precipitation characteristics of cations in solution for ammoniacal leach systems.

As the excessive extraction of high grade iron ore reserves has led to the rapid depletion of these ore bodies, there is a growing need to extract and upgrade low grade iron ores into more economically viable products with an iron content in excess of 50%. The beneficiation of low grade iron ores through the reverse cationic flotation procedure is gradually gaining popularity as a possible processing route of the future for South Africa’s iron industry. Reverse cationic flotation employs a reagent suite consisting of an amine compound which functions as a quartz collector in addition to providing the frothing effect in the flotation system, and hydrolysed starch which serves to depress hematite during flotation. The aim of this project was to investigate the effect of using five amine collectors with different molecular structures on the flotation recovery of quartz and the entrainment of hematite in the flotation of a South African iron ore and a Brazilian iron ore. Laboratory batch flotation tests were conducted on both ore samples and the grade and recovery of hematite were recorded. The collectors were characterised through surface tension measurements and pKa value analysis. An attempt at using different polysaccharides as hematite depressants entailed the use of a CMC and a guar gum in batch flotation tests of the Brazilian iron ore. With regards to the batch flotation of a South African iron ore, results showed that at the grind size of 80% passing 150 μm, there was no selective separation of quartz from hematite. When the most commonly employed collector, an ether monoamine (Flotigam EDA) at a dosage of 125 g/t was used in flotation, the iron grade changed from 39% to 40% whilst the quartz grade dropped from 38% to 36%. This change in grade was insignificant when experimental errors were considered. Flotation of a mill product with a grind size of 80% passing 75 μm using a hybrid flotation cell also exhibited poor flotation results when the same reagent suite was applied. The iron grade changed marginally from 39% before flotation to 41% after flotation. The ether diamine 1 (Flotigam 3135) and a blend of the Flotigam EDA and Flotigam 3135 at a dosage of 125 g/t also failed to selectively float quartz from hematite. The remaining collectors (ether diamine 2, an imidazoline and a quaternary ammonium salt) did not yield any mass recovery to the froth zone, therefore, they were the worst performing collectors in the batch flotation tests of a South African iron ore. QEMSCAN analysis of the South African iron ore revealed that the ore was texturally complex with finely disseminated mineral phases in the sizes of -20 μm. Poor liberation of the quartz and the hematite led to the poor flotation performance. Further milling of the ore merely produced high amounts of slimes which would lead to high losses of hematite through the entrainment of hematite during batch flotation. Batch flotation results from the flotation of a Brazilian iron ore at the standard grind size of 80% passing 150 μm yielded good flotation results for all five collectors that were investigated. Surprisingly, the ether diamine 2 (Flotigam 2835-2L) performed best as a collector for the Brazilian iron ore although it failed to perform in the flotation of the South African iron ore. A hematite concentrate with an iron grade of 65% and a quartz grade less of 1.5% was attained. At the same collector dosage of 125 g/t, the superiority of alky ether amines over the quaternary ammonium salt and the imidazoline was evident in the final iron grade and recovery. QEMSCAN analysis of the ore sample revealed that both quartz and hematite were above 90% liberated explaining the good response to reverse flotation using cationic collectors. In both mineral systems, entrainment of hematite was influenced by the type of collector employed as they differed in frothing strength. The use of a collector with a strong froth stabilising effect resulted in a higher degree of entrainment as seen when ether diamine 1 (Flotigam 3135) was utilised in flotation. The addition of an alcohol frother in systems employing the quaternary ammonium salt and the imidazoline salt as quartz collectors introduced a strong frothing effect to the system as well as an enhancement in quartz recovery to the froth zone. The presence of a neutral species in a flotation system either through the addition of an alcohol or the manipulation of chemical conditions i.e. the pH such that the alkyl ether amines generate neutral molecular species and ion-molecule complexes is necessary in order to achieve good flotation results. The neutral species and the ion-molecule complexes assist with frothing and adsorption of cationic collectors onto the quartz surface. Batch flotation using a CMC and a guar gum as hematite depressants showed that starch remains a superior hematite depressant. At the same depressant dosage of 1500 g/t a system using starch as a depressant yielded a hematite concentrate with an iron grade and recovery of 65% and 97%, respectively. The CMC and guar gum yielded a hematite concentrate with iron grades of 51% and 66% respectively and iron recoveries of 88% and 80%, respectively. The CMC, employed at a dosage of 1500 g/t, had a dispersive effect which resulted in more hematite reporting to the froth phase hence the lower iron recovery after flotation. At lower depressant dosages i.e. 250 g/t and 500 g/t, the system using a guar gum attained a higher iron grade of 63%. The results confirmed the strong flocculation effect of a guar gum in higher concentrations which results in the agglomeration of quartz causing it to remain in the pulp zone. In general, the study showed that the South African iron ore under investigation was not amenable to flotation. Further studies into the processing of ultrafine ore will be required if the desired iron grade is to be attained. Another key finding is that the collectors’ molecular chemistry alone does not accurately determine their flotation performance as their flotation behaviour can differ depending on the mineral system.

