The IronArc process is a novel process for a more sustainable production of liquid pig iron using electricity for heating and hydrocarbons for reduction. This thesis aims to facilitate its use by investigating possible refractory solutions and the gas blowing in the process which is done by a plasma generator. The process involves a slag with a high FeO content of 90 wt % and gangue content of approximately 5 wt % SiO2 and 5 wt % CaO. The interaction between such a slag and refractories of MgO, Al2O3, Cr2O3, SiC, ASZ,and C was investigated by high temperature experiments at 1700 K and by thermodynamic calculations in Thermo-calc and FactSage. In the high temperature experiments it was found that all of the studied refractory materials experienced signicant wear after 3 h, but the MgO-Al2O3 spinel refractories were the least affected. The thermodynamic calculations show fair agreement to the experiments, with the exception for the Cr2O3-spinel refractory which performed much worse than predicted by thermodynamic equilibrium calculations. It was concluded that the thermodynamic equilibrium calculations in Thermo-calc and Factsage can be used as an indicator for the stability of a refractory material, but with varying accuracy depending on the quality of the data in the database used. Since industrial refractory materials are not viable as refractory for the IronArc process a freeze-lining approach was evaluated by using CFD in ANSYS Fluent. The flow of a slag was simulated through two different designs of slag runner to investigate how well a freeze-lining protects the walls in a region with rapid flow and the cooling required to form and maintain said freeze-lining. It was found that the enthalpy porosity model in ANSYS Fluent in combination with the RSM turbulence model accurately predicts the thickness of a freeze lining when validated against experiments in the CaCl2-H2O system. For optimal protection of the refractory walls the reactor and runner should be designed to minimize the movement close to the walls as high near-wall turbulence will reduce the thickness and stability of the freeze-lining, leading to greater cooling requirements to maintain afreeze-lining. The IronArc process uses a plasma generator to supply heat to the reactor using electricity. By blowing gas and hydrocarbons through an electric arc, superheated gas is formed which when injected into the reactor provides both stirring and heating for the process. To study the behavior of the injected gas a simulation model was developed in OpenFOAM. The model for simulating gas blowing was tested in both incompressible and compressible simulations in the air-water system which were veried against an experimental study in the air-water system and found good agreement. The simulations of the plasma generator blowing were done in the compressible model to account for the high temperature and pressure present in the IronArc process. It was found that the stability of the gas blowing is dependent on the Froude number where low values cause an unstable and pulsating plume and higher values produce a more stable jet. It was also found that the empirical equation for penetration length is only valid for gas blowing with suciently high Froude numbers to produce a jetting behavior. It was found that the transition from pulsating to steady jetting in the IronArc system occurred around Froude numbers of 300 and higher values further increased the stability of the jet. For gas blowing below the transition region, the penetration length of the unstable and pulsating jet will be severely underpredicted by the empirical equation. This behavior must be considered when designing the gas blowing system for the IronArc process as the gas penetration length will signicantly influence the stirring in the reactor. Additionally, a pulsating and unstable jet produces large bubbles which risk coming in contact with the refractory walls which in previous studies has been shown to be very detrimental to the refractory lifetime. A decrease of the inlet diameter for the gas blowing increases the Froude number and the stability of the jet. By implementing the proposed refractory protection by freeze-lining and the small changes to the plasma generator inlet diameter the IronArc process can be developed into a promising industrial process capable of producing liquid pig iron in a more sustainable way. / Sammanfattning IronArc processen är en nytänkande metod för att producera flytande råjärn på ett mer hållbart sätt genom att använda elektricitet för uppvärmning och kolväten för reduktion. Denna avhandling ämnar att utvärdera möjliga metoder för att skydda infordingen i processen och undersöka gasblåsningen i processesen som görs med en plasma generator.   Ett av huvudstegen av IronArc processen är tillverkningen av en slagg med upp till 90 vikts % järnoxid samt 5 vikts % kiseldioxid och 5 vikts % kalciumoxid från gångarten. Interaktionen mellan en sådan slagg och olika infodringar baserade på MgO, Al2O3, Cr2O3, SiC, ASZ, och C undersöktes i högtemperaturexperiment vid 1700 K samt med termodynamiska beräknar i Thermo-calc och FactSage. Experimenten visade att alla de undersökta infodringsmaterialen bröts ned under de 3 timmar de var i kontakt med slaggen, men de två MgO-Al2O3 spinel baserade infodringarna visade högst motståndskraft mot slitaget. De termodynamiska beräkningarna överrensstämde bra med de experimentella resultaten för alla infodringsmaterial förutom den kromoxid baserade infodringen som bröts ned fullständigt trots att de termodynamiska beräkningarna påvisade viss stabilitet. Slutsatsen är att inget av de studerade infodringsmaterialen är bra anpassat för IronArc processen men att metoden som användes för de termodynamiska beräkningarna i Thermo-calc och FactSage kan användas för att ge en indikation om stabiliteten för olika infodringsmaterial i kontakt med slagg. Dock så kommer resultaten av de termodynamiska beräkningarna vara beroende av kvalitén av databasen som används för beräkningen.   