High grade ores have largely been depleted and those currently being treated are low grade, complex and sometimes finely disseminated, requiring fine grinding to liberate valuable minerals. For fine grinding applications, conventional tumbling mills are energy intensive. More energy efficient technologies such as stirred mills have been developed and widely used for fine and ultra-fine grinding. In this study, the effects of residence time, solids concentration, impeller speed, impeller type, media size and media density on energy consumption in a batch vertical stirred mill were investigated. The effect of energy on mill performance was assessed using the perfect mixing mill model. In addition, the effect of media stress intensity on grind and energy efficiency at constant residence time was also investigated. It was found that irrespective of the method of altering the energy input, the fineness of grind improved with increase in the specific energy input. This suggests that energy is the key driver for size reduction. The perfect mixing model can be used to assess mill performance and the breakage rates generally increased with increase in the specific energy input, impeller speed and solids concentration. The media stress intensity approach is useful in assessing mill performance in stirred mills at constant residence time. The fineness of grind improved when the media stress intensity was varied from 4.41×10-3 to 27.41×10-3Nm. In addition, the specific energy required to produce material below 25μm and 38μm decreased with an increase in the media stress intensity. When slurry density effects were considered, an optimum stress intensity was observed with respect to specific energy required to produce material below 25μm and 38μm. It was recommended that additional test work be carried out to investigate the effect of media size in the range -6.7mm + 2mm on energy efficiency. It was also recommended that tests be carried out at impeller speed between 600rpm and 1500rpm to assess how mill performance increases even at relatively high impeller speeds. In addition, a model predicting the specific energy using the impeller speed and solids concentration can also be developed.

In the western limb of the Bushveld Igneous Complex, platinum is mined predominantly from two reefs – Merensky and UG2. Most of the platinum group minerals (PGM’s) in Merensky ore are associated with base metal sulphides (BMS), and thus Merensky concentrators will usually resemble simple BMS circuits. However, the mineralogy of UG2 ore is more complicated, and thus UG2 circuits are also more complex. The UG2 reef is a chromitite layer in the critical zone of the Bushveld Igneous Complex, which results in high chromite content. Chromite causes significant complications in the downstream smelter process, and therefore chromite constraints are imposed on UG2 concentrators. A further aspect complicating the treatment of UG2 ore is that PGM’s are not only associated with BMS, but ultra-fine PGM’s are also locked in gangue minerals. This affects the milling and flotation characteristics of the circuits, as it is not possible to efficiently target the liberation and recovery of relatively large BMS and ultra-fine PGM’s in the same circuit. As a result UG2 circuits have evolved to deal with these issues in a number of ways. This thesis focuses on the design of milling and flotation circuits to optimise the recovery of coarse BMS (with associated PGM’s) and ultra-fine PGM’s contained in associated siliceous gangue minerals. In order to achieve this, UG2 circuits usually feature more than one milling and flotation stage. In the primary stage, a relatively coarse grind targets the liberation and recovery of BMS, while much finer grinds in subsequent milling stages target the liberation and recovery of PGM’s locked in siliceous minerals. In recent years stirred mills have been incorporated into UG2 circuits for the liberation of fine PGM’s locked in gangue. These mills grind exclusively via attrition type breakage, as opposed to the predominant impact breakage in a ball mill. The development of these multi-stage circuits to treat UG2 ore was based to a large extent on a qualitative evaluation of the ore and mineralogy, as well as by trial and error over many years of operation, rather than being built on a scientific foundation. Therefore, the aim of this project was to gain an understanding of the effect of breakage mechanism and circuit configuration on the floatability profile of a UG2 ore. This information would then be used to establish a formal design framework for the selection of mill type, and the number of milling and flotation stages to meet specified performance criteria. In order to achieve these goals, it was necessary to design a set of experiments that would investigate multiple millfloat circuits and various milling devices. The experiments were conducted on a pilot scale with a ball mill, IsaMill and Stirred Media Detritor (SMD), together with Lonmin’s flotation pilot plant. The ball mill represented a milling device in which impact breakage would predominate, using steel balls as a grinding media, whilst the IsaMill and SMD represented devices that grind exclusively via attrition, using inert grinding media. The mills and flotation cells were arranged into different millfloat configurations to test the effect of multiple milling and flotation stages on the PGM recovery of UG2 ore. The data was fitted to a kinetic model adapted from the literature to determine the floatability of the ore as a function of its treatment in the different circuits tested. Some new modelling methodologies were developed to incorporate disparate data sets into the same modelling framework. An analysis of the performance