Access to decent sanitation remains a problem in developing countries. At the same time, sanitation technology is constantly evolving specifically regarding resource recovery solutions. Some chemical elements found in human excreta derived from non-renewable resources, and the recycling of phosphorous from sewage in particular is a possible solution to the growing issue of resource scarcity. A potential way to recover phosphorous from urine or water-borne sewage is through struvite precipitation. Struvite (MgNH4PO4. 6H2O) is a mineral that can be used as a slowrelease magnesium, ammonium and phosphate based fertilizer and can be produced from urine by adding magnesium to the ammonium and phosphate rich urine. Usually, magnesium is dosed chemically using salts such as MgCl2, MgO, MgSO4 or bittern, together with pH regulating agents but these reactants produce unfavourable chemical by-products and the process tends to be expensive. Previous studies have proven that electrochemical dosing of magnesium is a feasible and reliable method of struvite precipitation. It not only produces high grade struvite that is valuable and marketable, but it also eliminates the need for alkalinity dosing in order to create a suitable pH environment for struvite precipitation. Further to that, electrochemical precipitation does not produce any harmful chemical by-products. Previous work shows that one main challenge that is associated with this method is the formation of a mineral layer on the magnesium anode called nesquehonite (MgCO3 · 3H2O). This leads to increased electrode potentials and hence high energy consumptions and may also lead to system failures of the reactor. Further to that, struvite generally precipitates as small crystals that are difficult to separate from the solution, leading to low mass recoveries of the product. These small crystals are formed as a result of the high supersaturation, which generally occurs for most processes that are employed to make struvite. In view of these problems, this dissertation presents an investigation of the potential improvements to the electrochemical precipitation of struvite from source-separated urine. The main aim is to minimise or eliminate the formation of mineral precipitates on the anode surface. It also looks into ways of increasing the crystal sizes of the struvite being precipitated in the electrochemical system. The methodology for this investigation involved modelling and experimental work. The specific objectives for this study were to: a) Investigate how thermodynamic modelling of struvite precipitation compares to the experimental results from an electrochemical precipitation reactor, b) Employ the aspect of seeding in an electrochemical reactor for struvite production and determine the technical feasibility of the proposed process, c) Establish how to minimise the formation of nesquehonite so that the quality of struvite produced in the electrochemical reactor is not compromised, d) Investigate how the crystal sizes of the struvite particles produced in the seeded electrochemical precipitation batch reactor setup compare to those produced in the continuously stirred reactor setup with a recycle that gives the particles a longer residence time, e) Investigate the economics and energy requirements of the SEP (Seeded electrochemical precipitation reactor). The results of the thermodynamic model suggested that the Mg:P molar ratio (magnesium to phosphorous molar ratio), the pH of the reaction environment and the presence of the constituent ions in adequate quantities play a critical role in the precipitation of struvite with regards to yield and purity. The model predicted that phosphorous conversion to struvite from stored urine can reach up to 98 % at Mg: P molar ratio of 1.2 and at a pH of 9.5. Complete P recoveries were obtainable at higher Mg:P molar ratios of up to 21, in a pH range of 9 to 10. However, nesquehonite also started to form at Mg: P molar ratios greater than 7.8 and between pH values of 6 and 12 due to excess magnesium ions and the presence of carbonate ions in the urine. Theoretically, the implication of this was a compromise on the quality of the struvite produced and, practically, also the likelihood of the formation of a passivation layer of nesquehonite on the sacrificial magnesium anode. As such, the magnesium concentration in the model was kept high enough to recover most of the orthophosphate present in the urine but low enough to avoid the formation of nesquehonite. The optimum Mg:P molar ratio and pH range were tested experimentally using an electrochemical precipitation batch reactor with struvite seeds. A Mg:P molar ratio of 1.2 and pH of 9.5 proved to provide good operating conditions for phosphorous recovery of up to 96 % instead of the 98% that had been predicted by the thermodynamic model. The results of the seeded electrochemical precipitation experiments, at different seeding and stirring conditions, showed that seeding does not affect the rate and extent of phosphate recovery and the rate at which equilibrium is reached. This was due to the fact that the