Equilibrium between titania slags and metallic iron

ELSTAD, H., ERIKSEN, J.M., HILDAL, A., ROSENQVIST, T., and SEIM, S. Equilibrium between titania slags and metallic iron. The 6th International Heavy Minerals Conference Back to Basics, The Southern African Institute of Mining and Metallurgy, 2007. Equilibrium between titania slags and metallic iron H. ELSTAD*, J.M. ERIKSEN, A. HILDAL*, T. ROSENQVIST, and S. SEIM *TINFOS TTI, Tyssedal, Norway TINFOS TTI, Department of Material Science and Engineering (MS&E), Norwegian University of Science and Technology (NTNU), Norway Department of Material Science and Engineering (MS&E), Norwegian, University of Science and Technology (NTNU), Norway Introduction The titanium industry demands high titanium feedstock for their processes. As rutile resources are limited, ilmenite has become the most important titanium source. Ilmenite smelting is an important part of upgrading ilmenite to high titanium feedstock. In the year 2005, upgrading of ilmenite by smelting, producing a slag with 80 90% TiO 2, accounted for approximately 40% of the feedstock to the TiO 2 pigment industry. In ilmenite smelting molten iron is produced together with a slag, which in addition to titanium oxide contains iron oxide as well as various impurity oxides of manganese, magnesium, silicon, aluminium, etc. Titanium is present mainly as TiO 2 but partly also as Ti 2 O 3. It is known that the concentration of Ti 2 O 3 increases with decreasing concentration of FeO, but this dependency is affected by the nature of the impurity oxides. Sigurdson and Moore 1 showed that slags high in CaO and MgO had a lower concentration of Ti 2 O 3 than slags high in SiO 2 for the same amount of FeO. Pistorius 2 showed in 2002 that if the impurity oxides MnO and MgO are counted, on a molar basis, as FeO, and Cr 2 O 3, V 2 O 3 and part of the Al 2 O 3 *), as Ti 2 O 3, industrial slags will have a composition close to that of the M 3 O 5 phase, where M denotes the elements Ti, Fe, Mn, etc, and which has the pseudobrookite structure. However, for equilibrium with liquid metallic iron he computed, by means of a computer program FACT, that the slag equilibrium composition is represented by a curved line in the ternary phase diagram FeTiO 3 -TiO 2 -Ti 2 O 3, but with a significantly higher concentration of trivalent titanium than actually observed in the industrial slags, and he discussed various mechanisms by which the slag during solidification could change its composition to a lower Ti 2 O 3 content, but the possibility that the computed compositions might be wrong was not mentioned. Tranell 3 found in her research on titaniferous silicate slags a lower Ti 3+ / Ti 4+ ratio than that which is computed by the FACT program. The purpose of the present investigation was to study the equilibrium composition of Fe-Ti-O slag in equilibrium with liquid metallic iron, as well as the effects of impurities such as SiO 2, CaO and MgO on the slag composition. Industrial slags from TINFOS Titan and Iron KS (TTI) will also be discussed and their compositions will be plotted in the ternary phase diagram. These are slags produced from Tellnes ilmenite, quite high in magnesium and iron. Experimental In the past, equilibrium studies of the Fe-Ti-O slag system were difficult when it came to producing the slag. It was difficult to find a suitable crucible for the meltings. The highly corrosive slag would attack all known refractories, and the metal would attack metallic crucibles such as those made of platinum and molybdenum. In the present investigations the slag-metal combination was prepared in a high-frequency induction furnace with a vertically segmented water-cooled copper crucible. The crucible is shown in Figures 1a and 1b. It was designed by the Australian Nuclear Science and Technology Organisation (ANSTO) with power from a 75 kva highfrequency generator (750 khz) supplied by Farfield Electronics Pty, Ltd, Australia. This enabled the slag-metal combination to be melted with turbulent stirring and is expected to be isothermal. The entire temperature gradient *An amount of Al 2 O 3 corresponding to 1/3 of the SiO 2 content is assumed to go into the silicate glass phase and is subtracted from the total Al 2 O 3 content. Figure 1a. 3-D sketch of the water-cooled copper crucible EQUILIBRIUM BETWEEN TITANIA SLAGS AND METALLIC IRON 35

