CO-PRODUCTION OF STEEL, TITANIUM AND HIGH GRADE OXTOE

This patent provides a method for directly smelting titaniferous materials such as ilmenite mineral concentrates to liquid titanium metal and or pigment-grade titanium dioxide by a process with an energy consumption believed to be less than one third of that of the currently best available commercial technology and potentially superior to other new technologies currently being considered as viable alternatives to the Kroll pyrometallurgical process route.

As pointed out in a recent paper (R. J. Fruehan, Y. Li and G. Carkin, Met &Mat Trans, 35B, August 2004, 617- 623) during the past 50 years there have been numerous attempts to produce aluminium by carbothermic reduction, eliminating the traditional Hall cell. None of this effort has yet led to a successful commercial process, but renewed efforts are apparently emerging, driven principally by a decrease in energy consumption and overall cost. The same factors drive attempts to lift titanium out of being considered almost an "exotic metal" and beyond the reach of most commercial applications in which the metal would clearly have exceptionally good metallurgical qualities.

The present inventor tried unsuccessfully to promote the theoretical attractions of primary aluminiizm via carbothermic reduction and vacuum distillation in the Research Department of Imperial Smelting (RTZ) in the UK (Unpublished Research N. A. Warner, Met. Research Statement No. 448, August 1970). This work was focused on direct smelting of bauxite to aluminium metal, eliminating both the Hall cell and the traditional Bayer process for upgrading bauxite. In those days, typical aluminium produced in the Hall cell contained as much as 800 ppm of iron and 600 ppm silicon. These days, the industry has moved to more high purity applications, so direct smelting from bauxite right through to primary alumimum of acceptable purity has apparently been ruled out by expert opinion.

In the case of direct smelting of titanium minerals, the issues concerning high purity are not so clear-cut as they are for present day alumimum. Despite this, the titanium pigment industry, for example, has to date been wedded to the necessity of others producing various feedstocks from ilmenite and other titaniferous materials to satisfy their preferred chloride route to high purity titanium dioxide, followed then by finishing steps to yield titanium pigment. Apparently, nobody in the industry has considered that it may be feasible to achieve purity levels in excess of 99.90% TiO ₂ by in-line refining continuously smelted ilmenite concentrates and thereby eliminate entirely the current chlorine-based technology, except perhaps as a means for dealing with those concentrates with unacceptably high thorium plus uranium contents and associated levels of radioactivity.

In taking up the challenge, the inventor was aware from the outset that for the carbothermic reduction of bauxite there was no stable aluminium oxide with a melting point below 1800 ⁰ C as there is in the Ti-oxygen system with TiO melting at 1750 ⁰ C but at the same time titanium carbide does not melt until 3067 ⁰ C. Furthermore, the existence of a complete range of stable solid solutions all possessing a simple cubic structure as there is with titanium oxycarbides, opens up exciting possibilities in terms of protective linings for ultra-high temperature reactors operating well above current temperature levels, say in excess of 2000 ⁰ C and upwards towards the generally accepted mechanical stability limit of structural graphite of around 2200 ⁰ C.

Titanium has valuable mechanical and corrosion resistant properties but according to a recent study performed for the US Department of Energy and Oak Ridge National Laboratory (Summary of Emerging Titanium Cost Reduction Technologies, January 2004), although titanium is the fourth most abundant structural material in the earth's crust after aluminium, iron and magnesium, the total US titanium production including scrap recycle was only around 48,000 tonnes in 1997 based on available statistics for that year.

A recent study published in the Journal of the Electrochemical Society (IC Dring et al., 152 (3), E104-113, 2005) claims that the quest for replacement of the current Kroll process has intensified and the cathodic deoxygenation of titanium dioxide (TiO ₂ ) originally proposed by Fray and co-workers at Cambridge University (WO 99/64638) is stimulating a great deal of interest.

The direct electrochemical reduction of TiO ₂ without the need to form titanium tetrachloride (TiCl ₄ ) involves a primary reduction reaction in which titanium oxide is deoxygenated by an electrochemical charge transfer of the type:

TiO _x + ne ^" <r> n0 ^"3 + Ti0 _x-n (1)

A preferred electrolyte is a molten calcium halide, which exhibits high oxygen anion solubility.

WO 03/046258 A3 describes a method for electrowinning titanium metal or alloy from titanium oxide containing compounds in the liquid state, which is claimed to have significant advantages over other emerging technologies. It involves direct electrowinning of titanium from molten titanium mixed oxide compounds in which the deoxygenation step is performed using either a consumable carbon anode, or an inert dimensionally stable anode or a gas diffusion anode. The preferred electrolyte is molten calcium fluoride. During the chemical electrochemical reduction, droplets of liquid titanium metal are produced at the oxide/electrolyte interface and sink by gravity, settling to the bottom of the electrochemical reactor, forming after coalescence, a pool of liquid titanium metal or alloy. The liquid metal is continuously siphoned or tapped under an inert atmosphere and cast into dense titanium metal ingots. The teachings of WO 03/046258 A3 form the basis of the present invention.

The initial production of metallic droplets of impurities such as metallic iron and other transition metals more noble than titanium e.g. Mn, Cr, V etc. advocated in WO 03/046258 A3, detracts seriously from the process in that cross-contamination would inevitably occur between the initial impure metal and the product titanium. Furthermore, once of the iron and other impurity metals have been removed, it is necessary to siphon off or tap these before titanium metal production commences. This means the process is not truly continuous, a shortfall also self-evident in the batch charging of molten titanium slag or other molten materials into the electrochemical reactor.