precipitation of struvite occurred extremely rapidly owing to its sparingly soluble nature and the degree of supersaturation of the solution, which also led to the fast precipitation of numerous small struvite crystals. Results of the thermodynamic model and those of the seeded electrochemical precipitation experiment were comparable with regards to the recovery of the orthophosphate from the source separated urine and the conversion of magnesium from the sacrificial anode. However, there were notable discrepancies with regards to the final pH and the conversion of the carbonate ions and thus the likelihood of the formation of nesquehonite. The formation of the passivation layer of nesquehonite on the anode surface over time led to an increase in the magnesium electrode potential which translated to an increase in the cell potential by 130% from 1 V to 2.3 V. It was postulated that this was due to the relatively high level of magnesium supersaturation at the anode surface compared to the rest of the system which favoured the formation of nesquehonite. This was not computed in the thermodynamic model which assumed the system to be well mixed, however the simulations did show that nesquehonite does indeed form at high magnesium concentration. The precipitates that were collected in suspension were analysed for magnesium, orthophosphate and ammonium ions and the results also showed that these ions were present in a 1:1:1 ratio. Further to this, XRD results confirmed that the layer on the anode was indeed nesquehonite. This led to the conclusion that the precipitates that formed in suspension were struvite and most of the nesquehonite that formed only built up on the anode surface and did not affect the quality of the struvite collected. Analysis of the size of the crystals that were produced in the electrochemical precipitation batch reactor over 2 hours showed that there was only a slight increase in the size of the resultant struvite particles after seeding. Two extreme stirring rates were also investigated and it was observed that high stirring rates produced larger particles than the lower stirring rate. This implies that the high stirring rate resulted in a decrease in the local supersaturation at the anode surface. This promotes particle size enlargement through a growth mechanism in the bulk solution rather than the formation of smaller particles through nucleation. Also, higher stirrer speeds could result in higher rate of particle-particle collisions and increased particle shear which would result in the formation of smaller particles. However, this appears not to be the case for this research. When the system was run continuously, i.e. by recycling the struvite particles back into the reactor with fresh urine in order to increase their residence time in the reactor, there was some growth of the crystals after 8 hours. The general shape exhibited by the particles was coffin like in all the images with other spherical and randomly shaped particles. This dissertation has thus shown that the electrochemical precipitation of struvite from source separated urine can be improved in terms of the functionality of the system and the quality of product by i) seeding ii) stirring. The cost of struvite precipitation in the seeded electrochemical precipitation reactor using a magnesium sacrificial anode was evaluated and compared against the cost of struvite precipitation by chemical dosage of magnesium with MgCl2. The electricity and magnesium electrode costs were evaluated for the seeded electrochemical precipitation method while only the chemical costs for the chemical dosage method were considered. Transport costs were neglected as they were assumed to be the same for both. The economic evaluation showed that the electrochemical precipitation method costs about ZAR 3.47 per kilogram of struvite while the chemical dosing method costs about ZAR 3.87 per kilogram of struvite produced. Since these costs are similar, it is recommended that this technology be investigated further. It is also important to note that chemical precipitation is an already proven technique for struvite production while electrochemical precipitation of struvite is relatively new, even though the core technology has existed for some time now. Therefore, there is likely more room for improvement and optimization in an electrochemical precipitation process while this might not be the case for chemical precipitation of struvite. Also, since it is difficult to control the magnesium concentration on the anode surface, it is recommended that in order to suppress the formation of nesquehonite on the anode surface, the urine should be pre-treated in order to remove the carbonate ions which leads to the formation of nesquehonite. Pre-treatment could involve acidifying the solution in order to convert the carbonate slowly into unstable carbonic acid and then to carbon dioxide gas. This would be beneficial especially when the system is run continuously in order to influence crystal growth and essentially avoid system failures which would be caused by the build-up of the nesquehonite layer.