the slags E.4 E.6 were cooled more slowly and solidified completely after about 20 minutes. The slags were analysed by X-ray fluorescence (XRF) for their metallic elements, and by wet chemical methods for their contents of Ti 2 O 3 as well as for droplets of metallic iron, which invariably were found suspended in the slag. Furthermore, their content of acid-insoluble oxide was determined, which gave a measure of the amounts of rutile phase in the slag. These analyses were carried out by the TINFOS Titan & Iron K/S Company. Furthermore, the button of metallic iron in the bottom of the slag ingot was weighed and in some cases analysed. Slag samples were also examined by X-ray diffraction (XRD) and microprobe analysis. Figure 1b. Cross-section of melting furnace is believed to be within the freeze lining, about 1 mm thick. Similar studies for oxide melts are described by Hong et al. 11. About ½ kg of slag-metal could be melted in each experiment. Slags with a molar ratio Ti 2 O 3 /FeO of unity were obtained by melting pure TiO 2 (laboratory grade) with a surplus of metallic iron, in which case the reaction: [1] occurred. Slags with lower or higher Ti 2 O 3 /FeO ratios were obtained by addition of either some Fe 2 O 3 or some metallic titanium, in which case the additional reactions: [2] Results (Fe-Ti-O) For slags without impurity oxides the results are given and discussed elsewhere 4, and are summarized in Figure 2. Here the two parameters N Fe = n Fe /(n Fe + n Ti ) and N O = n O /(n Fe + n Ti ) were calculated from the chemical analyses, the letter n denoting the molar concentration of the element in the slag, manganese being counted as iron. In Figure 2 different symbols denote the top, middle and bottom of the slag ingot, and the curve for iron -saturation is drawn between the points for the top and bottom parts. Also shown are the compositions from pilot plant smelting tests on rather pure ilmenite as reported by Swinden and Jones 5 and Pistorius and Coetzee 6, as well as the compositions for equilibrium with metallic iron as computed by Pistorius 2. The present saturation curve agrees closely with the one previously proposed by Rosenqvist 7. All the slags were examined by X-ray diffraction (XRD) and, with the exception of slag E.1, also by microprobe analysis, with the results given in Table I. Here TiO 2-x denotes a phase, which may be either a reduced rutile, or a Magnéli phase (Ti n O 2n-1 ) with some FeO in solid solution. For all slags except E.3 the dominating phase was an M 3 O 5 phase. If the X-ray spacings are compared with those given by Eriksson et al. 8 it appears that it is a solid solution between FeTi 2 O 5 and Ti 3 O 5. For experiment E.3 Ti 3 O 5 [3] occured. Furthermore, some slags were prepared with the addition of either SiO 2, MgO or CaO (slags S.1 S.6). For experiments E.1 E.4 metallic iron was added in the form of soft steel washers, which contained about 1% manganese as the main impurity. This gave rise to a few tenths of a per cent MnO in the slag, and made it possible to see how manganese distributed among the different slag phases. For experiments E.5, E.6 and S.1 S.6 only enough steel washers were used to initiate the slag melting, the remainder being added as high purity electrolytic iron. Iron was added to give a molar ratio O/M of about 1.5, which ensured saturation with metallic iron. The melting time was about 15 minutes for experiments E.1-E.6, and