A truly continuous process is not yet available, which is capable of accepting the electrically conductive titaniferous charge materials such as ilmenite mineral concentrates or titaniferous magnetite at one end of the spectrum right through to synthetic rutile or solid upgraded titanium slag at the other end of the spectrum. Their transformation in-line to high purity liquid titanium π oxide as the preferred continuous feed for titanium metal

production is the essence of the present invention. At the same time, it is clearly desirable to co-produce refined liquid steel in a state ready for continuous refining in advance of continuous casting.

Reduction of iron oxide in ilmenite using carbon (anthracite or char) in an electric furnace results in the formation of liquid iron and slag. With present technology, the need to maintain liquid slag with a low viscosity at temperatures in the range of 1700 ⁰ C to 1800 ⁰ C places an upper limit of the TiO ₂ at 88%. Thus all the iron is by design not reduced to metal, as some must remain in the ferrous state to act as a flux. This is advantageous in so far as the molten iron produced is largely free of trace elements, most impurities remaining in the slag and consequently a high purity pig iron is produced. Stanaway (Mining Engineering, December 1994, 1367-1370) states that before slagging in the electric furnace, the ilmenite feed can be roasted to remove sulphur, which he states is important for pig iron quality. In the inventor's opinion the emphasis should be redirected and focused on the quality of the oxidic melt not on the pig iron quality, because continuous refining of steel is a relatively straightforward procedure and is one which is likely to be adopted in the future by the industry.

This invention is an attempt to lead industry onwards into a new era and leave behind the conventional wisdom, which for so long, has thwarted development of a virile titanium metal industry. The hitherto upper temperature limit for ilmenite smelting furnaces, for example, will have to be abandoned and more advanced engineering designs incorporated in future plants.

To substantially further advance the technology for carbothermic reduction of ilmem ^' te, it is necessary to enter into the region of ultra-high temperature pyrometallurgy. Rather than considering current processing options dominated by freeze lining of electric furnaces, which after many years of commercial development, must surely now be considered to be operating close to their upper limits, the present invention starts with an entirely new perspective.

Carbothermic reduction of ilmenite and the associated formation of titanium oxycarbide are not new concepts and as long ago as 1992, B. S. Terry and P. Grieveson at Imperial College QSIJ International, Vol. 33 (1993), No. 1, pp 166-176) studied the reduction of ilmenite to iron and titanium carbide or titanium oxycarbide with the declared ultimate aim of achieving separation of the titanium and its subsequent conversion to pigment-grade titanium dioxide. The emphasis in their work was on solid-state reactions for titanium in association with liquid iron. More recently workers at the University of New South Wales (G. Zhang and O. Ostrovski, TMS Congress 2001, p 89-103) have proposed that reduction of titanium oxides to oxycarbide in the solid state with following chlorination might be an efficient alternative technology for processing of titanium minerals.

Also, workers in South Africa have made major contributions towards understanding the fundamental aspects of ilmenite smelting in electric furnaces and in particular have highlighted the fact that for large pyrometallurgical furnaces "it is difficult to reason intuitively about their dynamic behaviour" (J. H. Zietsman and P. C. Pistorius, Minerals Engineering, Vol. 19, Issue 3, March 2006, pp 262-279). They draw attention to the importance of the freeze lining behaviour in its role to contain liquid high titania slag, which aggressively attack refractories. They point out that the freeze lining of slag formed by water-cooling interacts both thermally and chemically with the slag bath. To make real progress in the future, this chemical interaction must be prevented so that ideally there is

no net chemical reaction between the melt and the lining or in other words the two phases must be in equilibrium with each other. The present invention achieves this objective and thus paves the way for ultra-high temperature processing for ilmenite treatment in the future.

The principle promulgated in the present invention for smelting right up to the engineering limit of structural graphite, say 2,200°C in terms of stability and mechanical strength, is that ultra-high temperature operations must be conducted relatively close to thermodynamic equilibrium between the phases in contact with each other. Normally in pyrometallurgy these phases are molten slag, molten metal, the solid in contact therewith with and the associated equilibrium gas phase at the prescribed constant operating temperature. Of these, the interaction of the slag and solid phases is crucial but the liquid metal/solid hearth contact must also be considered. For the particular case of continuous steelmaking in US 2005/0269752 Al, this means that the smelting system must be contained in residual unnielted steel at the hot face close to the composition for the solidus temperature and the melt itself must be close to the liquidus temperature for the particular ferrous melt involved, firstly in the ironmaking loop and secondly that in the steelmaking loop or loops. The combination of electrical conductive heating and melt circulation (N.A. Warner: Trans. Inst. Min. Metall. C, 2003, 112, C141-C154) is the key to making this a practical reality. By these means, a sequence of in-line processing steps can be performed in a series of melt circulation loops. It is accepted that relatively large areas are required, if close to equilibrium conditions are to apply, and this implies that use of what have been termed previously by the inventor, as "swimming pool reactors", but this is a small price to pay for the very considerable rewards potentially available. Fortunately, relatively large departures from equilibrium can be tolerated between the uppermost liquid phase and the bulk gas phase, if as is generally the case at ultra high temperatures, gaseous diffusion is rate controlling and thus equilibrium is automatically established at the interlace. This ameliorates the surface area requirements and allows acceptable diffusional driving forces to be maintained.