of the different circuits showed that PGM recovery increased with an increase in the number of mill-float stages, while a single mill-float stage produced higher initial grades. On a size-by-size basis additional mill-float stages resulted in an increase in PGM recovery in all size classes, while no minimum size was detected for optimum flotation (the highest recoveries for all three circuits were achieved in the finest size fraction). A detailed analysis of the data and floatability profiles revealed the following: (i) Multi-stage circuits minimise over-grinding, and this manifests as an improved recovery in the finest size fraction that was measured (-10m). (ii) Sub 10m liberated PGM’s displayed higher floatabilities than locked or partially liberated PGM’s in coarser size fractions, resulting in the higher recoveries observed in this size fraction for all circuits. (iii) Preferential liberation of PGM’s was observed for this ore. This meant that stagewise removal also favoured recovery in coarser sizes, since partially liberated PGM’s were recovered before being liberated to a finer size fraction by additional milling. (iv) Attrition breakage is more efficient than impact breakage at liberating valuables in the finest size fraction, but also more prone to over-grinding of liberated PGMbearing BMS particles. Therefore circuits with attritioning devices are more likely to benefit from multiple mill-float stages. Impact breakage is more efficient at liberating particles from coarse size fractions – the host will usually be shattered and the valuable mineral liberated to a smaller size fraction. (vi) In contrast, attrition breakage gradually chips away at the host particle, often achieving partial liberation of the valuable particle in a relatively coarse size fraction. Therefore, attrition breakage favours recovery in coarser sizes, as valuables are liberated in a coarser size fraction than with impact breakage. This does not indicate that attrition breakage is more efficient at liberating valuables from coarse sizes, as impact breakage will usually liberate more particles from coarse sizes, but these liberated particles are recovered in finer size fractions. (vii) For UG2 ore, it was found that significantly higher PGE concentrate grades could be achieved with stirred mills. Results indicate that this was driven by the liberation of PGM’s from BMS, since the recovery of fine, liberated PGM’s would result in higher PGE grades than the recovery of PGM’s associated with BMS. From these observations and findings it was possible to construct a framework for the milling and rougher design of UG2 circuits. The information required for the design is a target size distribution, size and associations of the valuable minerals and the hardness and breakage characteristics of the ore. From this information the number and type of milling devices, as well as the number of flotation stages can be selected. The target grind for each stage is driven to a large extent by economic return-on-investment considerations, although the recovery-grind relationship of the ore can assist in making that decision. Although the work and design framework was established for UG2 ore, it can be used to design the main stream milling and rougher circuits for most ore types.

This study was performed in context of a series of studies conducted investigating the use of acid zeolites for the synthesis of thymol by selective isopropylation of m‐cresol on the 6‐position. The aim of this study was to investigate the effect of the SiO2/Al2O3 ratio, i.e. the aluminium content of H‐MFI (H‐ZSM‐5) zeolites. The results are interpreted based on a deep analysis of the reaction network. Three commercial samples of the H‐MFI zeolite were employed with high, medium and low molar SiO2/Al2O3 ratios of 400, 90 and 20 respectively, whereby the first two samples consisted of agglomerates of very small crystallites, < 10 nm, while the last sample consisted of comparatively large crystallites with diameters of several 10 nm. m‐Cresol as the substrate was alkylated with isopropanol in a molar feed ratio of 1:1. Reactions were carried out in a tubular fixed‐bed reactor in the gas‐phase at reaction conditions of 200 ‐ 300°C, WHSV of 0.016 – 1.03 gm‐cresol/gcat.hr and 3 bar (abs) as the standard pressure. A pressure series at 275°C and WHSV of 0.25 – 1.03 gm‐cresol/gcat.hr was also carried out. The following was observed: Isopropanol reacts in two parallel ways namely, by O‐alkylation of the m‐cresol to form isopropyl‐3‐tolyl ether which occurs predominantly by its dehydration to propene and secondly, by alkylation of m‐cresol with propene with high preference on the 6‐position, forming thymol with high selectivity. The ether rearranges internally or transalkylates with m‐cresol to form thymol, 2‐isopropyl‐3‐methyl phenol and 4‐isopropyl‐3‐methyl phenol with a thymol selectivity < 50%. Higher reaction temperatures were found to favour the propene route. It appears that the major effect of the SiO2/Al2O3 ratio is, as expected, a direct effect on catalyst activity and an indirect effect on thymol selectivity via mass transfer control. Thymol is an intermediate product in the reaction network, that is, it can react further (by isomerisation or further alkylation to higher alkylated products). Mass transfer control on such reaction sequences reduce intermediate selectivity and hence thymol selectivity. Exactly this was observed over the low SiO2/Al2O3 ratio (i.e. high aluminium content), ‘large’ crystallite H‐MFI‐20 sample. The major conclusions and recommendations resulting from this study are the following:  Use propene as the alkylating agent rather than isopropanol in order to achieve high thymol selectivity  Avoid mass transfer control.