Mineral processing is the process that involves liberation and beneficiation of valuable constituents of an ore. Several physical beneficiation processes exist and one such process is classification. Screens are classification devices sometimes used in the classification stage of closed grinding circuits to separate mill product into different size classes. Poor classification of particles results in reduced throughput, high power consumption and over-grinding. Most of the research on screening has been done in scalping applications or classification at relatively large cut sizes. There is limited work done on screening at feed sizes of minus 150 μm and there are no robust models for wet fine screening application for use in circuit simulation studies. The effect of feed flow rate, solids concentration and aperture size on wet fine screening performance was evaluated in this study. The range of values of the factors investigated were the feed rate (9, 13, 19, 25, 30 and 35 t/h), screen aperture size (45, 75, 106 and 150 μm) and solids content (30, 40, 50 and 60%). A pilot Derrick screen plant at Mintek in Johannesburg was used for the experiments on a UG2 and chromite ore blend. Screen undersize and oversize samples were collected for particle size distribution analysis and mass balance calculations. The samples collected were filtered, de-lumped and split down to masses ranging between 200 and 300 grams for wet screening using the Malvern MasterSizer particle analyser. The results were used to analyse the effect of the investigated factors on the wet fine screening performance. These results were used to develop a wet fine screening model. Results indicate that increased feed flow rate and solids concentration lead to finer cut sizes, reduced sharpness of separation and higher water recoveries to the oversize. An increase in aperture size increased the sharpness of separation and decreased the water recoveries to the oversize. The solids concentration appeared to have a higher effect on cut size than the feed flow rate. The highest cut size and sharpness of separation and lowest water recovery to oversize were attained at the lowest feed rate. The lowest solids concentration produced the best performance with regards to all partition curve properties. The cut size approached the aperture size at the lowest throughput and solids concentration for all aperture sizes. All the efficiency curves exhibited fish hooks at fine particle sizes with the fish hooks becoming more pronounced at higher feed flow rates and solids concentration and smaller aperture sizes. A wet fine screen model that includes multi component ores as well as changes in operating conditions was developed using the 2-parameter Whiten screen model as a basis. The dimensional analysis approach was applied in developing the sub-models that relate the operating and design parameters to the Whiten model parameters. The dimensional analysis approach was further applied to develop the model that describes the fish hook effect for subsequent incorporation into the overall modified fine screen model. Generally, the modified model is capable of predicting the performance of the wet fine screen reasonably well with minor errors and accommodates for the data that exhibits the fish hook. The model also reduces the fitting process required in the original Whiten model.

High throughput and combinatorial experimentation is becoming more and more used in catalysis research. The benefits of parallel experiments are not only limited to shorten the time-to-market, but also give opportunities to study the process in more depth by performing more experiments. The influence of a parameter, for example the amount of the active metal and/or promoter, to the process is better understood with a broader parameter space investigated. To study the parameter space, multiple experiments need to be performed. It is of paramount importance to understand the variability of the data between these experiments. This is not always defined, specifically when literature gives contradictory results, most often due to the time for duplicate experiments necessary. In this project the reproducibility and variance in high throughput catalyst preparation and testing was determined and the use of parallel experimentation was demonstrated within a catalyst development study. The high throughput equipment was used for catalyst development studies for fuel processing, the production of fuel cell-grade hydrogen from hydrocarbon fuels. Fuel processing consists of three catalytic reactions, namely reforming, water-gas shift and a CO clean-up through either selective methanation or preferential oxidation. Focus has been placed on the first two reactions, steam methane reforming (SMR) and medium temperature water-gas shift (WGS), using platinum group metals (PGM). All catalysts in this study (except for the commercial WGS catalyst) were prepared using automated synthesis robot (Chemspeed ISYNTH) and the activity testing was performed on the Avantium Flowrence. For both reactions two types of studies were performed, one-to-many and many-tomany; referring to one catalyst tested in many reactors or many prepared catalysts (same composition, different batches) tested in many reactors. For the WGS one-tomany a commercial low temperature shift catalyst was selected and for SMR a single batch of Rh/Al2O3. The many-to-many experiments comprised of eight batches of prepared catalysts for both reactions. The WGS reaction was performed with 1 wt% Pt/Al2O3 catalysts and for the reforming reaction batches of 0.5 wt% Rh/Al2O3 was used. It was proven that in all these studies the experimental standard deviations in the data is 6%, from preparation to activity measurements. A study on the rhodium metal loading on alumina in the steam methane reforming catalyst was studied between 0.05 and 0.6 wt%. A 0.4 wt% Rh/Al2O3 was found to have the highest activity per amount of rhodium. Lower Rh content would require decreased space velocity, whereas higher metal content does not increase the conversion due to larger crystals sizes. This study has been performed up to a metal loading of 0.6 wt% and it is recommended to follow-up with studying the range of 0.6 to ~2.5 wt% to investigate the optimal metal loading. It was shown that the use of automated experimentation (parallel preparation and evaluation under same condition) for catalyst development results in highly reproducible results with a relative standard deviation of ~6% on the catalytic activity. The high throughput equipment was demonstrated to be a very powerful tool in catalyst research.