about 25 minutes for S.1 S.6. In order to prevent oxidation the meltings were carried out in an argon atmosphere in a gas-tight chamber, and the slags were allowed to solidify in the crucible under argon. As the freezing of the slag took place from the water-cooled bottom of the crucible and upwards, samples for analysis were in most cases cut from the bottom and/or top part of the solidified slag ingot and were analysed separately. Slags E.1 E.3 and S.1 S.6 were cooled rapidly by turning off the electric power, whereas Figure 2. Composition of slags. Also shown are slag compositions given by Swinden and Jones 5 and Pistorius and Coetzee 6 and computed line for saturation with metallic iron at liquidus temperature according to Pistorius 2, dashed line, and the stoichiometric M 3 O 5 composition dotted line, Eriksen, Robles and Rosenqvist 4 36 HEAVY MINERALS 2007

TableI XRD and microprobe analyses of slags without impurities Slag no. Microprobe analyses, atomic %, normalized to 100% XRD phases Ti Fe Mn O N Fe N O E.1 M 3 O 5 - - - - - - Rutile (trace) - - - - - - Fe (trace) - - - - - - E.2 M 3 O 5 30.0 7.0 0.13 62.0 0.19 1.68 Rutile (trace) - - - - - - Fe (trace) 2.0 92.3 0.00 3,0 - - E.3B Ti 3 O 5 37.6 <0.1 0.05 62.3 <0.001 1.65 Ti 2 O 3 39.0 0.4 0.20 60.4 0.02 1.53 Rutile (trace) - - - - - - Fe (trace) 39.0 90.5 0.50 5.0 - - E.4T M 3 O 5 31.3 7.1 0.05 61.5 0.18 1.60 TiO 2 - x 34.6 1.0 0.02 64.0 0.03 1.80 Fe (trace) - - - - - - E.4B M 3 O 5 34.0 4.0 0.05 62,0 0.11 1.63 E.5T M 3 O 5 28.7 9.1 0.04 62.1 0.24 1.64 M 2 O 3 21.8 17.1 0.90 60.2 0.45 1.51 Rutile 32.6 1.6 0.01 65.8 0.05 1.92 Fe(tr) - - - - - - E.5B M 3 O 5 29.3 8.9 0.02 61.8 0.23 1.61 M 2 O 3 21.3 18.3 0.12 60.0 0.46 1.50 Rutile (tr) - - - - - - Fe(tr) 0.0 97.3 0.00 2.7 - - E.6T M 3 O 5 33.9 3.5 0.02 62.5 0.09 1.67 Ti 3 O 5 (tr) 35.8 1.9 0.02 62.4 0.05 1.66 TiO 2 - x (tr) - - - - - - E.6B M 3 O 5 35.0 2.3 0.01 62.0 0.06 1.66 Ti 3 O 5 33.9 2.0 0.01 63.2 0.06 1.76 TiO 2-x (tr) - - - - - - The M 2 O 3 phase was for experiment E.3 practically pure Ti 2 O 3, and for experiment E.5 practically pure FeTiO 3. In both cases the MnO concentration was found to be significantly higher in the M 2 O 3 phase than in the adjacent M 3 O 5 phase. This shows that manganese is preferentially dissolved in the M 2 O 3 phase, which agrees with the findings of Grey et al. 10. Figure 3a. Back-scatter microprobe pictures for slag E.4 bottom. Light grey: M 3 O 5. dark grey: TiO 2-x phase denotes a phase (anosovite) with a structure slightly different from that of pseudobrookite. This slag also contained an M 2 O 3 phase, which was practically pure Ti 2 O 3. X-ray diffraction and microprobe analysis of slag E.6 showed the occurrence of both the M 3 O 5 phase (pseudobrookite), with N Fe 0.1 and the Ti 3 O 5 phase (anosovite), with N Fe 0.05. The microprobe values give the average of values obtained by a small number (2 3) of point analyses, and may not be fully representative for the phase as a whole. Furthermore, the pieces examined may not be fully representative for the entire slag due to segregation during freezing. For experiment E.4 backscatter pictures for the bottom and top parts are given in Figure 3. The bottom part appears to be mainly the M 3 O 5 phase with only small amounts of TiO 2-x, whereas the top part shows a typical eutectic structure with lamellas of M 3 O 5 and TiO 2-x phases. The composition of the TiO 2-x phase shows that this was practically free of iron (N Fe = 0.03) and had an N O value of 1.8 corresponding to the M 5 O 9 composition. Microprobe X- ray pictures from the bottom part of E.4 showed that the M 3 O 5 phase was not homogenous, the centre of the grains being low in iron and high in titanium, whereas the opposite was the case for the fringes of the solid solution grains. Evidently the composition of the solid solution had shifted towards a higher FeTi 2 O 5 concentration during solidification. Figure 3b. Back-scatter microprobe pictures for slag E.4 top. Light grey: M 3 O 5. Dark grey: TiO 2-x phase EQUILIBRIUM BETWEEN TITANIA SLAGS AND METALLIC IRON 37