Proton exchange membrane fuel cells (PEMFC) are identified as future energy conversion devices, for application in portable and transportation devices. The preferred catalyst for the PEMFC is a Pt-catalyst. However, due to the slow oxygen reduction reaction (ORR) kinetics, high Pt loadings have to be used. The high Pt loadings lead to high costs of the PEMFC. Pt-Co alloys have been identified as catalysts having higher ORR activity higher than of a Pt-catalyst. Therefore, in the present study, the Density Functional Theory (DFT) technique is used to gain fundamental insight into the ORR on the Pt3Co(111) surface. The calculations have been performed using the plane wave based code, the Vienna ab-initio Simulation Package (VASP). DFT spin-polarized calculations, utilizing the GGA-PW91 functional, have been used to study the adsorption of the ORR intermediates, viz. O2, O, OOH, OH, H2O and HOOH on the Pt3Co(111) surface. The results obtained on the Pt3Co(111) surface are compared to the results obtained on the Pt(111) surface. The adsorption strength of the ORR intermediates has been shown to be affected by the presence of Co to varying extents on the Pt3Co(111) surface relative to adsorption on the Pt(111) surface. The most strongly stabilised ORR intermediate on the Pt3Co(111) surface relative to adsorption on the Pt(111) surface is O: on the Pt3Co(111) surface O is 0.45 eV more strongly adsorbed than on the Pt(111) surface. The least affected ORR intermediate is H2O: H2O adsorption on the Pt3Co(111) surface is 0.20 eV more stable than on the Pt(111) surface. The energetically favorable, i.e. most strongly bound adsorption configurations for all the ORR intermediates involves a configuration in which the ORR intermediate is bonded to a surface Co atom. Therefore, the surface Co atom stabilizes the adsorption of the ORR intermediates, relative to adsorption on the Pt(111) surface. Coadsorbed configurations have been used to study the formation and dissociation of the ORR intermediates. From the coadsorption studies, it is shown that there is an energy cost associated with moving the adsorbates from their lowest energy sites, while separately adsorbed, to the higher energy coadsorbed state, prior to reaction. Hence, adsorbate-adsorbate interactions are expected to destabilize the coadsorbed state at the coverages considered in the present study. The Climbing Image Nudged Elastic Band (CI-NEB) method has been used to locate the transition states and to calculate the activation energies of the different elementary reaction steps. The calculated dissociation reaction activation energies for the Pt3Co(111) surface are found to be lower than the dissociation activation energies calculated on the Pt(111) surface. The most lowered dissociation activation energy is for the dissociation of O2: on the Pt3Co(111) surface the activation energy is 0.08 eV, whilst on the Pt(111) surface the activation energy is 0.59 eV. For the hydrogenation reaction steps, only the hydrogenation of O to form OH occurs with a lower activation energy of 0.86 eV on the Pt3Co(111) surface, compared to 0.95 eV on the Pt(111) surface. For other hydrogenation reaction steps, the activation energies on the Pt3Co(111) surface are higher than those on the Pt(111) surface. Based on the calculated activation energies of the elementary ORR reaction steps, the dissociative and the O-assisted H2O dissociation mechanisms are identified as the mechanisms most likely to be dominant on the Pt3Co(111) surface, due to having lower activation energies relative to the associative mechanisms. For both mechanisms, the reaction step with the highest activation energy is the step involving O, i.e. O hydrogenation to form OH for the dissociative mechanism, and the O* + H2O* ! 2OH* reaction for the O-assisted H2O dissociation mechanism. Thus, the reaction step involving the reaction of the strongly adsorbed O species, is identified as the potential rate limiting step of the ORR. Both the dissociative and the O-assisted H2O dissociation mechanisms are expected to be in competition on the Pt3Co(111) surface, since the potential rate limiting step for both mechanisms have similar activation energies. Hence, the preferred mechanism will depend on the relative abundances of the H species and H2O on the Pt3Co(111) surface. A microkinetic analysis would be need needed to fully account for concentration and entropic contributions to the rate of reaction for the different ORR elementary reaction steps.