Due to continually declining ore grades, increasing mineralogical complexity, and increasing metal demand, models for the design and optimisation of minerals processing operations are of critical importance. These models do not currently incorporate particle shape, which, although rarely quantified, is known to affect numerous unit operations. Automated Scanning Electron Microscopy (Auto-SEM-EDS) is a widely used tool for mineralogical analysis. It also provides an opportunity for simple, quantitative and mineral-specific shape characterisation. Existing mineralogical databases could therefore become useful resources to facilitate the incorporation of shape effects in minerals processing models. A robust Auto-SEM-EDS shape characterisation methodology is required to ensure that the particle shape information in these databases is interpreted appropriately. For this work, a novel methodology for Auto-SEM-EDS shape characterisation was developed that is suitable for the analysis of fine particles (<75 μm). This involved testing the response of various shape descriptors to image resolution, and measurement with different devices and image processing routines. The most widely used shape descriptor in minerals processing, circularity, was found to be highly dependent on both image resolution and image processing settings, making it a poor choice for shape characterisation of fine particles. Roundness and aspect ratio were found to be more robust descriptors. However, in the interest of being able to compare particulate shape measurements across different studies, the precise definition of aspect ratio is important as variation in ‘length’ and ‘width’ definitions can significantly impact aspect ratio measurements. The possibility that preferential orientation of particles would introduce bias to the 2-D cross-sectional measurements was also addressed through comparison of roundness distributions measured from orthogonal cross-sections of a particulate sample mounted within a block of resin. The excellent repeatability of these measurements indicated that the particles were randomly orientated, and thus it can be inferred that 2-D measurements of a sufficient number of particles will be directly related to the particulate sample’s 3-D properties. Roundness and aspect ratio were then used in conjunction to produce surface frequency distributions that allow for distinction between non-rounded particles that were smooth and elongated and non-rounded particles that were neither elongated nor smooth. Three applications of the shape characterisation methodology developed were then demonstrated, which highlighted some of the potential contributions that this methodology can make towards minerals processing. The applications were all based on a case study of the Upper Group 2 (UG2) Chromitite, a platinum group mineral (PGM) ore of key economic significance to South Africa. The first application of the shape characterisation methodology constituted a novel approach to quantify phase boundary fracture, a mode of breakage in which cracks tend to propagate along the boundaries between mineral grains. The mineral specific shape characterisation that is possible with Auto-SEM-EDS was used to assess the conservation of grain shape during breakage, quantifying the extent to which phase boundary fracture led to liberation by detachment of the chromite grains. The critical importance of phase boundary fracture to the liberation of platinum group minerals and thus the economic processing of UG2 ore was demonstrated. Both the data presented, and the novel approach, can provide valuable evidence to support the development of liberation models. The second application of the methodology involved fine grinding of UG2 ore in ball mills and stirred mills, with particle shape analysis used to provide insight into breakage mechanisms. The particle shape characteristics of feed and product samples from both laboratory and plantscale ball mills and stirred mills were compared. None of the milling conditions led to product particles being more rounded than feed particles. This suggested that the common perception that fine grinding involves a higher degree of abrasion than primary grinding is inappropriate if the definition of abrasion used is that which is commonly linked to rounding of particles. Particle shape was clearly dependent on ore texture and phase boundary fracture at coarse sizes, but below the chromite grain size distribution (less than 75 μm) there were minimal differences in shape between the major mineral components. The particle roundness distributions were similar for the laboratory stirred mill sample and the ball mill and IsaMill samples taken from the plant, suggesting that ore characteristics are the dominant factor controlling particle shape. The similarity in shape characteristics of samples from the plantscale ball mill and IsaMill led to the deduction that differences in the performance of a UG2 ore flotation circuit with an IsaMill on- or off-line are unlikely to be attributable to particle shape effects. Finally, entrainment is an important issue in most flotation circuits as it reduces concentrate grade. It is particularly important for UG2 ore due to the negative impact of chromite gangue on smelter performance. Little appears to exist in the