The X-ray diffraction diagrams and the large amounts of acid insolubles for slags E.4 E.6, which were cooled slowly, showed that these contained significant amounts of rutile or TiO 2-x 4. In contrast, industrial slags contain very little rutile. The small amounts of titanium and oxygen found in the metallic iron droplets may possibly be caused by contamination from adjacent oxide phases. Melting point determination The temperature during melting and solidification in the induction furnace could not be measured. Instead, small pieces of the slag, about 3 mm cross-section, were heated in a sessile drop furnace in argon, and the fusion of the slag was observed optically. The furnace temperature was measured with a two-colour optical pyrometer, calibrated against the melting points of pure iron and copper. The slag pieces rested on a substrate, which, for the first measurements, was either graphite or vitreous carbon, the latter being known to be extremely inert. As long as the slag was solid and had only point contact with the substrate, reaction with the substrate was not expected, and the initial melting temperature, the solidus temperature, could be observed. As soon as a liquid phase appeared, this reacted strongly with the substrate, regardless whether it was graphite or vitreous carbon. Microprobe analysis of the sample after the melting point experiments showed that practically all iron oxide had been reduced to metal. In contact with the graphite or vitreous carbon an oxycarbide phase with about equal amounts of TiO and TiC had been formed. The remaining part of the slag had been reduced to practically pure Ti 3 O 5 + Ti 2 O 3 and had a melting temperature of about 1710 1740 C for all slags. The lower value may represent the eutecticum between Ti 3 O 5 and Ti 2 O 3. One melting point experiment was done with a molybdenum substrate on a slag from experiment E.4 Top. Initial melting occurred at about 1675 C, and complete melting at 1718 C. Microprobe analysis of the slag after the experiment showed that practically all the iron and manganese in the slag had diffused into the molybdenum substrate forming a solid solution. The slag had been enriched in TiO 2 and had a composition with about N O = 1.74 and N Fe = 0.06. Another melting point measurement on slag E.4T on a TiO 2 substrate in a shallow graphite bowl gave initial melting at 1620 C and complete melting at 1720 C. The first of these probably corresponds to the eutective groove between M 3 O 5 and the TiO 2-x phase, the second to the liquidus surface. Microprobe analysis showed that the slag had reacted with the TiO 2 substrate and the graphite bowl to give an intimate mixture of titanium oxide, carbon and some metallic iron with no well-defined phases. For slag E.5 T on a TiO 2 substrate a sudden and almost complete melting occurred at 1540 C. This is believed to correspond closely to the ternary peritecticum between M 2 O 3, M 3 O 5 and TiO 2-x. Further heating was stopped, and the solidified slag was examined by microprobe analysis, which showed the occurrence of a TiO 2-x phase with N O = 1.86 corresponding to Ti 7 O 13 and an M 3 O 5 phase with N O = 1.67 and N Fe = 0.14. Slag E.6 T on a TiO 2 substrate showed initial melting at 1675 C and complete melting at 1725 C. Microprobe analysis after the melting point determination showed an almost iron free slag with N O = 1.60. Evidently reduction by the graphite bowl had taken place. Slag E.6 B collapsed suddenly at 1710 C without there being any noticeable solidus melting. This is believed to represent the eutecticum between Ti 2 O 3 and Ti 3 O 5. Phase diagrams The present results were combined with those of Eriksson et al. 8 to estimate the ternary