Ammonia oxidation is used in the production of nitric acid. The process is either run at high pressure or low pressure, with the latter requiring larger equipment. Platinum gauze is typically used as a catalyst operating at high temperatures (in the range of 810-940 OC). The platinum based catalyst is highly active and highly selective in producing the desired NOx products, with some formation of the undesired byproducts, i.e. N2 , N20 and N20 4. However, a significant amount of platinum is lost during the process due to platinum volatilisation resulting in plant operating times varying between 2-12 months. Furthermore, the loss of platinum is the 2nd largest expense of the operation. Platinum loss can only be minimised but not eliminated , thus a variety of metal oxide catalysts for oxidation of ammonia to nitrogen oxides have been studied. Cobalt oxide seems to be the most promising alternative for platinum exhibiting a high activity and selectivity towards NO. The aim of this study is to explore the use of a supported cobalt Co30 4 on silica catalyst for ammonia oxidation and compare some of the results with a commercial catalyst consisting of a pure, unsupported Co304. Both the synthesised and commercial catalyst showed a maximum conversion of ammonia at approximately 600 OC. A supported catalyst with a low cobalt loading and smaller crystallites yielded similar conversions of ammonia compared to the pure cobalt catalyst with much larger crystallites. However, the calculated intrinsic activity constant per m2 of Co30 4 revealed that the commercial catalyst was more active compared to the in-house prepared Co30JSi02 catalyst. Indicating that severe deactivation might have taken place on the synthesised Co30,JSi02 catalyst under ammonia oxidation conditions using an ammonia content of 7.1 vol.-% at 450- 800 OC. A high ammonia conversion can be achieved by adjusting the space time. The NO content as a fraction of NO plus N20, increases with increasing temperature before the catalyst is completely deactivated at temperatures above 800 OC. The inhouse prepared Co30,JSi02 catalyst displayed a higher relative NO selectivity compared to the commercial Co304 catalyst under industrially relevant conditions at complete conversion of ammonia. Applying the rate equation proposed by Saykov et al. (2000) and operating under a regime where Knudsen diffusion is the dominant diffusion mechanism, the in-house synthesised supported catalyst and the commercial catalyst showed severe mass transport limitations, indicating inefficient use of the catalyst. The heat transfer limitations were assumed to be negligible with minimal temperature gradients with the catalyst and boundary layer. The supported Co3041'Si02 catalyst showed severe mass transfer limitations with the effectiveness factor less than 0.5 for particles greater than 300 microns. The mass transfer parameters (Thiele modulus, effectiveness factor, rintrinsic and r observed) exhibit small changes over the catalyst bed at low conversions of NH3 and displayed major significant changes at more industrially relevant conditions where higher conversion of NH3 is achieved., The conversion of ammonia decreases rapidly at higher temperatures. It is deduced that sintering of the catalyst is not a major concern. Ammonia oxidation proceeds via the Mars and Van Krevelen mechanism. Therefore the deactivation of the catalyst might be caused by the reduction of active Co30 4 phase to the catalytically inactive CoO phase. Since the mechanism involves the lattice oxygen, the deactivation mechanism is thought to be reversible by utilising excess air. However for the supported catalyst, the CoO acts as an intermediate in the formation of cobalt silicate (Co2Si04) , resulting in an irreversible deactivation. In conclusion , a support material which can react with CoO under hydrothermal conditions, i.e. silica and alumina should be avoided in the preparation of cobalt catalysts for ammonia oxidation.

Sustainability in the mineral processing industry is guaranteed by the efficient operation of unit processes to maximize profitability within the prevailing economic conditions at the time. Efficient operation can be measured based on cost, metallurgical and energy efficiency of the operation or as a combination of the three, with energy efficiency becoming increasingly more important in the world and current South African context even more so. With ore bodies becoming more complex and finely disseminated the need for ultra fine grinding to maximise mineral recoveries in PGM concentrating operations by means of mainstream grinding has become more important. Conventional grinding technology was found to be inefficient to address this challenge. Introduction of new stirred milling technology for mainstream grinding using ceramic media successfully addressed the challenge posed by finely disseminated and complex PGM ore bodies. Introduction of new technology also brings about optimisation opportunities on various levels to ensure that the new technology is performing optimally in order to maximise profitability. As relatively new technology, several potential optimisation areas for the horizontal stirred mill has been identified for study. Operating costs for IsaMillsTM horizontal mills can be classified into 3 groups i.e. media, maintenance and energy cost29. Ceramic media type used can potentially play a role in determining the cost contribution of each of these groups to the overall mill operating costs. An area found very important to study, taking into consideration the enormous strain on the South African electricity supply system, was the impact of ceramic media properties on the energy efficiency of the horizontal stirred mill. Historical ceramic media test data5, 14, 17 indicated general trends between ceramic media properties and energy consumption. However the datasets available contained little data on the physical properties of the media to conclusively define the relative role of the media properties on energy usage. Five (5) different ceramic media samples from various sources