Thesis (PhD) -- Stellenbosch University, 2014. / ENGLISH ABSTRACT: The process route for the production of base and platinum-group metals from natural sulfide ores commonly requires the conversion of high-iron furnace matte into an iron-lean converter matte. This is followed by pre-treatment through cooling of the iron-lean molten matte, physical processing of the solidified matte and hydrometallurgical metal extraction. Lonmin is the third largest producer of platinum-group metals in the world and utilizes Peirce-Smith converters for blowing high-iron furnace matte with air to a final iron concentration or endpoint. The molten matte is water granulated and solidification occurs via fast-cooling. The solidified matte is ground in a closed circuit ball mill with hydrocyclone classification and subjected to first stage atmospheric leaching. The specification of an ideal or desirable converter iron endpoint requires careful consideration. Most importantly, it must ensure the crystallization of converter matte with mineralogical qualities that are within the setpoints of the downstream unit processes and techniques. An additional consideration is for the final blown converter matte to achieve an optimum bulk concentration of the base metals Ni and Cu and platinum-group metals Pt, Pd, Rh, Ru and Ir. Mattes characteristic of variable iron endpoints were regularly produced at the Lonmin converter plant section. Uncertainty by plant metallurgists in knowing the desirable iron endpoint, particularly within the context of the Lonmin base metal refinery, and poor control has had detrimental effects on the mineralogical quality of the final matte and hence on the processing characteristics of the solidified matte particles downstream. A desirable iron endpoint required investigation, selection and implementation at Lonmin. The primary focus of this study was therefore to quantify the effect of a specific iron endpoint on the mineralogy and mineral chemistry of solidified converter matte. A fundamental examination of the solidification process upon cooling was regarded as critical to an in-depth understanding of the attained mineralogy and mineral chemistry as a function of a specific iron endpoint. It became equally important to quantify the effect of the resultant mineralogy, and hence iron endpoint, on the physical property of mineral structures in relation to downstream grinding, liberation and leaching characteristics. Despite considerable industry context, limited in-depth and coherent studies on the effect of a specific iron endpoint on fast-cooled converter matte systems were found in both industrial and scholarly literature. Previous findings in literature offered a limited quantitative understanding of the effect on mineralogy and mineral chemistry. Phase and cooling equilibria of multi-component, iron endpoint specific Ni-Cu-S matte systems were also not fully available. These would have been particularly useful in understanding the complexities of converter matte solidification as a function of iron endpoint. Physical property knowledge of converter matte mineral structures was hardly available and even less so in relation to grinding, liberation and leaching processes. A comprehensive investigation was therefore required to address these extensive knowledge gaps with respect to fastcooled converter matte systems in an industrial framework. Three Peirce-Smith converter production samples, representative of the extent in variability of iron endpoints attained at the converter plant, were used in a systematic investigation coupled to a novel combination of modern analytical techniques, computational thermochemistry and metallurgical testwork. The modern analytical techniques included the application of high resolution transmission electron microscopy and focused ion beam scanning electron microscopy tomography. Computational thermochemistry was applied through the use of MTDATA phase diagram software. Metallurgical testwork involved laboratory batch grinding at various specific