It is of course quite obvious that swimming pool reactors can only be entertained, if the processing technology is truly fully continuous, with provision for the withdrawal upwards of certain plant items, when processing is interrupted and the melt freezes over. Even right at the liquidus temperature, freezing over is not instantaneous but rather a two-phase solid/liquid region is entered as heat is dissipated. Accordingly, this is entirely feasible but it does imply electrical conductive heating and the residual unmelted solid shell approach, to be able to return reliably to steady-state operation. This steady state operation then has to be maintained indefinitely without cyclic variation.

In order to construct a fully engineered reactor with prefrabricated solid linings, use is made optionally of individual bricks, slabs, blocks or preformed shapes comprised of materials of the appropriate composition, which will be commensurate with being in a state of thermodynamic equilibrium with the slag at the operating temperature. Clearly, it is necessary to envisage that availability of acceptably priced raw materials to supply on- site requirements will not be an undue problem.

In the particular case of titanium oxycarbide unmelted solid shell linings, there is the very considerable advantage that TiO and TiC form a complete range of solid solutions, so provision of oxycarbide linings is simplified very considerably. For synthesis of group IVB metals, oxycarbide carboreduction reactions of the type:

TiO ₂ + (a + b) C -» (TiC _a O ₂ . _b ) + bCO (2)

have been reported by Mazzoni and Conconi (Materials Research, 5, (4), Oct/Dec 2002), covering the range TiCo.i6θo.84 to TiC0.73O0.27, and involving heating the finely inter-dispersed reactants at a temperature of 125O ⁰ C to 165O ⁰ C with a reaction time of 120 minutes.

Anatase, rutile and synthetic rutile are the preferred raw materials for the synthesis of the oxycarbide linings. The other reactant is preferably carbon black, again a material commercially available at reasonable prices and is used, for example, in formulating the mix for automobile tyres. Under certain circumstances, fabrication of the linings in-situ is preferable, particularly those associated with complex arrangements such as underflow and overflow weirs and linings or whole component parts associated with siphons and gas-lift pumping devices.

Alternatively, a material mix corresponding to the required solidus composition can be melted in specialist facilities away from site and quenched rapidly to capture the solidus phase composition. The cooled material is then reground and then pressed and sintered in a controlled atmosphere with a CO/CO ₂ ratio as indicated on the HSC4 program into refractory blocks from which the melt circulation reactor can be fabricated on site.

For large swimming pool sized reactors, monolithic linings composed of various grades of titanium oxycarbide, is the preferred approach. Away from site, material mix corresponding to the required solidus composition can be melted and quenched rapidly to capture the required single-phase equilibrium condition. This material is re- ground and stored under inert gas protection for transportation to the site. On site construction of steel formwork is undertaken with electrical conductive heating in mind so that an applied low voltage generates uniform temperature throughout the steel formwork, which itself is encased in supporting refractory enclosures. Provision is made for ramming the material mix into the formwork and then the whole assembled lining and associated structure is conductively heated in an argon atmosphere to enable the mixture to consolidate by sintering to form a monolithic lining. Depending on particular requirements and assessment by ceramic experts, the titanium oxycarbide may in some cases be the inner lining of a composite lining with more conventional refractory mixtures backing on to outer steel/alloy shells. On cooling, the inner steel casing is removed, leaving the outer steel case in place and the whole volume enclosed is filled with material of the liquidus composition. Again under a protective argon atmosphere, the whole mass is conductively heated with perhaps graphite electrodes also installed above to melt the inner material, whilst heat is extracted from the external walls and base. Melt at initially the liquidus temperature infiltrates into the sintered side walls and base and further material mixture is added over a period of time to raise the oxide melt level to the operating level. At this stage iron containing carbon, etc., as per the HSC4 formulation is added to build up a bottom layer of liquid iron melt and in so doing some of the current flow changes from the oxide melt to that of the iron hearth melt. This procedure is repeated as required along the length of the swimming pool reactor in a number of campaigns to complete the task.

In the present invention, operation of the continuous process line close to thermodynamic equilibrium is mandatory. For a particular sub-process ideally there must not be any change in overall chemical composition of the principal components of each phase and the temperature throughout must be maintained essentially constant.

This is approached ideally by what is termed a fully back-mixed reactor system. The preferred mechanism for achieving this fully back-mixed condition is melt circulation within a closed loop employing a carrier melt, comprised of the sub-process product both in temperature and composition, circulated at a rate many times that of the melt overflow or other means of passage to the next sub-process. For example, PCT/GB2006/000302 relates to a melt circulation ratio in excess of 1500/1 in these terms, and of course electrical conductive heating can be applied in striving for temperature uniformity. The energy implications of such high melt circulation ratios are minimal, provided the head to which the melt is being circulated is itself relatively low.