were selected for further subsequent tests in an attempt to decouple the effect of the various media properties on mill power draw. Various physical properties of the media evaluated were measured during the study. Physical properties that were measured were: friction coefficient, size, density, and aspect ratio of every media. M4 IsaMillTM tests were also conducted in water and slurry environments to determine the relative contribution of the media properties on energy consumption. Analysis of historical data and results from the water-media study indicates that media mass and density contributes minimally if at all to the energy consumption of the M4 IsaMillTM. Energy consumption in the M4 IsaMillTM, operating in a water ceramic media environment, was found to be driven by a combination of media load and machine contributing factors. Machine contributing factors can be defined as the No Load Power (NLP) of the mill. Media load factor was found to be proportional to the product of the friction coefficient of the media, the total surface area of the grinding media charge and the aspect ratio of the media. From the results, a methodology was developed to predict the power draw of the M4 IsaMillTM in a water-ceramic media system that incorporates the media properties contributing to energy consumption. The model was successfully applied to a separate set of tests, with the average error in the predicted M4 IsaMillTM power draw being less than 2% of the measured mill power. The model developed in a water-media system was modified to incorporate the friction coefficient measured in slurry and used to predict the power draw of the M4 IsaMillTM during slurry tests. The prediction of power using the modified model was inaccurate because ore particle properties and the energy requirements for particle breakage were not incorporated in the modified water model. Stress Intensity and stress number is proportional to specific energy required for breakage in horizontal stirred mills33. As the ore and ceramic media properties will influence the stress intensity and the stress number the slurry model was expanded to incorporate a breakage energy component After fitting breakage (kb) and media charge (kc) constants to the model fairly accurate prediction of the power draw could be accomplished. Attempts were made to validate the model at lower volumetric media filling degrees with varying success. At very low media loads (50% less than model development data), the prediction was very inaccurate and based on the work from other authors43, it was clear that the model requires further development. Results of the study and the methodologies developed serves as a good foundation for future research into defining and quantifying the impact of ceramic media properties on all areas of horizontal stirred mill operation. The methodology also indicates promise in decoupling the factors contributing to the energy consumption of horizontal stirred mills. Further work is required to define the impact of particle size on media friction coefficient, the overall impact of ore to media ratio, and the effect of slurry rheology on energy consumption of horizontal stirred mills. It also serves as a good foundation for future scale up studies to test the wider application of the methodology on pilot and industrial scale mills. If scale up is reasonable the models could be potentially incorporated into a process control system to provide online monitoring of mill load conditions in the mill.

Energy is known to play an important role in particle-bubble contacting in flotation. This thesis investigates the effect of energy input (or agitation) on the flotation kinetics of quartz in a novel oscillating grid flotation cell. The effects of bubble size and particle size have been recognized as important variables affecting particle-bubble contacting in turbulent systems and are investigated in this thesis. The research work done in this thesis is a continuation of the work done by the Centre for Minerals Research by Deglon (1998) who investigated the effects of energy in a batch mechanical flotation cell. However, this system has a very complex hydrodynamic environment, resulting from the large disparities in turbulence intensity. Previously Breytenbach (1995) had constructed a hybrid flotation column cell, which was essentially a column flotation cell that could be modified into a Jameson cell or a mechanically agitated column cell. He used this to compare particle collection efficiency in these different particle-bubble contacting environments. The third phase of the work was the oscillating baffle column (OBC), a novel flotation column that attains agitation by oscillating a set of orifice baffles through the slurry, thereby producing a more uniform shear rate distribution than would be obtained in an impeller driven system (Anderson, 2008). The OBC unfortunately has significant oscillatory flow and has high shear rates, which often result in detachment effects becoming appreciable. Oscillating grids generate near ideal hydrodynamic environments, characterised by turbulence that is relatively homogeneous and isotropic. The oscillating grid flotation cell used in this study was based on the oscillatory multi-grid mixer used by Bache and Rasool (2001). The oscillatory multi-grid mixer was purchased from these authors and retrofitted to produce the oscillating grid flotation cell. The novel oscillating grid cell consists of a 10 litre tank agitated by 19 grids with a mesh size of 8 mm and grid spacing of 18 mm. The grids were oscillated at a fixed amplitude, equal to the grid spacing, and over a range of frequencies, using a variable speed drive. Frother was added at 100 ppm to be consistent with the work of Deglon (2002) and Ahmed and Jameson (1985). A low gas flow-rate (100 ml/min) and solids concentration were specifically chosen in order that there was minimal influence on the structure of turbulence in the oscillating grid cell, as Bache and Rasool (2001) took measurements in water. Flotation tests were performed on methylated quartz particles (P80 = 100 μm) over a range of power intensities (0.015–0.60 W/kg) and using three different bubble sizes, generated by sintered glass discs (0.13, 0.24 and 0.82 mm). The flotation rate constant was found to increase approximately linearly with increasing particle size for all three bubble sizes. This was due to the increased probability of collision for larger particles and is well established in the flotation literature. A number of researchers have found that the flotation rate constant for quartz particles increases almost linearly with particle size, at low power intensities. An inverse power relationship was observed between bubble size and flotation rate constant for all fine, middling and coarse particle size ranges. This inverse power relationship was due to the increased probability of collision for smaller bubbles and is also well established in the flotation literature. More significantly, the flotation rate constant was found to increase almost linearly with increasing power intensity for all particle and bubble sizes used in this study. The majority of theoretical and experimental studies have found energy input to have less of an effect than the proportional/linear dependence observed in this study. In addition, the increase in the flotation rate constant with increasing power intensity was observed to depend on particle size, but to be less dependent on bubble size. These findings suggest that energy input and bubble size may respectively play more and less of a role in promoting particle–bubble contacting in turbulent environments than was noted in the flotation literature. However, a recent study by Newell and Grano (2006) done using a stirred tank also noted this linear dependence. Given the findings of this thesis, it is strongly recommended that further work be done to investigate the OGC at higher energy intensities (~3W/kg) and to scale it up so that it can be more comparable to the widely used mechanical flotation cells. The homogeneous and nearly isotropic turbulence generated by the OGC also makes it an ideal environment to characterize floatability for different ores.

Comminution involves crushing and grinding operations. The grinding operations use the traditional tumbling mills and stirred mills to reduce the ore to the required fineness. This thesis intends to investigate the influence of design and operating variables on the IsaMillTM specific energy and product size, when grinding UG2 platinum-bearing ore. The main objectives of this work were to study the effects of operating variables on specific energy consumption and product fineness, and to investigate IsaMillTM scale-up protocol. The experimental studies were conducted using the M4 IsaMillTM on a laboratory scale and the M10 000 IsaMillTM on an industrial scale. The laboratory scale M4 IsaMillTM was used to investigate the effects of some design and operating variables on the energy consumption and product fineness for UG2 PGM (Platinum Group Metals) bearing ore. There are many operating and design variables that have been shown to have a significant influence on the stirred mill operations (Clark et al., 2004; Jankovic, 2003; Pease et al., 2005; Zheng et al., 1996). This study, however, has focused only on variables that influence specific energy consumption and the product fineness of grind. The variables investigated include; stirrer speed, media load, media size, feed size, solids concentration and flow rate. Sampling campaigns were conducted on the M10 000 industrial scale IsaMillTM to evaluate the performance of large scale units. The campaigns were conducted at Anglo Platinum’s Waterval UG2 Concentrator and Western Limb Tailings Re-treatment Plant (WLTRP), which are both located in the Rustenburg area in South Africa. Waterval Concentrator treats UG2 platinum ore and has two IsaMillTM operating in mainstream grinding. The WLTRP, on the other hand, treats reclaimed material from the old Klipfontein tailings dam. The reclaimed material contains a mixture of UG2 and Merensky ore. The WLTRP has one IsaMillTM installed in fine-grinding applications to re-grind concentrates. While the results obtained from the test work have shown that when the IsaMillTM mill is operated at different speeds the energy required to grind UG2 ore of feed F80 = 120μm to a given product size (P80) varies. It was seen that lower speeds (1500 rpm) required more energy for given product size, increasing the speed to 1800 rpm resulted in a decrease in energy and further increase led to higher energy utilization. This indicates an existence of optimum speed when grinding UG2 ore of F80 = 120μm at 1800 rpm. In terms of the effect of media load, it was found that different media loads are required to efficiently grind various feed sizes (F80) of UG2 ore to desired product. Therefore, the IsaMillTM media load should be optimized at different levels for efficient grinding process. This study has indicated that the optimum media load is dependant on feed particle size. This test work also indicates that the performance of the IsaMillTM is greatly affected by the media size and feed particle size distribution. For the same stirrer speed, media load and slurry percent solids, it was found that the best grinding efficiency when grinding UG2 ore of feed sizes F80 = 55μm and F80 = 120μm were achieved when the 2mm media was used. However, the 3.5mm media was the most efficient media to grind the UG2 ore of F80 = 250μm. Therefore, the ratio of media to feed size must be matched and optimized in order to maximize the grinding efficiencies. Comparison of the data obtained from the sampling campaigns conducted on M10000 and the M4 IsaMillTM test work exhibited a consistent behaviour of industrial IsaMillTM that is closely matched by results achieved using the M4 mill. This suggests that the M4 laboratory scale IsaMillTM can be utilized to accurately estimate the energy required to prepare desired product fineness in M10000 industrial scale IsaMillTM. Therefore, the M4 can be used to generate data for design and operations optimization of industrial scale IsaMillTM.