published literature correlating particle shape with entrainment, and therefore, the third application for the shape characterisation methodology was focused on this topic. This required an adaptation of the methodology in order to ensure that particle size was not confounding the results. The findings based on samples taken from a UG2 ore concentrator indicated that the chromite particles recovered to the concentrate streams were significantly more rounded than chromite particles of the same size in the feed and tails streams. This suggests that shape could have a significant effect on entrainment, although in this study it was not possible to identify the mechanisms responsible, and more research would be required to determine whether the observations are generalizable. The approach to decouple the effects of shape from size could be applied to other areas of minerals processing research, and could prove particularly useful for the incorporation of effects of shape in classification models for devices such as screens, hydrocyclones, thickeners, reflux classifiers and air classifiers.

Kansanshi mine is the largest copper producer in Africa. The deposit is mineralogically and texturally complex due to supergene enrichment resulting in the presence of a variety of primary and secondary copper minerals. This necessitates the processing of ore through three separate circuits: sulphide flotation, mixed flotation and oxide leach, followed by solvent extraction and electro-winning. This study revisits the process mineralogy of the ore using modern mineralogy tools, which for such a large and complex deposit cannot but deliver significant value. Specific focus is given to copper mineralisation and the flotation of the sulphide ores in compliment to another MSc study from the Centre for Minerals Research focusing on mixed ore flotation (Kalichini, 2015). A series of hand samples and grab samples representing the variation in mineralogy and texture of the Kansanshi ore, as well as two run of mine sulphide ore flotation feed samples were used for this investigation. Process mineralogical characterisation entailed optical microscopy, XRF, QXRD, QEMSCAN and EPMA investigations, alongside a series of laboratory scale batch flotation tests of two sulphide ores at two grinds (80% passing 150 μm, 80% passing 212 μm). Copper mineralisation at Kansanshi occurs as both vein-hosted mineralisation, and to a lesser extent sediment-hosted mineralisation. Later breccia-hosted and supergene mineralisation have overprinted all previous mineralisation styles. Chalcopyrite is the main ore mineral for both vein-hosted and sediment-hosted mineralisation styles. Vein-hosted mineralisation is characterized by an overall coarse-grained texture (>0.5 mm), compared to sediment-hosted mineralisation that is characterised by fine-grained disseminated textures that occur parallel to the bedding and foliation planes. Breccia-hosted and supergene related mineralisation have led to the formation of an array of secondary copper minerals, such as chalcocite, covellite, malachite and chrysocolla. These minerals show a variety of complex intergrowth textures between one another. Secondary copper oxide mineralisation is commonly associated with distinctive stockwork and boxwork textures, with replacement being partial or complete depending on the extent of oxidation. The variety of textures related to the replacement reactions result in grain size variations that cause a decrease in the chalcopyrite grain size and produce secondary copper sulphides that are of equivalent to or of a finer grain size (< 0.2 mm) than that of the primary copper sulphide. Mineralogical investigations of two run of mine sulphide flotation feed samples showed that the dominant ore mineral is chalcopyrite with an overall coarse-grained (> 0.5 mm) texture with minimal fine composite particles, which results in good chalcopyrite liberation. Results of this laboratory study show good copper recoveries (~89%) during rougher flotation, because chalcopyrite liberation was over 90% at a grind of 80% passing 150 μm. The effect of coarsening the grind caused an insignificant loss of copper recovery. This good performance during flotation can be attributed to a number of mineralogical characteristics, including minimal fine composite particles, the natural hydrophobicity of chalcopyrite and the high degree of liberation of chalcopyrite associated with the overall coarse texture of the sulphide ore. Mineralogical investigations suggest that the relatively low copper grades from batch flotation cannot be attributed to the presence of composite particles, and can potentially be improved using a series of cleaner floats. The effects of supergene enrichment on mineralogy and texture, and its influence on processing, have been used to develop a simplified process mineralogy matrix for Kansanshi. The matrix demonstrates the continuum of mineralogy and textures due to supergene enrichment and their potential influences on mineral processing. Some ideas for regular on-site use of mineralogical analyses at Kansanshi have also been proposed. Ultimately, this information can be incorporated into the existing geometallurgical framework at Kansanshi, adding to the understanding and predictability of the ore being fed into each circuit.