Eftersom infodringsmaterialen inte kunde motstå slitaget från slaggen undersöktes en dynamisk infodring för slaggrännan i IronArc processen. Detta gjordes genom att simulera flödet och stelningen av slagg i flödesberäkningar i ANSYS Fluent i två olika typer av slaggrännor. Studien visade att enthalpy-porosity modellen för stelning samt RSM modellen för turbulens kunde förutspå stelningsförloppet i slaggrännan samt beskriva hur väl den dynamiska infodringen skyddar väggen och vilken kyleffekt som krävs för att bibehålla den. Denna modell validerades mot experimentella studier i CaCl2-H2O systemet med god överrensstämmelse. För optimalt skydd av väggarna i IronArc processen borde reaktorn och slaggrännan utformas så att flödet nära väggarna minimeras då ett turbulent flöde nära väggen är negativt för stabiliteten och tjockleken hos den dynamiska infodringen.   IronArc proceesen använder sig av en plasmagenerator för att förse processen med värme via elektricitet. Genom att blåsa gas och kolväten genom en ljusbåge värms gasblandningen och trycks in i reaktorn  vilket ger både värme och omrörning till processen. För att undersöka hur den varma gasen beter sig i reaktorn utvecklades en simuleringsmodell i OpenFOAM. Modellen utformades som både inkompressibel och kompressibel för blåsning av luft i vatten och jämfördes med experiment där gas blåstes i vatten. De båda modellerna överrensstämde bra med de experimentella resultaten och kunde därför användas för att studera gasflödet i IronArc processen. För simuleringen av IronArc processen valdes den kompressibla versionen av simuleringen då den tar hänsyn till de höga temperaturer och tryck som uppstår i reaktorn.   Simuleringarna visade att den inblåsta gasen kan ge en stabil gas-jet om Froude-talet för inblåset är tillräckligt högt. Om Froude-talet för gasblåsningen är för lågt så kommer gasen pulsera på ett instablit sätt och skapa stora bubblor som kommer i kontakt med infodringsmaterialet, vilket tidigare har påvisats orsaka ökat slitage på infodringsmaterialet. För IronArc processen krävdes ett Froude tal på ca 300 eller högre för att skapa en stabil jet av gas, där högre värden vidare ökar gas-jettens stabilitet. Studien visade också att den empiriska ekvationen som används för att beräkna penetrationslängden vid gasblåsning endast är korrekt om gasen är en stabil jet. Om ekvationen används för att beräkna penetrationslängden för gasblåsning med mindre än det krävda Froude talet kommer penetrationslängden kraftigt underskattas vilket kan medföra att fel beslut tas när en process utformas. Genom att minska diametern på dysan som används för gasblåsningen ökas Froude-talet och därmed stabiliteten av gasjetten, vilket gör den mer förutsägbar och bättre för processen.   För att vidare utveckla IronArc processen så bör den undersökta dynamiska infodringen samt de föreslagna modifieringarna till gasblåsningen användas. Då kan en lovande industriell process utformas som har möjlighet att producera flytande råjärn på ett mer hållbart sätt.    Keywords: IronArc, infodringsslitage, plasmagenerator, dynamisk infodring

This study has focused on determining the heat flows in a production blast furnace hearth. This part of the blast furnace is exposed to high temperatures. In order to increase the campaign length of the lining an improved knowledge of heat flows are necessary. Thus, it has been studied both experimentally and numerically by heat transfer modeling. Measurements of outer surface temperatures in the lower part of a production blast furnace were carried out. In the experimental study, relations were established between lining temperatures and outer surface temperatures. These relations were used as boundary conditions in a mathematical model, in which the temperature profiles in the hearth lining are calculated. The predictions show that the corner between the wall and the bottom is the most sensitive part of the hearth. Furthermore, the predictions show that no studied part of the lining had an inner temperature higher than the critical temperature 1150°C, where the iron melt can be in contact with the lining. / QC 20101124

To be able to supervise the quality of a sintering process, furnace control pieces are therefore used. The current furnace control piece is not sensitive enough small variations during insert production. The goal of the project was to find and evaluate a new suitable cemented carbide grade, which better captures temperature variations during sintering process, likewise, evaluate the alternative placing in the production furnace and methods to supervise and follow the status of the sintering process. The cemented carbide grade 592, which is a DQ-grade, captured larger temperature variations during the sintering compared to the current furnace control piece. The process charge for 592 that captured the largest variations also had the highest charge weight, which indicates that the new grade is also sensitive to the charge weight. The purpose of the project was fulfilled when a more suitable cemented carbide grade, 592, was evaluated for both sintering temperatures 1410 ℃ and 1450 ℃.