Hypersaline inorganic brines are generated from many global mining operations and the volume of these brines is increasing at an exponential rate. The environment and water resources in the vicinity of these mining operations are at a risk of being polluted as a result of this increase in brine volume. These are the key reasons why these brines need treatment. Eutectic Freeze Crystallization (EFC) has been identified as a possible novel brine treatment method, but to date it has not been applied to multi-component streams such as brines. Therefore, the aim of this thesis was to develop a brine treatment protocol and to demonstrate the "proof of concept" of EFC as a brine treatment method. Three key aspects essential to the brine treatment protocol were identified as being crucial to the treatment process. These key aspects were brine analysis, thermodynamic modelling and kinetic aspects. A combination of standard water analysis techniques and wet chemistry were used to characterize the brine, while OLI Stream Analyser was used to perform the thermodynamic modelling of the brine. It was found that the difference between the total cations and total anions (ion imbalance) from the analysis of two brine samples, Brine 1 and Brine 2, were 5.8% and 6.3% respectively. The brines were also very dilute with a total dissolved solid content of 29.77g/L for Brine 1 and 31.26g/L for Brine 2. OLI Stream Analyser was able to predict and simulate the phase equilibria of an aqueous system over a wide temperature range by estimating the standard state terms and the excess terms with the use of various thermodynamic frameworks. This was an important step because the identities of the potential salts, the temperatures at which they would crystallize and the potential yields of the various products could be predicted before any experiments were conducted. The thermodynamic modelling predicted that the brine samples were saturated with respect to CaS04·2H20. The modelling also predicted that ice, Na2S04·10H20 as well as K2S04·CaSOdH20 would crystallize in a narrow temperature range from -O.8°C to -2.2°C. The thermodynamic results also showed that a high overall ion recovery (85% for Brine 1 and 71 % for Brine 2) would be obtained at an operating temperature of -5°C. However, the thermodynamics merely offered a prediction. It was only by investigating the kinetic aspects of the system that the identity, crystallization temperatures and yield of products could be confirmed. The kinetic aspects incorporated three phases. The first phase focused on determining the metastable zone (MSZ) of ice in a binary sodium sulphate solution, as well as the MSZ of ice for the brine. This essentially defined the operating region in which heterogeneous ice nucleation could occur. The results confirmed the inherently stochastic nature of nucleation and showed that a number of experiments were needed in order to define the MSZ. The MSZ for a 1 wt% sodium sulphate solution (1.8ml volume) was 2.56 ~oC for a cooling rate of 2°C/hour, 2.76 ~oC for a cooling rate of 4°C/hour and 4.76 ~oC for a cooling rate of 8°C/hour. The effect of solution volume on the nucleation process and hence the MSZ was determined. The MSZ for a 250ml, lwt% solution and a 1000ml, lwt% solution were similar (2.3 ~OC and 2.2 ~OC respectively). The second kinetic phase investigated the problem associated with the crystallization of two salts and ice. The EFC process is based on the principle that ice can be separated from a salt because of their density differences, thus producing pure products. However, if two salts crystallize at the same conditions, then salt contamination would occur (similar densities for the salts). This problem was avoided by utilizing knowledge obtained from phase diagrams and by seeding with a specific salt. It was found that seeding as a separation technique was feasible in a ternary Na2S04-MgS04-H20 system. The addition of Na2SOd OH20 seeds to a supersaturated solution at 12°C resulted in pure Na2S04.10H20 (96% purity) being formed. The third kinetic phase focused primarily on the sequential removal of pure salts from a single brine during EFC conditions. The experimental work showed that pure calcium sulphate (98.0% purity), pure sodium sulphate (96.4% purity) and potable water (ice) could be formed with a brine mass reduction of ~97%. The problem with the brines initially being saturated with respect to calcium sulphate was also solved by successfully removing calcium sulphate and ice under EFC conditions. This meant that pure sodium sulphate and ice could be removed in the subsequent stage. This thesis ultimately showed that EFC could be used to treat multi-component streams and that pure salts could be sequentially produced along with potable water.