energies. Closely associated leach experiments were also considered relevant to this wide-ranging investigation. The Peirce-Smith converter samples investigated were indicative of mattes that attained specific endpoints of 5.17%, 0.99% and 0.15 weight% Fe. The highest combined bulk concentration of the important base and platinum-group metals was achieved in the matte which attained a specific iron endpoint of 0.99%. The mineralogy of all three converter mattes was dominated by nickel sulfide mineral structures matched to the natural mineral of heazlewoodite. Mineral structures of copper sulfide, NiCu-alloy, spinel and OsRu-alloy were also constituents of the different converter mattes. The attainment of a specific iron endpoint was found to result in measurable mineralogical differences with respect to relative mineral abundances, external morphological characteristics and mineral chemistry. The mineralogical differences were particularly distinct between mineral structures of the high (5.17%) and low (0.99% and 0.15%) iron mattes. Subtle mineralogical differences were evident between mineral structures of the low iron mattes. The 0.99% Fe matte was characteristic of a significantly higher NiCu-alloy relative abundance, compared to the 5.17% Fe matte. The NiCu-alloy structures were found to act as the primary collectors of the economically significant platinum-group metals. Mineralogical observations were used to develop an understanding of the underlying mineralization mechanism of NiCu-alloy structures. High-fidelity color and grayscale 3D reconstructions were produced of the resultant mineralized structures. It was shown theoretically that variations in iron endpoint specific starting compositions of oxygen-free liquid matte systems alter the solidification pathway towards the eutectic. Moreover, a quantitative understanding of liquid phase solidification of the high and low iron matte systems, including oxygen, was developed to within ±2.5 oC. Most of the specific energy available for grinding was expended breaking the nickel sulfide matrix, particularly of the high iron matte. The breakage rates of copper sulfide mineral structures in the 5.17% Fe matte were calculated to be higher than in the 0.15% Fe matte at 25kWh/t specific energy. The degree of copper sulfide liberation was shown to be higher for the 5.17% Fe matte than for the 0.15% Fe matte at the same specific energy of grinding. A higher degree of Ni extraction and Cu cementation could be achieved when leaching low iron matte particles. The production of converter matte attaining a specific iron endpoint of 0.99% was found to be the most suitable with respect to endpoint selection criteria. A practical iron endpoint range of 1.6% to 1.0% was recommended for the production of converter matte with a resultant mineralogical quality within the constraints of the Lonmin base metal refinery. This study offers an integrated understanding of base and platinum-group metals production as a function of a desirable iron endpoint at Lonmin. This was not previously available in metal production literature. New technology for the monitoring and consistent control of such a practical iron endpoint range can subsequently be implemented. / AFRIKAANSE OPSOMMING: Die prosesroete vir die produksie van onedel en platinumgroepmetale uit natuurlike swawelertse vereis gewoonlik die omsetting van ’n ysterryke hoogoondmat in ’n ysterarm omsettermat. Hierna volg voorbehandeling deur die afkoeling van die ysterarm gesmelte mat, fisiese verwerking van die soliede mat, en hidrometallurgiese metaalekstraksie. Lonmin is die derde grootste produsent van platinumgroepmetale ter wêreld en gebruik Peirce-Smith-omsetters om ysterryke hoogoondmat met lug te blaas totdat dit ’n finale ysterkonsentrasie- of ystereindpunt bereik. Die gesmelte mat word met water granuleer, en solidifikasie vind deur middel van snelafkoeling plaas. Die soliede mat word in ’n geslotekringbalmeul met hidrosikloonklassifikasie gemaal en aan eerstestadium- atmosferiese loging