For ilmenite processing, the first of the sub-processes is the formation and recovery of a molten iron alloy for subsequent continuous processing to liquid steel ready for continuous casting. Thus a gas phase and two liquid phases are involved. All of these must be essentially at equilibrium with each other throughout the associated melt circulation loop. If the three phases are in a state of thermodynamic equilibrium, then provided a phase diagram is available, the composition of a suitable solid phase can be identified. Ceramists and others skilled in the art can specify the so-called liquidus and solidus compositions for a particular operating temperature. If phase equilibrium data are not available, clearly experimentation may then be required. The reactor hearth, walls and contact areas of equipment such as lances, snorkels, overflow and underflow weirs immersed in the melt must all be prefabricated from material of the solidus composition.

The gas phase in each of the sub-processes is not back-mixed but is essentially in a state of what is termed "plug flow". Careful consideration needs to be given to this aspect and provision must be made for ready access and ultimate roof replacement if necessary. For example, in the highly reducing atmospheres characteristic of the sub- processes involved in the present invention, structural graphite can be used backed by graphitic thermal insulation, provided attention is focused on restricting gas-phase mass transfer, which at the very high temperatures involved can be expected to dominate the interfacial rate phenomena involved in the interaction of the furnace atmospheres and the roofs within each loop. In theory, transfer of gases between each of the melt circulation loops, even at extreme temperatures, does not pose a problem, if the solid lining is again at the identified solidus composition and temperature, because the offgas and its solid lining are by definition in a state of thermodynamic equilibrium,

The generic melt circulation technology advocated by the inventor on many prior occasions is characterised by circulation of a single liquid phase. In certain circumstances, where erosion of the hearth refractory has been perceived as a potential problem, as for example in complex sulphide smelting, based on circulation of copper sulphide (matte), the carrier melt is copper-saturated, so it is clearly advantageous to maintain a stationary pool of molten copper below the circulating copper matte.

In the present invention, because of the prime importance of operating close to thermodynamic equilibrium between solid and liquid phases within each of the sub-process loops, it follows that composition and temperature gradients within the bulk phases must also be eliminated. Thus if two liquid phases are involved in a particular sub-process, then each of these phases is preferably independently circulated under turbulent flow conditions to promote good mixing. Alternatively, perhaps the lower phase (metal) is circulated and the upper phase (slag) subjected to intense top blowing and vice versa, if suitable tuyeres can be installed to provide good mixing throughout the lower phase (metal) with the upper liquid phase (slag) force circulated. To illustrate the

aforementioned general approach, further attention throughout the description will be directed at what is believed to be a totally new approach to pyrometallurgy, viz. forced circulation of both slag and metal phases.

TMs invention does not prescribe a single preferred reductant for carbothermic processing of ilmenite, as individual locations throughout the world have different priorities and opportunities for accessing their own preferred reductant. Natural gas, calcined anthracite coal, petroleum coke, carbon black or wood charcoal are among reductants chemically feasible, but hydrogen and carbon monoxide or indeed manufactured gas with low methane content are very much less efficient thermodynamically, unless used in conjunction with a carbon source at unit activity. A prime requirement is very low sulphur content because of the detrimental effects of sulphur on attaining high quality products in the present case. Natural gas of the purity specification supplied to power stations is ideal in terms of sulphur content. For other reductants some prior desulphurisation may be necessary. In terms of sustainability, wood charcoal is obviously a renewable resource and is attractive for avoiding carbon emission penalties.

Fossil feel decarbonisation for mitigating global warming and heralding in a hydrogen economy is a possible scenario for the future. Thermal decomposition of methane is central to the approach. The by-product carbon black could well become readily available for carbothermic reduction under these circumstances. Carbon black used alone to form a floating carbon layer in a melt circulation loop, or pneumatically conveyed carbon black entrained in carbon monoxide, for example, for top jetting onto or into a titanium oxycarbide melt, or carbon black co-charged with ilmenite concentrates in a pelletised condition, all offer potentially viable approaches for carbothermic processing of ilmenite.

Effective management of the oxycarbide protective lining is crucial to the success of the new technology. The inventive strategy is based on the ability to transfer carbon reliably at very high efficiency to the melt in order to sustain equilibria throughout without precipitation of oxycarbide or other solid phases in the bulk of the melt. Such precipitation would lead to uncontrolled deposition and accretion build-up as well as deterioration in the physical properties of the single phase melt, especially viscosity, which is important to avoid, to ensure that the various liquid phase transport processes are not impeded. Although the thermochemical reduction requirements can be provided by solid carbon addition, carburisation by blowing natural gas onto the surface of the melt is instantly controllable and provides locally advantageous mass transfer conditions conducive to effective assimilation of dissolved carbide. This is all achieved without carry-over of solid carbon particles into the gas phase above the melt surface. The essential prerequisites are turbulent flow in the bulk melt and carefully designed top blow nozzles so that the carbon utilisation efficiency is very close to the predicted 100% theoretical value.

Natural gas top blowing is also the foundation for satisfying other important issues, which dominate the potential viability of the new continuous processing technology. In particular, because of the extremely high temperatures involved, health and safety issues demand that the process plant is fully under control at all times. In the event of an adventitious departure from normal or safe conditions, automatic control strategies must be in place to take over the operation without manual intervention. Straightforward top blowing of the melt surface from above, coupled with controlled melt circulation rate simply by changing the inert gas flowrate to the gas-lift pump

arrangement and electrical conductive heating to supply the thermal demands, are fail-safe measures which can be exploited collectively to safely implement automatic control procedures.