Fe-Ti-O phase diagram and its liquidus surface, which is shown in Figure 4. As all experiments, with the exception of experiment E.3, showed the occurrence of either rutile or a TiO 2-x phase, it may be concluded that this phase is formed during solidification. The small and variable amounts of iron droplets found in the slag are believed to have been in suspension and have not been part of the oxide melt at its liquidus (syntectic) temperature. This is supported by the fact that its concentration increases from the top to the bottom of the ingot, and is larger in the rapidly than in the slowly cooled slag. The top sections for the experiments E.2, E.4 and E.6 most likely represent the composition of the eutectic groove between the M 3 O 5 and the rutile or TiO 2-x phase. According to Eriksson et al. 8 Ti 3 O 5 melts congruently at 1718 C and its eutecticum with TiO 2-x occurs at 1665 C and N O = 1.73, and with Ti 2 O 3 at about 1680 C and N O = 1.62. As mentioned above the latter eutecticum is more likely to be around 1710 C. Furthermore, FeTi 2 O 5 melts peritectically at 1455 C, and FeTiO 3 melts peritectically at 1377 C. As mentioned above the top part of slag E.5 is believed to correspond closely to the ternary peritecticum between M 2 O 3, M 3 O 5 and rutile with a melting temperature of about 1540 C. Also shown in Figure 4 is the 1500 C isotherm given by Pesl and Eric 9. As seen, this shows a much wider melting region than found in the present study. Figure 5 shows a suggested melting point diagram for the Fe-TiO 2 composition line. It should be pointed out that this is not a true binary diagram as the composition of the M 3 O 5 phase changes from a somewhat higher Ti 3 O 5 content at the syntectic temperature of about 1710 C to a somewhat higher FeTi 2 O 5 content at the eutectic temperature of about 1620 C. At the same time the N O value increases from maybe 1.63 to 1.67. According to this diagram a slag melted at, say 1750 C, will on cooling to the syntectic temperature first eject a small amount of iron droplets. At the syntectic temperature a M 3 O 5 phase with a Ti 3 O 5 surplus and a small oxygen FeTi 2O 5 1455 C 0.4 N Fe 0.3 0.2 0.1 TiO 2 0.0 1857 C 2.0 0.5 1377 C 1.5 FeTiO 3 0.0 0.2 0.4 0.6 0.8 1.0 Ti 2O 3 1842 C Figure 4. Proposed ternary melting diagram for slags without impurities 1.9 1.8 N O 1665 C 1.7 Ti 3O 5 1718 C 1710 C 1.6 38 HEAVY MINERALS 2007

Figure 5. Proposed pseudobinary section Fe-TiO 2 as a function of temperature deficiency will precipitate. On further cooling an M 3 O 5 phase with decreasing Ti 3 O 5 content and increasing oxygen content will precipitate, and the slag will solidify at the eutectic temperature of about 1620 C. Under equilibrium conditions the M 3 O 5 phase will decompose below about 1325 C to give metallic iron and rutile or some Magnéli phase 8. In practice this decomposition is very slow and the phase combination M 3 O 5 + TiO 2-x is retained at room temperature. Effects of impurities Three slags were prepared with a Ti 2 O 3 /FeO molar ratio of unity and three with a Ti 2 O 3 /FeO molar ratio of three and with addition of 5% of respectively SiO 2, MgO and CaO, and with enough iron to ensure metal saturation. The melting time was about 25 minutes, and the slags were cooled rapidly in the crucible. The analytical results for representative samples of the slags are shown in Table II. It will be noticed that the slags S.1 and S.6 (with SiO 2 ) contained the highest content of Ti 2 O 3 and slag S.3 (with CaO) contained the lowest content of FeO. Slag S.4 (with CaO) contained the highest amounts of insolubles. X- ray diffraction and microprobe analyses (Table III) showed for all slags mainly the M 3 O 5 phase and traces of metallic iron. Slags S.1 and S.6 showed in addition a silicate glass phase, and S.3 and S.4 showed perovskite, CaTiO 3. Slag S.1 and S.3 showed traces of rutile, while slags