onderwerp. Die spesifikasie van ’n ideale of gewenste ystereindpunt verg deeglike oorweging. Bowenal moet dit verseker dat die omsettermat kristalliseer met mineralogiese eienskappe wat binne die setpunte van die eenheidsprosesse en - tegnieke verder af in die prosesstroom val. ’n Bykomende oorweging is dat die uiteindelike geblaasde omsettermat ’n optimale massakonsentrasie van die onedel metale Ni en Cu en die platinumgroepmetale Pt, Pd, Rh, Ru en Ir moet bevat. Matte met die kenmerke van wisselende ystereindpunte is gereeld by die Lonminomsetteraanleg geproduseer. Die onsekerheid van metallurge by die aanleg oor die gewenste ystereindpunt – veral binne die konteks van die Lonmin-raffinadery vir onedel metale – sowel as swak beheer het ’n nadelige uitwerking gehad op die mineralogiese gehalte van die uiteindelike mat, en dus ook op die verwerkingskenmerke van die soliede matdeeltjies verder af in die prosesstroom. Die bepaling van die gewenste ystereindpunt het sorgvuldige ondersoek, seleksie en toepassing deur Lonmin vereis. Hierdie studie is dus hoofsaaklik uitgevoer om die uitwerking van ’n spesifieke ystereindpunt op die mineralogie en minerale chemie van soliede omsettermat te kwantifiseer. ’n Grondliggende ondersoek na die solidifikasieproses by afkoeling is as noodsaaklik beskou vir ’n diepgaande begrip van die verworwe mineralogie en minerale chemie as ’n funksie van ’n spesifieke ystereindpunt. Mettertyd het dit egter ewe belangrik geword om die uitwerking van die gevolglike mineralogie, en dus die ystereindpunt, op die fisiese eienskappe van minerale strukture met betrekking tot maling-, vrystellings- en loogprosesse verder af in die prosesstroom te kwantifiseer. Ondanks heelwat bedryfskonteks, het nóg bedryfs- nóg vakkundige literatuur veel diepte- en samehangende studies oor die uitwerking van ’n spesifieke ystereindpunt op snelafgekoelde omsettermatstelsels opgelewer. Vorige bevindinge in die literatuur het boonop ’n beperkte kwantitatiewe begrip van die uitwerking op mineralogie en minerale chemie getoon. Die fase- en afkoelingsekwilibriums van ystereindpuntspesifieke Ni-Cu-S-matstelsels met veelvuldige komponente was ook nie ten volle beskikbaar nie. Dít sou veral goed te pas gekom het om die kompleksiteite van omsettermatsolidifikasie as ’n funksie van ystereindpunt te verstaan. Kennis van die fisiese eienskappe van die minerale strukture van omsettermat was kwalik beskikbaar, terwyl selfs minder inligting oor maling-, vrystellings- en loogprosesse opgespoor kon word. Daarom was ’n omvattende ondersoek nodig om hierdie beduidende kennisleemtes met betrekking tot snelafgekoelde omsettermatstelsels in ’n nywerheidsraamwerk aan te vul. Drie Peirce-Smith-omsetterproduksiemonsters wat die wisselende bestek van ystereindpunte by die omsetteraanleg verteenwoordig, is in ’n stelselmatige ondersoek gebruik, tesame met ’n vernuwende kombinasie van moderne ontledingstegnieke, gerekenariseerde termochemiese bewerkings en metallurgiese toetswerk. Die moderne ontledingstegnieke sluit onder andere in hoëresolusie-transmissie-elektronmikroskopie (HRTEM) en gefokusdeioonstraalskandering-elektron-mikroskopie (FIB SEM) tomografie. Die gerekenariseerde termochemiese bewerkings is met behulp van MTDATAfasediagramsagteware uitgevoer. Metallurgiese toetswerk het die maling van laboratoriumlotte teen verskillende spesifieke energieë behels. Nou verwante loogproefnemings is ook as relevant vir hierdie omvattende studie beskou. Die bestudeerde Peirce-Smith-omsettermonsters het op matte met spesifieke eindpunte van 5.17%, 0.99% en 0.15 gewig% Fe gedui. Die hoogste gekombineerde massakonsentrasie van die belangrike onedel en platinumgroepmetale is in die mat met ’n spesifieke ystereindpunt van 0.99% gevind. Die mineralogie van ál drie omsettermatte is oorheers deur die minerale strukture van nikkelsulfied, wat met die natuurlike mineraal heazlewoodiet ooreenstem. Die verskillende omsettermatte het ook die minerale strukture van kopersulfied, NiCu-allooi, spinel en OsRu-allooi bevat. Daar