Unlike its predecessors in steelmaMng, the melt circulation loops cannot readily be heated radiatively by post- combustion of the gaseous products of reduction. The emphasis now must switch to electrical conductive or direct resistance heating of the melts for supply of a major portion of the energy input and in this respect the operation is analogous to the zinc process proposed by the inventor for direct zinc smelting using a molten copper matte as the melt circulation medium. There are strong parallels between the properties of copper matte and titanium oxycarbide melts. Both are electronic conductors with relatively high thermal conductivities and because post combustion cannot be used to heat oxycarbide or matte melts directly, conductive heating and melt circulation are mandatory requirements for both. This effectively means that cheap electrical power is a prerequisite for the proposed new technology, as indeed is already the case for electric furnace smelting of titaniferous materials in Canada (hydroelectriciry) and South Africa (coal-based power). With reference to electrical conductive heating, it is noteworthy that there has been a recent change in sentiment towards nuclear power and it has been suggested that global warming and climate change may lead to resurgence in the nuclear power industry. According to an article published in Materials World (April 2005), some 17% of global electricity is generated by nuclear power. France, for example, has 78% of its total power requirements provided by nuclear energy.

Referring back to the titanium oxycarbide furnace linings, the ductile to brittle fracture transition for such materials is very favourable and they would appear to be able to operate over a temperature range from above 2000 ⁰ C to say 800 ⁰ C. This facilitates maintenance of the unmelted shell approach rather than attempting to use so-called "skull formation" with water-cooled hearths. According to the Handbook of Refractory Carbides and Nitrides

(H. O. Pierson: "Handbook of Refractory Carbides and Nitrides", NJ. Westwood: Noyes Publications, 1996, p. 66), the transition metal carbides have the ability to deform plastically above a given temperature, referred to as the ductile-to-brittle transition temperature. Below that temperature titanium carbide fails in a brittle manner, while above it, it shows ductile behaviour and undergoes plastic deformation. For TiC this is in the region of 800 ⁰ C and because of the cubic structure of titanium oxycarbide over the whole range of solid-state stability, high temperature linings of titanium oxycarbide can reasonably be expected to behave in a similar fashion. This is extremely helpful in terms of accommodation of thermal stresses resulting from thermal expansion without fracturing or forming cracks in the lining, which could lead to problems with melt containment and also very desirable in maintaining the electrical and thermal conductivity integrity of the titanium oxycarbide solid lining.

At the present time, the market for titania pigments far outstrips the demand for titanium metal, so the invention is focused not only on a possible major future increase in demand for titanium metal, but also on production of pigment-grade titanium dioxide directly by carbothermic smelting of ilmenite and related titanium minerals without the involvement of chlorine-based technology, which is central to state-of-the-art white pigment manufacture. The disadvantages of the chloride route to white pigments include the need for high-grade feedstocks and the potential hazards associated with large amounts of chlorine and titanium tetrachloride.

An embodiment of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which: -

Fig. 1 is an overview of the plant for continuous smelting of ilmenite concentrates employing three melt circulation loops in series to feed an electrochemical deoxygenalion reactor to produce titanium metal with the oxidic melt overflow then undergoing various processing options to produce either or both pigment-grade titanium dioxide product, or, if exceptionally high purity specifications so demand, a process intermediate of titanium tetrachloride by direct chlorination of the oxycarbide melt at high temperature.

Fig. 2 repeats the essentials of Fig. 1 except that electrochemical deoxygenation to titanium metal is carried out at the end of the in-line processing, rather than being upstream of the downstream processing options, so that the only melt undergoing electrochemical deoxygenation is that destined to be converted to titanium metal.

Fig. 3 is schematic general arrangement of the first of the melt circulation loops for carbothermic reduction of ilmenite using natural gas as the reductant, showing ilmenite distribution onto the top of the oxidic melt, natural gas distribution to headers for top blowing the melt surface, the means for quenching the off-gas with quench gas recycled by a turbocompressor after heat recovery in a waste heat boiler and hot gas filtration with a candle filter, circulation of the melt in a closed loop and its continual progression to downstream in-line processing stations and continuous production of liquid steel in advance of refining prior to continuous casting.

Fig. 4 is schematic half sectional elevation of one arm of a melt circulation loop containing a single oxidic melt liquid phase showing the "cavity-wall" type of construction comprised of an inner hot face lining of solidus composition titanium oxycarbide, a free space containing support skids to allow unimpeded thermal expansion and contraction of the arm, boiler tubes for steam raising, superheating or closed-loop steam reheating service as appropriate on safety grounds, so positioned that they receive direct thermal radiation from the cooler face of the oxycarbide lining, and an outer backing of conventional refractory and insulating materials, all encased externally in a gas-tight steel shell.

Fig. 5 is a schematic sectional elevation of one end of the reduction arm (typically 60-90 m in overall length), shown in cross-section across the width in Fig. 4, when operating at the full design temperature.

Fig. 6 is a schematic cross-sectional elevation of the two side-by-side arms of the first of the melt circulation loops, the top blown reactor (TBRl), which employs forced circulation of both the oxidic melt and denser molten metal, each based on gas injection from above with overhead lances or other gas sparger devices. Fig. 6 shows the suspended flat arch roof construction and one row of an array of top blowing lances passing across the hot face lining of titanium oxycarbide, positioned very close to the top surface of the oxidic melt.