S.2 and S.3T showed traces of an M 2 O 3 phase (ilmenite). Slag S.3 contained also an oxide phase TiO 2-x, which will be discussed later. It is evident that the three impurity oxides behaved in three different ways during solidification of the slags. Thus in slag S.1 silicon was present exclusively in the silicate phase which, on the average, contained in atomic per cent: 28.5 Si, 2.9 Ti and 0.6 Fe. In weight per cent this means that the silicate inclusions contained 86% SiO 2, 11.6% TiO 2 and 2.4% FeO. As the slag as a whole contained 4.3% SiO 2 this means that also 0.6% TiO 2 and 0.1% FeO were present in the silicate inclusions. These were subtracted from the total analyses given in Table III, and the N Fe and N O values for the oxide melt were calculated to 0.19 and 1.68 respectively, and are shown in Figure 6. Similar correction was made for slag S.6. For slags S.2 and S.5 microprobe analyses showed that all magnesium in the slag substituted for iron in the M 3 O 5 and M 2 O 3 phases. Therefore, for these slags the parameters N Fe and N O give the ratios (n Fe +n Mg )/(n Ti +n Fe +n Mg ) and n O / (n Ti +n Fe +n Mg ) respectively. It is also evident that magnesium, contrary to manganese, is preferentially enriched in the M 3 O 5 rather than in the M 2 O 3 phase. For slag S.3 and S.4 calcium was present mainly in the perovskite phase, which on the average contained in atomic per cent: 21.2 Ti, 0.2 Fe, 19.7 Ca and 58.9 O, i.e. close to Figure 6. Composition of slags with added impurities and of industrial slags from Tinfos Titan & Iron K/S. For slag S.1 and S.6 the SiO 2 content formed a separate silicate glass phase. For slags S.2 and S.5 MgO is counted as FeO. For slag S.3 and S.4 the CaO content is assumed present as CaTiO 3. For TTI slags the procedure by Pistorius 2 is applied Table II Composition of slags with impurities in weight%, and solidus temperature Slag no. TiO 2 tot FeO tot SiO 2 MgO CaO Ti 2 O 3 as TiO 2 Fe met Insoluble Solidus temp (C) S.1 81.6 18.4 4.3 - - 26.1 0.9 0.9 1491 S.2 81.6 16.3-5.1-17.1 1.0 0.1 1547 S.3 79.7 14.0 - - 6.0 21.8 0.2 0.9 1382 S.4 88.4 6.8 - - 5.7 31.9 0.2 7.0 1420 S.5 93.1 5.5-5.6-34.3 0.8 0.1 1735 S.6 91.1 8.9 5.1 - - 43.6 0.8 0.4 1575 EQUILIBRIUM BETWEEN TITANIA SLAGS AND METALLIC IRON 39

Table III Microprobe analyses in atomic%, normalized to 100% for slags with impurities Slag No Phases Ti Fe Si Mg Ca O N Fe N O S.1 T* M 3 O 5 30.8 5.1 64.0 0.14 1.78 Silicate 2.0 0.5 29.5 68.1 Fe (tr) 1.8 96.9 1.2 S.1 B M 3 O 5 32.3 4.4 63.2 0.12 1.72 Rutile 31.4 2.1 66.1 0.06 1.97 Silicate 3.8 0.7 27.5 68.6 Fe (tr) 2.6 96.3 1.1 S.2 T M 3 O 5 29.3 4.1 4.2 62.5 0.22 1.66 M 2 O 3 21.8 15.6 1.0 61.5 0.43 1.60 Fe (tr) 2.6 96.3 1.1 S.2 B M 3 O 5 29.8 3.1 4.5 62.6 0.20 1.67 M 2 O 3 21.2 16.3 1.8 60.7 0.46 1.55 Fe (tr) 2.9 96.0 1.0 S.3 T M 3 O 5 29.9 9.5 0.36 62.7 0.24 1.60 TiO 2-x 29.4 3.7 3.11 63.6 0.11 1.92 Rutile 32.8 1.2 0.10 65.9 0.03 1.94 Perovskite 20.5 0.6 19.70 59.3 M 2 O 3 20.7 17.5 0.50 61.4 0.46 1.61 S.3 B M 3 O 5 31.0 5.4 0.10 63.6 0.15 1.75 TiO 2-x 29.2 3.7 3.00 64.1 0.11 1.95 Rutile 32.7 0.9 0.15 66.3 0.03 1.97 S.4 T M 3 O 5-x 36.3 2.0 0.03 61.7 0.05 1.61 TiO 2-x 34.1 0.9 64.6 0.03 1.85 Perovskite 21.8 0.4 19.40 58.4 S.4 B M 3 O 5-x 36.1 2.5 0.12 61.2 0.06 1.58 TiO 2-x 34.3 0.8 0.18 64.0 0.02 1.78 Perovskite 22.0 0.2 20.00 57.7 S.5 T M 3 O 5-x 32.3 1.3 4.0 62.2 0.14 1.64 Fe (tr) 0.7 98.3 0.0 1.0 S.5 B M 3 O 5-x 33.5 1.6 3.8 61.2 0.14 1.62 Fe (tr) 3.2 95.5 0.1 1.2 S.6 T M 3 O 5-x 37.0 2.6 0.1 60.4 0.07 1.53 Silicate 2.7 0.4 33.1 63.9 Fe (tr) 3.4 94.7 0.3 1.7 S.6 B M 3 O 5-x 37.5 1.9 0.1 60.5 0.05 1.53 Silicate 3.2 0.3 32.8 63.8 Fe (tr) 4.3 92.8 0.7 2.3 *The XRD diagram for slag S.1 showed exclusively the M 3 O 5 pattern whereas microprobe analysis gave N O 1.75 corresponding to an M 4 O 7 composition. The reason for this is not clear. For slag S.2 N F e and N O are calculated by including MgO, on a molar