There is a growing importance in the mineral processing industry to find ways which are economic and effective in improving the recovery of minerals in the flotation process. The focus of this study was on the recovery by flotation of minerals found in the Merensky reef, which is one of the major reefs in the Bushveld complex. In that reef, base metal sulphide (BMS) minerals are commonly associated with PGMs and this has an effect on the way in which these minerals are concentrated by flotation (Vermaak et al. 2004; Wiese et al. 2005b; Miller et al. 2005; Schouwstra et al. 2000). A major problem in this process has been reported to be losses of valuable minerals (PGMs) associated with the loss of BMS (Wiese et al. 2005b), during flotation. The present investigation has focused on studying the relationship between the flotation of sulphide minerals using xanthates as collectors and the electrochemical properties of the flotation system. It is well known that electrochemical mechanisms in flotation systems have a major influence on flotation since the reactions occurring at the mineral/solution interface are of critical importance in the process (Woods, 1971). The aim of this study was to investigate the extent to which there was a relationship between the electrochemical reactions occurring in this ore which could indicate the effectiveness of the flotation process. The electrochemical reactions were studied by determining the redox potential changes occurring when various changes were made. These were the length of the alkyl chain length of the xanthate collector, changing the pH or using various chemical reagents to change the potential of the system. It was found from the rest potential measurements, that collectors of different chain length have different extents of interaction with mineral surface. A greater interaction, which is indicated by a greater change in the mixed potential after addition of the collector, is considered to be indicative of a greater adsorption of the collector at the mineral surface. It was hypothesized that this stronger adsorption by collectors of longer alkyl chain length would result in improved flotation performance. However, this was not observed to be the case and that was consistent with previous results on the relationship between the recovery of sulphide minerals in the Merensky ore and xanthates of different chain lengths. Thus it was shown that there was no correlation between the interactions between collectors of different alkyl chain lengths as determined through electrochemical studies and the flotation performance of valuable minerals under the tests conditions used. When NaClO was used as a potential modifier it was found that it was possible to change the Eh values without any change in pH. This was important since it allowed the effect of Eh alone to be investigated. The measurements of rest potentials of sulphide minerals showed that the addition of NaCIO increased these potentials to varying degrees. It was hypothesised that an increase in surface potential would promote collector-mineral interactions and thus possibly the formation of hydrophobic species such as dixanthogen. Depending on whether the potential is greater than a threshold value of between 120 – 150 mV, the formation of dixanthogen would be preferred and that would result in a higher degree of hydrophobicity, and hence a possible improvement in floatability of valuable minerals. However, from the findings in this study copper recoveries and grades remained largely unchanged at Eh values produced by the addition of NaClO which were in the range from (100 – 200 mV) to (500 – 600 mV) at a pH of 9. At pH 11 the Eh produced by the addition of NaClO was in the range between (0 – 100 mV) and (200 – 300 mV) and similar results for copper were observed in the presence of a collector and/or NaCIO. Under all the conditions nickel recoveries and grades were high only in the presence of a collector. This project has contributed to a further understanding of the effect of changing electrochemical potential by chemical means on the flotation of sulphides. Usually such potentials are changed by changing either the dissolved oxygen content or the pH (as illustrated in Pourbaix diagrams). The actual mechanisms involved when using chemical reagents are still generally not well understood (Chanturiya and Vigdergauz, 2009), and consequently some of the proposed mechanisms are inevitably speculative.