Duplex stainless steels (DSS), with a ferritic-austenitic microstructure, are used ina wide range of applications thanks to their high corrosion resistance and excellent mechanical properties. However, efficient and successful production and joining of DSS require precise control of processes and an in-depth understanding o frelations between composition, processing thermal cycles, resulting microstructures and properties. In this study laser welding, laser reheating, and laser additive manufacturing using Laser Metal Deposition with Wire (LMDw) ofDSS and resulting weld and component microstructures and properties are explored. In the first part a lean FDX 27 duplex stainless steel, showing the transformation induced plasticity (TRIP) effect, was autogenously laser welded and laser reheated using pure argon or pure nitrogen as shielding gas. The weld metal austenite fraction was 22% for argon-shielding and 39% for nitrogen-shielding in as-welded conditions. Less nitrides were found with nitrogen-shielding compared to argonshielding. Laser reheating did not significantly affect nitride content or austenite fraction for argon-shielding. However, laser reheating of the nitrogen shieldedweld removed nitrides and increased the austenite fraction to 57% illustrating the effectiveness of this approach. Phase fraction analysis is important for DSS since the balance between ferrite and austenite affects properties. For TRIP steels the possibility of austenite tomartensite transformation during sample preparation also has to be considered. Phases in the laser welded and reheated FDX 27 DSS were identified and quantified using light optical microscopy (LOM) and electron backscatter diffraction (EBSD) analysis. An optimized Beraha color etching procedure was developed for identification of martensite by LOM. A novel step-by-step EBSD methodology was also introduced, which successfully identified and quantified martensite as well as ferrite and austenite. It was found that mechanical polishing produced up to 26% strain-induced martensite, while no martensite was observed after electrolytic polishing.In the second part a systematic four-stage methodology was applied to develop procedures for additive manufacturing of standard 22% Cr duplex stainless steel components using LMDw combined with the hot wire technology. In the four stages, single-bead passes, a single-bead wall, a block, and finally a cylinder with an inner diameter of 160 mm, thickness of 30 mm, and height of 140 mm were produced. The as-deposited microstructure was inhomogeneous and repetitive including highly ferritic regions with nitrides and regions with high fractions ofaustenite. Heat treatment for 1 hour at 1100 ̊C homogenized the microstructure, removed nitrides, and produced an austenite fraction of about 50%. Strength, ductility, and toughness were at a high level for the cylinder, comparable to those of wrought type 2205 steel, both as-deposited and after heat treatment. The highest strength was achieved for the as-deposited condition with a yield strength of 765 MPa and a tensile strength of 865 MPa, while the highest elongation of 35% was found after heat treatment. Epitaxial growth of ferrite during solidification, giving elongated grains along the build direction, resulted in anisotropy of toughness properties. The highest impact toughness energies were measured for specimens with the notch perpendicular to the build direction after heat treatment with close to 300 J at -10oC. It was concluded that implementing a systematic methodology with a stepwise increase in the deposited volume and geometrical complexity can successfully be used when developing additive manufacturing procedures for significantly sized metallic components. This study has illustrated that a laser beam can successfully be used as heat source in processing of duplex stainless steel both for welding and additive manufacturing. However, challenges like nitrogen loss, low austenite fractions and nitride formation have to be handled by precise process control and/or heat treatment. / Duplexa rostfria stål (DSS) är viktiga konstruktionsmaterial tack vare derasutmärkta mekaniska egenskaper och goda korrosionsbeständighet. Vid svetsningoch additiv tillverkning krävs noggrann styrning av parametrar och kunskap om processernas inverkan på mikrostrukturen för att uppnå önskade egenskaper.Lasersvetsning, värmebehandling med laser och additiv tillverkning i form av lasermetalldeponering med tråd (LMDw) har därför studerats för DSS. Det duplexa stålet FDX 27 lasersvetsades utan tillsatsmaterial och med argon ellerkväve som skyddsgas. Kvävgasskydd