basis, in the FeO value. For slags S.4 S.6 XRD showed mainly the M 3 O 5 phase whereas microprobe analyses gave N O values between 1.53 and 1.64 i.e. less than the stoichiometric composition (N O = 1.67). It is conceivable that the M 3 O 5 phase may exist with an oxygen deficiency, and may be denoted M 3 O 5-x. the stoichiometric CaTiO 3 composition. This phase was found mainly in the top part of the slag ingot, which shows that it was formed during the final solidification. Furthermore, these slags contained a phase TiO 2-x with N O 1.90, with significantly less iron and more calcium than in the adjacent M 3 O 5 phase. This indicates that some Magnéli phase may take some CaO into solid solution with correspondingly less FeO. For simplicity, for these slags the N Fe and N O values in Figure 6 were calculated by assuming all CaO tied up as CaTiO 3. Nevertheless, these slags showed, even after subtraction of CaTiO 3, higher N O values than the others. Evidently CaO has stabilized Ti 4+ in the slag. The melting temperature of slag S.1 (with SiO 2 ) was studied on a single crystal silicon carbide substrate. Initial melting, the solidus temperature, was observed at about 1490ºC. As soon as liquid phase was established this reacted rapidly with the silicon carbide substrate under vigorous gas evolution to give a metallic melt and an oxide which did not melt below 1720ºC. Subsequent microprobe analysis showed that the metallic phase contained, in atomic per cent, around 50 iron, about 35 silicon, 1 4 titanium and 15 17 carbon. The oxide phase was almost pure rutile. The melting temperature of slag S.2 (with MgO) was studied on a TiO 2 substrate in a shallow graphite cup. Initial melting occurred at about 1550ºC and the slag was completely melted at 1685ºC. Microprobe analysis showed no traces of a metallic phase. The oxide phases were rutile and M 3 O 5 with low Fe contents. 40 HEAVY MINERALS 2007

Slag S.3 (with CaO) on a TiO 2 substrate showed initial melting at 1382ºC and was completely melted at about 1525ºC. Microprobe analysis showed that the metallic phase contained, in atomic per cent, 3.9 Ti, 2.1 O and 93.2 Fe. Furthermore, the slag contained a M 3 O 5 phase with about 6.5 at% iron, and a rutile phase which contained ~1 at% iron. For slag S.4 (with CaO) the melting temperature was studied on a TiB 2 substrate. Initial melting occurred at about 1420ºC and complete melting above 1700ºC. Slag S.5 (with MgO) was also studied on a TiB 2 substrate, and showed initial melting at about 1735ºC. Complete melting was not reached, even at 1800ºC. Thus all impurities tend to lower the eutectic temperature between the M 3 O 5 and rutile phases shown in Figures 4 and 5. The largest freezing point depression is found for slags with CaO addition, the lowest with addition of MgO. This means that the impurities tend to stabilize the liquid oxide phase, and also extend its range of stability to lower oxygen contents. This may be the reason why industrial slags show mainly the M 3 O 5 phase with only traces of rutile (acid insolubles) Figure 5. Addition of MgO and CaO tends to stabilize the Ti 4+ ion with corresponding less Ti 3+. Slags from Tinfos Titan & Iron K/S Three slags from Tinfos Titan & Iron K/S were examined. These were taken during tapping of the furnace by dipping a cold steel bar into the slag flow. The slag solidified and cooled rapidly, and no noticeably oxidation was expected. The composition of the slags is given in Table IV. The N Fe and N O values were calculated by the procedure given by Pistorius 2, by counting MnO and MgO, on a molar basis, as FeO, and V 2 O 5, Cr 2 O 3 and part of Al 2 O 3 as Ti 2 O 3. This gave for the three slags N Fe from 0.190 to 0.231 and N O values from 1.656 to 1.665. Within the limits of analytical accuracy the N O values all agree with the stoichiometric composition M 3 O 5. These values are plotted in Figure 