is bevind dat die verkryging van ’n spesifieke ystereindpunt tot meetbare mineralogiese verskille in die relatiewe volopheid van minerale, die eksterne morfologiese kenmerke sowel as minerale chemie lei. Die mineralogiese verskille was veral duidelik te sien tussen die minerale strukture van die ysterryke (5.17% Fe) en ysterarm (0.99% en 0.15% Fe) matte. Fyn mineralogiese verskille is ook tussen die minerale strukture van die ysterarm matte bespeur. Die 0.99% Fe-mat het tipies beduidend meer NiCu-allooi as die 5.17% Fe-mat bevat. Die NiCu-allooistrukture tree oënskynlik op as die hoofversamelaars van die ekonomies belangrike platinumgroepmetale. Mineralogiese waarnemings is gebruik om ’n begrip te ontwikkel van die onderliggende mineralisasiemeganisme van NiCuallooistrukture. Die gevolglike gemineraliseerde strukture is met behulp van driedimensionele rekonstruksies met hoë kleurgetrouheid sowel as in grysskaal voorgestel. Daar is teoreties aangetoon dat variasies in ystereindpuntspesifieke beginsamestellings van suurstofvrye vloeibare matstelsels die solidifikasieroete na die eutetikum wysig. Daarbenewens is die vloeifasesolidifikasie van die ysterryke en ysterarm matstelsels, wat suurstof insluit, op sowat ±2.5 oC gekwantifiseer. Die meeste van die spesifieke energie wat vir maling beskikbaar was, is gebruik om die nikkelsulfiedmatriks te breek, veral vir die ysterryke mat. Berekeninge toon dat die breektempo’s van die minerale strukture van kopersulfied by die 5.17% Fe-mat hoër was as by die 0.15% Fe-mat teen ’n spesifieke energie van 25 kWh/t. Die mate van kopersulfiedvrystelling was hoër by die 5.17% Fe-mat as by die 0.15% Fe-mat by dieselfde spesifieke energie vir maling. ’n Hoër mate van Ni-ekstraksie en Cu-sementasie is verkry toe ysterarm matdeeltjies geloog is. Wat eindpuntseleksiemaatstawwe betref, is die produksie van ’n omsettermat met ’n spesifieke ystereindpunt van 0,99% as die mees geskikte aangewys. ’n Praktiese ystereindpuntbestek van 1.6% tot 1.0% word aanbeveel vir die produksie van ’n omsettermat met ’n gevolglike mineralogiese gehalte wat binne die perke van die Lonmin-raffinadery vir onedel metale val. Hierdie studie bied ’n geïntegreerde begrip van die produksie van onedel en platinumgroepmetale as ’n funksie van ’n gewenste ystereindpunt by Lonmin. Hierdie inligting was nie voorheen in literatuur oor metaalproduksie beskikbaar nie. Nuwe tegnologie vir die monitering en konsekwente beheer van so ’n praktiese ystereindpuntbestek kan dus op grond hiervan in werking gestel word.

Thesis (MScEng (Process Engineering))--University of Stellenbosch, 2011. / Please refer to full text to view abstract.

Thesis (MScEng)--Stellenbosch University, 2011. / ENGLISH ABSTRACT: In a first stage atmospheric leach at the Lonmin Marikana base metals refinery, nickel-copper-iron-sulphur Peirce Smith converter matte is leached in recycled electrolyte from the electrowinning section. The electrolyte contains sulphuric acid, copper and nickel sulphates, and a small amount of iron sulphate. The converter matte contains mostly nickel, copper and sulphur (typically 48 %, 28 % and 23 %, respectively), but also minor amounts (<5 %) iron and cobalt. The matte also contains platinum group elements (PGEs) and other precious metals totalling 0.2 – 0.7 % (platinum, palladium, iridium, rhodium, ruthenium, osmium and some gold). The predominant mineral phases are heazlewoodite, chalcocite and a nickel-copper alloy phase, as well as some entrained slag and spinel minerals. The purpose of the first stage leach is to extract nickel, while simultaneously precipitating copper and PGEs contained in the recycled electrolyte. Nickel, cobalt and iron are leached by acid and oxygen. Copper is precipitated by a redox reaction in which copper ions oxidise nickel from the matte. The purpose of this study was to determine the effects of key variables on the performance of the first stage leach (specifically on the removal of PGEs and copper from solution and the overall extraction of nickel) and to