gav mer austenit och färre nitrider änargonskydd. En efterföljande laservärmebehandling löste upp nitriderna då kväve användes som skyddsgas och austenithalten ökade till 57%. Austeniten i FDX 27kan vid deformation omvandlas till martensit. Två metoder för identifiering av martensit utvecklades därför: en färgetsmetod för ljusoptisk mikroskopi samt en metod som utnyttjar bakåtspridda elektroner (EBSD) vid elektronmikroskopi.Som mest bildades 26% martensit vid mekanisk provpreparering medan elektropolerade prover endast innehöll austenit och ferrit. Procedurer togs fram för additiv tillverkning av komponenter, i 22% krom duplexa rostfria stål, med LMDw kombinerat med varmtrådsteknik. Slutprodukten var en 140 mm hög cylinder med 160 mm inre diameter och tjocklek av 30 mm. Mikrostrukturen var inhomogen med periodiskt omväxlande ferritiska områden med nitrider, och områden med stor andel austenit.Värmebehandling under 1 timme vid 1100oC eliminerade nitriderna och gav en homogen struktur med ca. 50% austenit. De mekaniska egenskaperna var, både före och efter värmebehandling, jämförbara med de typiska för motsvarande stål. Högst hållfasthet uppmättes före värmebehandling med sträckgränsen 765 MPa och brottgränsen 865 MPa, medan den största förlängningen var 35% efter värmebehandling. Slagsegheten var upp till 300 J vid -10oC men varierade med hur provstavens brottanvisning var orienterad relativt byggriktningen.Laser är en lämplig energikälla vid svetsning och additiv tillverkning av duplexa rostfria stål. Utmaningar som kväveförlust, låga austenithalter och nitridbildning kan hanteras med noggrann processkontroll och/eller värmebehandling.

As competition in the car market increases, the techniques for car manufacturing are developed and becomes more advanced to be able to keep up with the pace. The development process of car body components has shifted over the years to involve more simulation driven testing than ever before to save time and money in the early stages of development. As the importance of reliable sheet metal forming simulations grows, inconsistencies between simulations and physical stamping can be detrimental to the development time if stamping dies need to be reworked because of poor correlation between physics and simulations.                        The aim of this study is to improve the coherence between physical stamping and the simulation software used by Volvo Cars. The coherence is determined by studying different properties of the result in simulations and comparing them to measurements taken on the corresponding physical stamped parts. A comparison was done between the current standard simulation software, Autoform Forming R8 and a beta version of Autoform Forming R10. The objectives of this study were to compare the sheet thickness, strain, draw-in and ability to predict material failure between the two simulation software to see which of them correlate best to the physical measured parts.                       The workflow consisted of initially setting up the simulations in Autoform Forming R8. Some of the simulations could begin testing right away, while others required needed some geometry rework as the physical tested parts had been stamped with modified stamping dies. When the simulation setups were completed copies of the simulations were taken and run on Autoform Forming R10 to compare with. The simulations were run with a varied Triboform friction models and some of the simulations were run using symmetry to reduce the simulation time. When data was compared Autoform Forming was used when possible and when additional tools were needed the simulated geometries were exported and compared in software such as SVIEW and GOM Correlate.                       The result showed relatively low differences in the comparisons of sheet thickness and major and minor strain as neither of the simulations seemed to give more accurate values compared to the measurements. A slight improvement in the draw-in comparison was found for the Autoform Forming R10 compared to the R8 simulation. In the material failure prediction a major difference was found where the Autoform Forming R10 simulations were better at determining splits than R8. However the splits were only discovered with 2 of the 4 tested friction models in the R10 simulations while the 1 of the 4 simulations indicated risks for a split in the R8 simulation.                       In conclusion the simulations run on Autoform Forming R10 seem to be better at predicting splits and draw-in dimensions while no major differences were found in the comparisons of strain and sheet thickness.