Fig. 7 is a schematic sectional elevation view of the first of the melt circulation loops, the top blown reactor (TBRl), which employs forced circulation of both the oxidic melt and denser molten metal, each based on gas injection from above with overhead lances or other gas sparger devices.

Fig. 8 is a comparison of three sectional elevations taken across the melt circulation loop constituting TBRl, moving from left to right beyond the ceiling/melt close proximity top- blow area of Fig.7:

(a) shows side-by-side arms of the loop separated longitudinally by a central partition of the cavity wall type of construction with titanium oxycarbide of solidus composition on both sides of the free space with support skids, steam tubes, etc., and cross-hatching all removed for clarity purposes

(b) shows as above, a section through the gas-lift pumping arrangements for the top oxidic melt circulation

(c) shows as above, the system for gas-lift pumping of the bottom molten metal phase melt circulation.

Fig. 9 is a schematic sectional plan (AA in Fig. 8) of the molten oxidic melt cross-over between the two principal side-by-side arms of the upper oxidic melt circulation loop of TBRl, comprising, in the direction of melt cross- flow, the overflow weir from the main bath (submerged not visible), a melt calming section followed by an underflow weir all of titanium oxycarbide of the solidus composition with cavity wall type of construction and then the active melt sump for gas-lifting the oxidic melt across and overflow weir of the same solidus composition as the underflow and principal linings back into the other principal arm and means for top lancing or sparging as required for gas admission to promote two-phase flow upstream of the overflow weir.

Fig. 10 is a schematic sectional plan (BB in Fig. 8) of the molten metal cross-over between the two principal side- by-side arms of the lower molten metal circulation loop of TBRl, comprising in the direction of melt cross-flow, the weir from the main bath (above section not shown), a melt calming section followed by an underflow weir of titanium oxycarbide of the solidus composition with cavity wall type of construction and the active melt sump for gas-lifting the molten metal across the overflow weir of the same solidus composition as the underflow and principal linings back into the other principal arm and means for top lancing or sparging as required for gas admission to promote two-phase flow upstream of the overflow weir.

Fig. 11 is schematic cross-sectional elevation of the two side-by-side arms of the melt circulation loop that constitutes in-line continuous vacuum refining (CVR) of the liquid titanium oxycarbide produced in TBR2, showing both arms of the loop contained within a single vacuum vessel.

Fig. 12 shows the key features of chlorine-based technology seen to be included as one of the "downstream processing options" referred to in Figs. 1 and 2.

Referring now to Fig. 1, in explanation of what is termed "downstream processing", it is necessary to know that in the chloride process, titanium dioxide feedstock is reacted with chlorine in the presence of coke in a fluidised bed at around 900 ^u C to produce impure titanium tetrachloride. In the present invention, as one of the process options implied in Fig. 1, the already refined titanium monoxide and can be reacted with chlorine to produce titanium tetrachloride as an intermediate product, which with current technology can then be purified so that all objectionable impurities are reduced down to extremely low levels. In the invention the existing carbon in the oxycarbide melt contributes to the overall carbon requirement for carbochlorination. With titanium oxide in the reduced state, the consumption of carbon for carbochlorination is one half of that required for carbochlorination of

the equivalent Ti content TϊO ₂ . Accordingly, the preferred embodiment, if chlorination is deemed necessary to reduce residual impurity levels even further than can be accomplished in-line by vacuum refining, is to complement the existing carbon in the melt after electrochemical deoxygenation, with additional carbon added to the melt either as elemental carbon in one of the forms, already referred to, for carbothermic reduction to permit carbochlorination to proceed by reaction (3).

TiO + C + 2Cl _2fg) = TiCl^fe) í COfe) (3)

For the corresponding reaction employing natural gas, the chlorine may by injected into the melt to avoid hydrogen chloride formation in advance of carbochlorination, but it is recognised that excess CI ₂ will react to form HCl unless steps are taken to keep the gaseous reactants apart. Accordingly, top blowing with CI ₂ an already carburised melt by natural gas as top blowing in an entirely separate compartment is the preferred approach and it is proposed that another melt circulation loop to conduct this carbochlorination operation is fully justified. On one arm (the first arm) for reaction with natural gas and the other for top blowing with chlorine is the preferred embodiment. The off-gas from the first arm is a further stream of hydrogen enriched "fuel gas" to be added to this exported commodity or used in-plant for power generation.

An overwhelming attraction of top blowing with natural gas to provide the additional carbon not already in the melt for carbochlorination is that the exothermicity is reduced to manageable levels by incorporating reaction (4) into the sequence of chemical processing. Per mole CH ₁ , the reaction exothermicity is reduced to a mere 17 kcal, if both reactant gases are initially at 25°C and the TiO and the three gaseous reaction products all at 1900 ⁰ C.

TiO + CHLK ₈ ) + 2Cl ₂ ( _S ) = Ti Cl ₄ ^ + CO® + 2H _2(g ) (4)

The present invention takes as received mineral concentrates and transforms them continuously to titanium oxide with impurity levels so low that the pigment requirement for minimal optical absorption at visible wave lengths is met for adequate whiteness and opacity. For titanium dioxide this in turn requires high chemical purity and in particular transition metal impurities must be virtually eliminated. Small amounts of impurities such as iron, manganese, chromium or vanadium darken the pure titanium dioxide crystal. Therefore, the priority in the present invention is to reduce as far as is thermodynamically possible the concentration levels of these four deleterious metals. At the same time, it was considered worthwhile to aim for minimum concentration levels of all other contaminants.