improve fundamental understanding of these effects. Batch leaching tests were carried out to investigate the effects of the following factors: availability of oxygen, initial acid concentration, initial copper concentration, iron endpoint (iron content of the matte), solids/liquid ratio and stirring rate. Liquid samples were analysed with Atomic Absorption Spectroscopy (AA) to determine leaching kinetics. Characterisation of solid samples from leach tests by quantitative X-Ray diffraction (XRD) and scanning electron microscopy with an energy dispersive system (SEM-EDS) helped to improve understanding of the leaching mechanism. The oxidative leaching mechanism entails an initial period in which the alloy phase is leached by acid and oxygen, while copper reacts with the nickel-copper-alloy and heazlewoodite phases (which react galvanically with each other) to form a chalcocite precipitate. In a second reaction period, heazlewoodite was transformed to millerite by acid leaching and the particle structure became more porous. The rate of copper precipitation and nickel extraction were faster during the second reaction period than the first reaction period. Some copper leaching occurred once the leachable nickel (60 – 70 %) had been dissolved, provided that the solution was strongly acidic (pH < 2). The non-oxidative leaching mechanism entails a galvanic interaction, between the nickel-copper-alloy and heazlewoodite phases, in which nickel is leached from both phases and copper is precipitated as chalcocite. Leaching by acid was negligible in most non-oxidative tests. An initial fast period of copper precipitation was followed by a second slower period. The decrease in reaction rate can probably be linked to the decreasing availability of the nickel-copper-alloy phase. During non-oxidative leaching, the particle structure remained mostly intact. Copper precipitation kinetics under non-oxidative conditions was found to be slower than under oxidative conditions. The faster copper precipitation kinetics under oxidative conditions is most likely caused by an increase in porosity and reaction area as nickel is leached from the matte by acid and oxygen. The initial acid concentration, solids/liquid ratio and Fe-endpoint were the most important factors determining reaction kinetics under oxidative conditions. Low initial acid concentrations (37 g/L) and a high solids/liquid ratio improved the extent of copper precipitation. Nickel extraction was enhanced by low solids/liquid ratios and high initial acid concentrations (74 g/L). Nickel extraction was significantly less (56 % less in one instance) when leaching high iron mattes (5.7 % Fe) rather than low iron mattes (< 1 % Fe). Copper precipitation was initially faster when leaching a high iron matte, but slower nickel leaching from high iron mattes led to an excess of available acid, which resulted in copper being leached. The results suggest that high iron mattes will lead to poor copper and PGE precipitation in the first stage leach and also to lower nickel extractions. Consequently, Peirce Smith converting at the plant must be carefully controlled to avoid high iron mattes. Under non-oxidative conditions, the solids/liquid ratio and Fe-endpoint were the most important factors. The rate of copper precipitation was faster when a high iron matte was leached, so that a higher percentage copper was precipitated and more nickel was extracted from the matte. / AFRIKAANSE OPSOMMING: As ‘n eerste stap in die Lonmin Marikana basis-metale veredelingsaanleg word nikkel-koper-yster-swawel Peirce-Smith-converter-mat geloog in elektroliet wat hersirkuleer word vanaf die aanleg se koper-elektroplaterings-afdeling. Die loging word by atmosferiese druk uitgevoer. Die elektroliet bevat swawelsuur, koper- en nikkel-sulfate en ‘n klein hoeveelheid ystersulfaat. Die mat bevat hoofsaaklik nikkel, koper en swawel (tipies 48 %, 28 % en 23 %), maar ook klein hoeveelhede (< 5 %) yster en kobalt. Verder maak Platinum Groep Elemente (PGE’s) en ander waardevolle metale (platinum, palladium, iridium, rhodium, ruthenium, osmium en goud) 0.2 % tot 0.7 % van die massa van die mat uit. In terme van minerale bestaan die materiaal hoofsaaklik uit heazlewoodite, chalcocite en ‘n nikkel-koper allooi fase, asook slak en spinel minerale, wat tydens Peirce-Smith-converting weens meesleuring in die mat rapporteer. Die doel van die eerste stadium loog is om nikkel op te los, terwyl koper en PGE’s wat in die elektroliet voorkom presipiteer moet word. Nikkel, kobalt en koper word geloog in reaksies met suurstof en swawelsuur. Koper word presipiteer deur middel van ‘n redoks reaksie waarin koper-ione nikkel in die mat oksideer. Die doel van hierdie studie was om die effekte van sleutelveranderlikes op die proses te bepaal (spesifiek hoe nikkel-loging en koper presipitasie affekteer word) en om fundamentele begrip van die veranderlikes en hul effekte te verkry. Lot loogtoetse is uitgevoer op ‘n laboratorium-skaal en die effekte van die volgende faktore is ondersoek: beskibaarheid van suurstof, begin suurkonsentrasie, yster eindpunt (die ysterinhoud van die mat), vastestof/vloeistof verhouding en die roertempo. Vloeistof monsters geneem tydens loogtoetse is geanaliseer met behulp van Atoom Absorpsie Spektroskopie (AA) om kinetika te bepaal. Vastestof monsters is ook geneem tydens loogtoetse en kwantitatiewe X-straal diffraksie (XRD), asook skanderings-elektron-mikroskopie met ‘n energie dispersie sisteem (SEM-EDS) is gebruik om die materiaal te karakteriseer en die logingsmeganisme te verduidelik. Die oksidatiewe logingsmeganisme behels ‘n aanvanklike periode waartydens die allooi fase geloog word deur suur en suurstof, terwyl koper presipiteer om chalcocite te vorm as gevolg van ‘n reaksie waarin galvanise interaksie tussen die nikkel-koperallooi en heazlewoodite fases ‘n belangrike rol speel. In ‘n tweede reaksie periode is heazlewoodite geloog deur suur om millerite te vorm. Tydens hierdie tweede fase het die partikel struktuur meer porieus geword. Die tempo van koper presipitasie en nikkel loging was vinniger tydens die tweede reaksie periode as tydens die eerste. Koper is geloog indien die oplossing baie suur was (pH < 2) en die loogbare nikkel (60 – 70 %) reeds opgelos het. Die nie-oksidatiewe logingsmeganisme behels galvaniese interaksie tussen die nikkel-koper-allooi en heazlewoodite fases, wat lei tot koper presipitasie as chalcocite. Loging deur swawelsuur was onbeduidend. ‘n Aanvanklike vinnige periode van koper presipitasie tydens nie-oksidatiewe toetse is gevolg deur ‘n tweede stadiger periode. Die afname in reaksietempo kan waarskynlik verklaar word deur die afnemende beskikbaarheid van die nikkel-koper-allooi fase. Tydens nieoksidatiewe loging het die partikel struktuur redelik onveranderd gebly. Koper presipitasie kinetika in nie-oksidatiewe toetse was stadiger as in oksidatiewe toetse. Die belangrikste faktore wat kinetika in oksidatiewe toetse beïnvloed het was die suurkonsentrasie, vastestof/vloeistof verhouding en die yster-eindpunt. Lae beginsuurkonsentrasies (37 g/L) en ‘n hoë vastestof/vloeistof verhouding het gelei daartoe dat meer koper uit die elektroliet herwin is. Nikkel ekstraksie was hoër indien die vastestof/vloeistof verhouding laag was en die begin suurkonsentrasie hoog (74 g/L). Nikkel ekstraksie was beduidend laer (56 % laer in een geval) wanneer hoë-yster mat (5.7 % Fe) geloog is, eerder as lae yster mat (< 1 % Fe). Wanneer ‘n hoë yster mat geloog is, was koper presipitasie aanvanklik vinniger, maar weens stadige nikkel-ekstraksie-tempos was ‘n oormaat van suur beskikbaar sodat koper uiteindelik geloog is. PGE presipitasie is ook nadelig beïnvloed wanneer koper geloog is en veral tydens toetse met hoë yster mat. Die mees belangrike faktore wat nie-oksidatiewe loging beïnvloed het was die vastestof/vloeistof verhouding en die yster-eindpunt. Die tempo van koper presipitasie was vinniger in toetse met ‘n hoë yster mat, sodat ‘n hoër persentasie koper presipiteer het en meer nikkel opgelos het wanneer ‘n hoë yster mat geloog is.