The acceptance of additive manufacturing (AM) depends on the quality of final parts and the process repeatability. Recently, many studies have been dedicated to the establishment of the relationship between the process behavior and material performance. Phenomena such as laser-material interaction, melt pool dynamics, ejecta formation and particle movement behavior on a powder bed are of a particular interest for the AM community as these events directly influence the outcome of the process. Another aspect, which hinders the adoption of AM, is the need for cost-efficient powder materials and their sustainable processing and subsequent recycling.  The research work presented in this thesis, to a certain degree, covers the above mentioned scientific aspects and focuses on the behavior of gas and water atomized steel powders in laser powder bed fusion (LPBF).  Paper I demonstrates a comparative study of dissimilarly-shaped gas and water atomized low alloy steel powders regarding their processability, packing capacities, particle movement behavior and powder performance in LPBF. The impact of chemical composition and morphology of the powders on the process behavior was revealed. Powder spattering and melt pool instabilities were discussed in detail.  Paper II explains the role of ejecta in the recycled powder and the changing behavior of the material due to ejecta pick-up. The impact of multiple powder recycling on the degradation of low alloy steel powder in laser powder bed fusion was studied. Oxygen content, particle size and ejecta occurrence gradually increased after each recycling step and were identified as the main contributors to the property alterations observed in the powder during recycling. In addition, a direct correlation between the increase in oxygen with repeated recycling and a more frequent spatter ejection after each recycle was established.  Paper III is a successor of Paper I and contains a research on the particle movement and denudation behavior on a powder bed when using near-spherical and non-spherical steel powders. The influence of particle morphology on the dynamics of arbitrary-shaped powder particles was studied by applying an analytical correlation formula to calculate the drag force exerted on powder particles of various shape. Particle entrainment of gas and water atomized powders in front of the laser beam was measured, revealing a significant difference in the powder transfer towards the melt pool.

High-temperature abrasion is an expensive issue in industrial fields such as glass and cement production or mining and processing industry. Yet its effects on steel are not well documented. This study investigates and analyses the behaviour of six different steel grades placed in hot abrasive conditions similar conditions encountered in the industry to enables better material selection.  Abrasion tests in a slurry pot were done at room temperature and at 500 °C. Impact and tensile tests were also performed at different temperatures. To complete the mechanical properties evaluation, hardness measurements were executed before and after tempering at 500  °C. Wear rates assessed at room temperature or at 500 °C, are independent of the mechanical properties of the material. At high temperature, it was shown that wear rates and performance of the steels were influenced by tempering and leading to a unique microstructures for all steel grades investigated and equalize their performances. To conclude, high temperature wear of the investigated grades does not depend on their mechanical properties, however, it can be influenced by their tempering resistance. As the temperature increase, steel tempers, its mechanical properties decrease and homogenise with other steel grades' performances, but some grades keep their properties longer at high temperature.

Syftet med denna studie är att undersöka hantverkskunnandet och metallens egenskaper vidarbetet med ciseleringsteknik på olika metallfolier. Genom studien besvaras frågorna: Hursvarar de olika metallfolier när de ciseleras? Hur ser hantverkskunnandet i ciselering ut?Studien kan beskrivas som en kvalitativ undersökning med inspiration av den experimenterandemetoden samt autoetnografiska metoden. Hantverket i studien har genomförts med ettexperimentellt arbetssätt genom bearbetning av koppar, mässing, tenn och aluminiummetallfolier. Den experimentella metoden genomförs genom en kontrollerad undersökning medhjälp av konstanta delar samt variabler med syfte att upptäcka materialens lämplighet iförhållandet till ciselering. Datainsamlingen har skett via observationer av det egna praktiskaarbetet. Dokumentationen består av bilder, skriftligt material i form av en processdagbok samten ifylld matris med förutbestämda kriterier angående materialets relevanta egenskaper.Varje metallfolie har sina speciella egenskaper, men det visar sig också att de metallfoliernasom jag har testat också har vissa gemensamma egenskaper. Alla inköpta metallfolier visade siglämpliga för ciseleringstekniken även om de hade olik hårdhet och stabilitets förmåga. Genommatrisen som producerades under studiens gång möjliggjordes val av lämpligt material för varjeändamål. Erfarenheter skapades genom experimenterandet med både tekniken och materialetmen också genom reflektioner som i sin tur leder till nya handlingar och djupare kunskap. Dettafenomen skedde under hela den undersökande processen och kan vara relevant förinlärningsprocesser även i andra sorts hantverk. I diskussionskapitlet diskuteras studiens resultatutifrån flera perspektiv, såsom tradition som inspiration, ciselering i skolslöjden och ciseleringutifrån ett hållbarhetsperspektiv.