6. All slags showed almost exclusively the M 3 O 5 phase with only traces of rutile and metallic iron. Thus the procedure suggested by Pistorius 2 is well suited to predict the phase combination for industrial slags. Summary The equilibrium between titania slags and liquid iron was studied in an induction furnace with a multiple segmented water-cooled copper crucible. The slags were either pure Fe-Ti-O melts or melts with addition of about 5% of either SiO 2, MgO or CaO, and the Ti 2 O 3 /FeO ratio of the slags was varied by addition of either metallic Ti or Fe 2 O 3. The composition of the slag in equilibrium with liquid iron was found to follow an arched curve in the FeTiO 3 -TiO 2 -Ti 2 O 3 ternary diagram, and with its apex slightly above the composition of the FeTi 2 O 5 -Ti 3 O 5 solid solution line (the M 3 O 5 phase). The melting temperature in the Fe-Ti-O system was studied, and showed a eutectic groove between the M 3 O 5 phase and rutile or Magnéli phases (Ti n O 2n-x ). Addition of about 5% of respectively SiO 2, MgO or CaO showed that these lowered the eutectic temperature and shifted the composition of the slag in equilibrium with iron closer to the M 3 O 5 line. SiO 2 formed a separate silicate glass phase with small amounts of TiO 2 and FeO. MgO went into solid solution in the M 3 O 5 phase where it substituted for FeO. CaO solidified mainly as perovskite, CaTiO 3, whereas small amounts went into solid solution in a Magnéli phase. Table IV Slags from Tinfos Titan & Iron K/S TTI 1 TTI 2 TTI 3 TiO 2 tot 81.7 78.6 79.7 FeO 6.9 11.2 9.1 Ti 2 O 3 as TiO 2 27.1 20.8 23.9 MgO 5.8 5.8 5.9 MnO 0.58 0.56 0.56 Al 2 O 3 1.7 1.7 1.6 V 2 O 5 0.35 0.35 0.35 Cr 2 O 3 0.11 0.13 0.13 Others 5.14 5.19 5.16 Fe met 0.23 0.33 0.46 N Fe 0.190 0.231 0.212 N O 1.665 1.656 1.660 XRD phases M 3 O 5 M 3 O 5 M 3 O 5 Rutile (tr) Rutile (tr) Rutile (tr) Fe (tr) Fe (tr) Fe (tr) Acknowledgement The authors are grateful to Sean Gaal for assistance with preparing the slags and melting point determinations References 1. SIGURDSON, H. and MOORE, C.H. Petrology of High Titanium Slags, Trans AIME, vol. 185, 1949. pp. 914 919. 2. PISTORIUS, P.C. The relationship between FeO and Ti 2 O 3 in ilmenite smelter slags, Scand. J. Met. vol. 31, 1949. pp. 120 125. 3. TRANELL, G. SINTEF Materials Technology. Trondheim. Norway. Private communication. 4. ERIKSEN, J.M., ROBLES, E.C., and ROSENQVIST, T. Equilibrium between liquid Fe-Ti-O slags and metallic iron, Steel Research Int., 2007, In print. 5. SWINDEN, D.J. and JONES, D.G. Arc-furnace smelting of Western Australian beach sand ilmenite, Trans. Instn. Min. Met. 1978: p. C87: pp. C83 87. 6. PISTORIUS, P.C. and COETZEE, C. Physicochemical Aspects of Titanium Slag production and Solidification, Met. and Mater. Trans. vol. 34 B, 2003. pp. 581 588. 7. ROSENQVIST, T. Ilmenite Smelting, Trans. Techn. Univ. Kosice, vol. 2, 1992. pp. 40 46. 8. ERIKSSON, G., PELTON, A.D., WOERMANN, E., and ENDER, A. Measurement and Thermodynamic Evaluation of Phase Equilibria in the Fe-Ti-O System, Ber. Bunsenges. Phys. Chem., vol. 100, 1996. pp. 1839 1849. EQUILIBRIUM BETWEEN TITANIA SLAGS AND METALLIC IRON 41

9. PESL, J. and ERIC, R.H. High-Temperature Phase relations and Thermodynamics in the Iron-Titanium- Oxygen System, Met. and Mater. Trans. B, vol. 30B, 1999. pp. 695 705. 10. GREY, I.E., REID, A.F., and JONES, D.G. Reaction sequences in the reduction of ilmenite: part 4: interpretation in terms of the Fe-Ti-O and Fe-Ti- Mn-O phase diagram, Trans. Inst. Min. Met. 1974. pp. C83, pp. 105 111. See also Grey, I.E., Li, C., and Reid, A.F. Phase equilibrium in the system MnO- TiO 2 - Ti 2 O 3 at 1473 ºK Journ. Solid State Chem. vol. 17, 1976. pp. 343 352. 11. HONG, S.W., MIN, B.T., SONG, J.H., and KIM, H.D. Application of cold crucible for melting of UO 2 /ZrO 2 mixture, Materials Science and Engineering A357, 2003. pp. 297 303. 42 HEAVY MINERALS 2007