The Anglo Research Nickel (ArNi) process is a novel extractive metallurgical process that arose out of the need to develop a processing route for the recovery of nickel from lateritic ore deposits that is both economical and environmentally acceptable. Kieserite (MgSO4·H2O) crystallisation is a critical step in the process which leads to the regeneration of reagents (HCl, H2SO4 and MgO). Hence, the regeneration of reagents is dependent on the amount of magnesium sulphate that precipitates out within the limits of the operating conditions. These conditions include temperature and the ioninteractions of the background aqueous environment. Hence, by manipulating these parameters the optimal region and hence, operating conditions where the minimum solubility of the solute lies can be identified. This novel ArNi process demonstrates the power of manipulating aqueous chemical environments in order to regenerate reagents and hence, develop more sustainable processes. Thus, the ArNi process has provided the building blocks to reinventing the way mining processes are designed, implemented and perceived. Therefore, in order to develop a broader understanding of how aqueous environments can be manipulated in order to process different types of ores, the solubility of NaCl in hypersaline brines was also investigated as a 2nd model system. Temperature and ion interactions are the most important factors affecting both the solubility of slightly soluble magnesium sulphate, and highly soluble sodium chloride salts, as well as for the type of hydrate formed. However, there is a lack of data for the thermodynamic properties of these salts in multi-component systems, especially their solubilites at high temperatures. Thus, there is scope for the development of a better understanding of the ion interactions in multi-component systems under different aqueous environments and temperature conditions, and how these affect the precipitated solute. To achieve the objectives of the study, experiments were conducted in 450 ml glass reactors. The desired operating temperatures were attained using heating bands and maintained with temperature controllers. Spiral reflux condensers were fitted to condense any vapour that evolved and ensure that the volume of solvent remained constant. Face-centred central composite designs and central composite factorial designs were adopted for the FeCl3-MgCl2-HCl-MgSO4-H2O and ZnCl2-HCl-NaCl-H2O systems respectively. The factors that were varied were the concentrations of FeCl3, MgCl2 and HCl for the FeCl3-MgCl2-HCl-MgSO4-H2O system at 105°C and the concentrations of ZnCl2 and HCl at temperatures of 40°C, 80°C and 107°C for the ZnCl2- HCl-NaCl-H2O system. The measured responses were the solubility of MgSO4 and NaCl. Characterisations of the hydrates of MgSO4 that formed under different aqueous environments were also established for the FeCl3-MgCl2-HCl-MgSO4-H2O system. FeCl3-MgCl2-HCl-MgSO4-H2O system Statistical analysis of each of the factorial phases established that a second order model best fits the experimental data and accounts for 98.3%, 96.1% and 98.3% of the variation in the solubility of MgSO4. Within each phase, MgCl2 concentration had the most significant effect on the solubility of MgSO4 of all the varied factors. MgCl2 suppressed the solubility of MgSO4 due to the presence of the common Mg2+ ion. HCl had the opposite effect on the solubility of MgSO4 i.e. increasing the concentration of HCl resulted in an increase in the solubility of MgSO4 due to an increase in ionic strength. At low concentrations of MgCl2 and HCl, increasing the concentration of FeCl3 decreased the solubility of MgSO4 due to the bond formation between SO4 2- ions and ferric hydroxyl complexes. At high concentrations of MgCl2 and HCl, increasing the concentration of FeCl3 had a minimal effect on the solubility of MgSO4. MgSO4·H2O precipitated independently or with a combination of MgSO4·1.25H2O or MgSO4·6H2O at each of the different concentration limits of MgCl2, FeCl3 and HCl. The FeCl3 factor did not have an influence on the hydrate or hydrates that formed. However, the presence of MgCl2 and HCl had a dehydrating action on the formation of the hydrates with HCl having a more pronounced effect. ZnCl2-HCl-NaCl-H2O system Statistical analysis of the central composite factorial designs at each of the temperatures investigated, found that a linear order model best fits the experimental data and accounted for 69.1%, 62.7% and 55.1% of the variation in the solubility of NaCl. The concentration of ZnCl2 had the most pronounced influence on the solubility of NaCl. An increase in the concentration of ZnCl2 increased the solubility of NaCl on account of the formation of homo-polar bonds which decreases the ionization of the solution. The increase in the concentration of HCl decreased the solubility of NaCl. The effect of temperature did not have a significant effect on the solubility of NaCl because of the flat solubility line. The findings in this study have shown that ion interactions play a crucial role in the solubility of salts in hypersaline brines. In addition, each ion has a different effect (common ion effect, ionic strength effects or complex formation) on the solubility of a specific salt and is unique to its individual system. Thermodynamic modelling can predict salt solubility trends. However, in order to gain a fundamental understanding of a system, especially complex systems, experimental measurements are a necessity. The experimental measurements provide an in-depth understanding of specific systems which can lead to the manipulation of aqueous environments towards the development of more sustainable processes and hence, a whole new approach to extractive metallurgy.