Austenitization followed by quenching is a well-known conventional heat-treating procedure which is widely used on carbon steels with the aim to obtain high strength in as-quenched condition. Such quenching is usually done by immersing a steel product into the cooling medium which provides a uniform cooling of the surface. The cooling rate can be adjusted to a certain degree on a “component” length-scale by using different cooling mediums such as water, oil, polymer solution, etc. However, certain steel products such as beams, pillars in automobile industry or different machinery parts in agriculture require a proper and controllable cooling gradient and thus mechanical property gradient within the product. It is difficult to control the cooling rates locally on the length-scale smaller than the product only by replacing the quenching medium. In addition, quenching by immersing the product into the cooling medium is accompanied by thermal stresses due to the different cooling rates of the surface and the core, and also accompanied by transformation stresses due to the volume change during phase transformations. These stresses may lead to negative effects such as undesired residual stresses or even cracks. Therefore, cooling must be properly optimized and controlled to eliminate these drawbacks. Such a controllable cooling can be performed by several impingements of the water jets onto a hot austenitized surface at certain locations. By controlling the water flow, number of jets, their locations and other parameters, the global and the local cooling rates can be optimized for a specific industrial application later on.  This thesis demonstrates the potential and capability of the water Impinging Jet Quenching Technique (IJQT) to provide a flexible and controllable cooling for both differential and for uniform quenching cases. The test rig of IJQT was developed in the University of Gävle and was used to perform quenching experiments in this study: differential cooling of thick sheets and uniform quenching of bars to different depths. Differential cooling was performed on square-shaped carbon steel sheets with thickness of 15 mm, and the uniform quenching with different flow rates was performed on carbon steel cylindrical bars with 100 mm in diameter. Along with the physical experiments, Comsol Multiphysics 5.6 software was used to solve a 1D heat transfer problem to estimate the cooling rate profile along the radius of the bar. The experiments were verified by observations and characterization of the microstructure using light optical microscopy (LOM), and by examining the mechanical properties through tensile tests and hardness measurements. The results of the quenching experiments and verifications showed a high potential and flexibility of the IJQT in differential cooling case as well as in the uniform quenching case. / <p>Funding agencies: For financial support Sweden’s Innovation Agency Vinnova, SSAB,Väderstad Components, Swedish Knowledge Foundation and Universityof Gävle are acknowledged.</p>

Ferrosilicon containing 50-75% Si and 1.0-5.0% Zr is used as inoculant in the cast iron industry. Zr can be added to liquid ferrosilicon by use of Zr metal or zirconium ferroalloy (FeSiZr). Then the recovery of Zr, i.e. the fraction of Zr transferred from the additive to the ferrosilicon, as well as the hit rate on specification is high. The aim of this study was to investigate the recovery of Zr from zircon sand, ZrSiO4, and zirconia, ZrO2, in comparison to zirconium ferroalloy when added to liquid ferrosilicon with 75% Si at 1600⁰C.  Also the refining effect of the different additives on Al was investigated. The experiments were carried out by stirring samples of controlled amounts of ferrosilicon and Zr additive in a graphite crucible at 1600⁰C and under inert Ar atmosphere for certain amounts of time. The reaction between ferrosilicon and Zr additive was stopped by rapid cooling of the samples. ICP-OES provided the concentration of Zr and Al and LECO O/N the concentration of O. SEM-ESD was used to examine the microstructures of ferrosilicon and Zr additive after experiments. It was found that ZrO2 was reduced by Si at the particle surface to yield dissolved Zr and ZrSiO4. The ZrSiO4 additive decomposed via two simultaneous reactions, one yielding ZrO2, Si and O2 and the other Zr, Si and O2. The recovery of Zr from ZrO2 and ZrSiO4 was significantly lower than from FeSiZr. Of ZrO2 and ZrSiO4, ZrO2 yielded the highest Zr recovery; the difference was much bigger than predicted by thermodynamics. It was discussed if that could be due to a higher reaction rate of the ZrO2, caused by the smaller size (APS 1 µm compared to d50 91 µm) and larger surface area of this addition. It was also found that utilization of density differences to separate the ferrosilicon and Zr additive did not work for zirconia under the same conditions as it worked for zircon sand, although zirconia has a higher density than zircon sand. The reason was the smaller particle size of the ZrO2 powder. No refining of Al was observed.