At first sight, the possible need to use chlorine to meet high purity specifications may appear to be an admission of defeat. In-line pyrorefining can reduce all the critical impurities down to single digit ppm levels. If higher purity is mandatory for pigments, commercial purity titanium metal is well within reach of in-line refining. However, it should be stressed that the chloride route to pigments, even with an up-graded titania slag as the feedstock, still produces significant quantities of waste metal chlorides with less than 10% processed for sale. Currently none of the chlorine in the waste is recycled. The present invention, if tetrachloride is needed for ultimate purity, is considered to improve the waste chloride issue by at least one order of magnitude.

Referring now to Fig. 2, the plant for continuous smelting of ilmenite concentrates employs at least one melt circulation loop at around atmospheric pressure, or perhaps somewhat higher pressure to secure benefits, feeding continuously into a continuous vacuum refining (CYR) melt circulation loop discharging downstream into various processing options in advance of the residual melt finally undergoing electrochemical deoxygenation. This is in recognition of the fact that there is some degree of miscibility between the oxidic melt and the molten electrolyte as well as acknowledging other possible reactions that may contaminate the oxidic melt, such that it is no longer of the very high purity produced by CVR The most readily identifiable benefits of operating above atmospheric pressure are to facilitate gas- UfI pump action and also to reduce fluoride or chloride gaseous emissions associated with the molten electrolytes involved in titanium metal production.

Referring now to Fig. 3, preheated ilmenite concentrates 1 are distributed onto the surface of the circulating oxidic melt 2 in the top blown reactor TBRl on either or both of the reactor arms 3. Natural gas 4, preheated no hotter than 350 ⁰ C to avoid carbon decomposition, is piped to an array of top blowing nozzles or jets 5. The very hot off- gases 6 (raw fuel gas) are quenched by special means, which have been incorporated so that the copious fume generated on quenching does not form accretions on solid surfaces in the immediate vicinity.

The preferred method of quenching is to exit the gases through nozzles at relatively high velocity so that the quench gas 7 is entrained radially into the hot gas jets to effect cooling without intervention of solid surfaces. The gases are then passed onwards to a waste heat boiler 8. In the present invention, the quench gas itself is recycled hydrogen enriched "fuel gas" at say 400 - 500 ⁰ C. After solids removal from the gases 9 leaving the waste heat boiler 8 in a candle filter 10, if hot gas cleanup is dictated for on-site power generation or, alternatively at a lower temperature if traditional bag filters are used. In the latter event, the hydrogen enriched "fuel gas" is exported to a nearby power station. In either case, a turbocompressor 11 is installed both for recycling purposes as well as for onwards transmission of the low sulphur fuel gas 12 directly to either gas turbine combustors for combined cycle power generation or pipeline distribution to customers.

The solids 13 collected by the candle filter 10 are reverted to the process feed 14 so that no fine solid waste disposal problems arise and valuable iron units are recovered. For fully continuous operation the equivalent amount of liquid steel 15a actually produced within TBRl has to be withdrawn from the melt circulation loop as unrefined product liquid steel 15a. Similarly, the equivalent amount of oxidic melt 16a formed within TBRl has to be withdrawn continuously from melt circulation.

In consequence of the relatively small flow rates involved with both 15a and 16a in comparison with the overall melt circulation 2, special consideration must be given to ensuring the flow of both 15a and 16a proceeds reliably without malfunction, which could jeopardise the whole continuous processing operation. The inventor is very aware of the crucial importance of this aspect and has outlined procedures for attaining this objective using electrical conductive heating in a Patent Application entitled "Molten Siphon with Internal and External Heater" (WO 2004/074524 Al) and also referred to in a published paper entitled "Conductive heating and melt circulation in pyrometallurgy (N. A. Warner, Trans. Inst. Mm. Metctll. C, December 2003, Vol. 112, C141-154). Because of the presence of dissolved titanium and aluminium, typically in the region of 1% and 500 ppm respectively, there

will be very little oxygen in the liquid steel, so the propensity for disruptive sub-surface carbon monoxide evolution will be virtually zero. Accordingly, the siphons can utilise a full vacuum if so required.

Referring now to Fig. 4, for rectangular furnaces 17, projected to be typically about 60-90 metres in overall length, the expansion of the unmelted oxycarbide linings 18 from room temperature to say 2000°C is estimated to be around 2 metres. Accordingly, very special measures need to be taken in the design of such furnaces to accommodate differential expansion internally, whilst keeping the outside surfaces of the furnaces moderately cool. Freedom of the hearth to expand or contract without excessive friction is crucial to the success of the proposed continuous smelting technology. Conventional skid mounting 19 on heat resistant alloy shells 20 encasing the cooler faces of the oxycarbide linings for mechanical integrity and structural stability, or perhaps more sophisticated "bogey" rail tracking may be necessary for this purpose so that the rather long hearths involved can expand or contract freely. In this connection it must be borne in mind that unscheduled shutdowns have to be accommodated and the prospect of the hearths cooling to room temperature must be addressed at the design stage. Also sufficient clearances must be provided inside the furnace interiors to permit free expansion and contraction to take place differentially with respect to outer steel encasement 21 or associated pressure/vacuum vessels.