The flotation process has been used for more than a century to separate valuable minerals from bulk ores. The separation process is based on utilising the differences in the physico-chemical properties of liberated particles, mainly the particle hydrophobicity which allows the particles to be attached to air bubbles rising from the pulp phase into the froth phase and subsequently collected to the launder. The stability of the froth phase which is be defined as the ability of bubbles to resist coalescing and bursting (Triffet & Cilliers, 2004), has been shown to have a significant effect on the efficiency of the flotation process. An unstable froth will result in poor valuable mineral recovery as these desired hydrophobic particles are detached from air bubbles and drain with the water back into the pulp phase due to bubble coalescence. On the other hand, a very stable froth may result in poor concentrate grade as the unwanted gangue materials are unselectively entrained to the concentrate. As a result, a substantial amount of research has been performed on improving control of froth stability by the manipulation of frother type and dosage. A recent study investigated the manipulation of flotation operating parameters such as air rate, froth height and depressant dosage which resulted in minimal changes in froth stability. The present study then investigated the effect of particle size and solids concentration on the stability of the froth phase using a UG2 ore and an Itabirite ore. Froth stability was determined using Bikerman tests on a laboratory scale noncontinuous stability column. A novel continuously operated agitated hybrid cell was also used to assess froth stability, with water recovery and froth recovery used as proxies for froth stability. The agitated hybrid cell was then included in the experimental design as it allowed for continuous floatation system to be evaluated which resembles more industrial operations as compared to the stability column. The hybrid also incorporated the agitation zone benefits of a lab scale batch flotation cell which allows for better attachment of coarse particles and also allowing for the formation of deeper froths enabling improved froth stability measurements. The viability of using the top froth average bubble size and the side of froth axial bubble coalescence rate as froth stability proxies was also evaluated as the columns were clear glass. An evaluation of the particle size distributions of the feed and the concentrate reporting to the launder showed that the concentrate was consistently finer than the feed. Feed particle sizes for the UG2 ore ranged from 157-78 μm with concentrate sizes ranging from 83-39 μm from the coarsest to the finest grind. Feed particle sizes for the iron ore ranged from 29-62% passing 38 μm with concentrate sizes ranging from 49-82% passing 38 μm from the coarsest to the finest grind. It is hypothesised that this was due to the increased weight of the coarse particles resulting in the particles draining back into the pulp zone at a faster rate. As a result, a smaller fraction of the coarser particles reports to the concentrate resulting in the finer particle size distribution. The effect was shown to be more pronounced for the UG2 ore as compared to the iron ore, as the UG2 ore forms a less stable froth which has a higher rate of bubble coalescence. Changing the feed particle size was also shown to alter the concentrate particle size thereby allowing for the investigation into the effect that the size of particles present in the actual froth has on froth stability. Test results show that froth stability increased with decreasing particle size for both ores. It was hypothesised that a decrease in particle size would result in an increase in the maximum capillary pressure thereby reducing capillary drainage. It was also hypothesised that a decrease in the particle size of the entrained particles would increase the viscosity of the interfilm fluid, thereby reducing drainage rate and increasing stability. Froth stability was shown to follow a decreasing power law relationship with feed particle size. Froth stability was also shown to decrease sharply with increasing particle size over the fine feed size range of less than 100 μm, with the effect becoming less pronounced with increasing particle size over the coarser range. The steep decrease was shown to correspond to concentrate particle sizes approximately less than 50 μm, the range in which particles are expected to report to the froth through entrainment. Froth stability followed an increasing linear relationship with feed specific area for the size range tested (UG2 ore: 78-157 μm and iron ore: 48-118 μm). More importantly, froth stability was assessed as a function of total surface area imparted by the concentrate particles and stability was shown to increase with increased total surface area. Decreasing the feed particle size was shown to result in higher solids recovery and finer concentrate particle size thus higher specific area, therefore total surface area imparted was critical to the assessment of froth stability. Increasing solids concentration was shown to result in an increase in froth stability and the effect was shown to be less pronounced for UG2 ore as it is a sparsely mineralised PGM-bearing ore. The rheology of the interfilm is suggested to play a more significant role in particle stabilisation of the froth as the stability increased steeply in the particle size range where entrainment is expected as shown by the entrainment factor increasing steeply in the same size range. Future work evaluating the rheology of the froth is recommended to further clarify this.