Typically, the width of the furnaces involved could be in excess of 8 metres, so suspended arch construction is necessary and for the physical configuration demanded in carbothermic reduction using natural gas. The refractory roof or "arch" 22 has to be flat so as to maintain a small clearance 23, probably in the region of 10 cm between the liquid melt 16 and the flat refractory roof 22 immediately above it. The furnace hearth cannot support this roof because it expands very considerably itself on being heated to operating temperature.

Provided the foregoing precautions are taken to ameliorate the deleterious effects of hearth expansion, the remaining issue concerns the 1 metre or so expansion of the flat refractory roof. In the present invention, the full length of the flat refractory roof and its associated structural steel work 24 is supported on steel pontoons floating on liquid metal or fusible alloy contained in launders or troughs 28 on each side of the hearth (un-melted oxycarbide 18) containing the melt 16 and extending the full length of the furnace. By pumping liquid metal in and out of these launders, the pontoons 26 can be made to float and thus during heating up from room temperature to say 2000 ⁰ C at the hot face of the lining, the structure is free to expand both longitudinally and laterally across the width of the hearth. When operating temperature is reached, liquid metal 27 can be partially removed from the troughs 28 so that the pontoon supported structures no longer float but rather can bear down onto refractory fibreboard 29 in a controlled fashion to form a gas-tight seal, assisted by the positioning of load cell devices at appropriate locations. If the plant is to be shutdown from operating mode, the pontoons 26 can be floated again by pumping liquid metal 27 back into the troughs 28 so that the roof structure 25 and its associated refractory flat arch 22 can return eventually to the cold position.

Continuing reference to Fig. 4, the removable lid 30 covers the whole extent of the furnace and may be a single unit or multiplicity of units to achieve the same effect and affords the means for ready access. In this diagram the unsectioned area on the right illustrates the external hearth channel enclosure 17. A skid-mounted system 19 permitting thermal expansion or contraction of the oxycarbide lining 18 and a row of steam raising boiler tubes 31

to stabilise the munelted oxycarbide 18 at a prescribed steady state thickness are shown schematically in this diagram.

It is recommended that a protective gas atmosphere is maintained at locations not already referred to as containing natural gas or products of combustion, in order to prevent carburisation or hydrogen embrittlement of steel or alloy components exposed to relatively high temperatures, such as the steam boiler tubes 31 and Hie metal 20 sheathing the cooler faces of the oxycarbide lining 18.

Referring now to Fig. 5, it can be seen that at the design operating temperature level, differential expansion between the highest temperature oxycarbide un-melted shell lining 18 of the furnace and the relatively cool external enclosure 21 almost fully consumes the expansion space 18a on one side, provided for skid-mounted free movement from its initial position when cold.

Also apparent in Fig. 5 is the flat suspended arch 22 with its steel support joist girders 25, which are permitted to move independently of both the external enclosure 21 and the un-melted shell of oxycarbide 18 constituting the furnace hearth, side and end walls in order to accommodate differential expansion during warm-up. At full operating temperature once steady state is reached, the fusible alloy launder/pontoon system, discussed in Fig. 4, in relation to the flat refractory arch 22 and its supporting steelwork, would be activated so that sufficient weight is bearing down to effect compression of fibreboard sealing arrangements and also enough force applied laterally on the fibreboard seal at 29 to compress it adequately for gas sealing purposes. It may also be observed that this particular drawing depicts operation somewhat above atmospheric pressure. The higher-pressure gas region is associated with the gas space 23 and the corresponding gas off-take 45, whilst near to atmospheric pressure conditions prevail above the melt in the gas at 46.

Referring now to Fig. 6, external furnace enclosures have been removed and the cavity-wall type of construction associated with the oxycarbide linings 18 is simplified by deleting cross-hatching, skid mounts, boiler tubes, etc. The refractory suspended flat arch 22 and its associated steel joist girders 25 and supporting steelwork 24 is shown to span across both arms of the melt circulation loop. Depending on the width of the melt baths concerned, an alternative is to have two independent support systems each with its own pair of steel pontoons 26 floating on liquid metal or fusible alloy 17 contained in launders or troughs 28 to permit individual expansion or contraction, if only one of the arms of the melt circulation loop is required to be shut down. Fig.6 relates specifically to TBRl in which both the oxidic melt 16 and the lower steel melt 15 are circulated independently of each other. Also shown in this diagram is the narrow gap or clearance 23 between the oxidic melt 16 and the hot face of solidus composition titanium oxycarbide roof elements 22. Individual lances or jets 32 that form an extensive top blown array for admitting natural gas are protected with at least one and preferably two or more cylindrical concentric radiation shields fabricated from the solidus composition oxycarbide and other ceramic materials as the temperature decreases. Each of the top blow lances 32 is steam cooled on its outer surface at least over the relatively short length exposed to the highest temperature within the titanium oxycarbide facing of the suspended arch or ceiling immediately above the melt. By these means the temperature of the natural gas is maintained below 350 ⁰ C to prevent methane decomposition and carbon deposition. One option is that such cooling

constitutes the initial reheating of a closed loop steam system in association with an advanced steam turbine for power generation.

Now referring to Fig. 7, this relates specifically to TBRl, a melt circulation loop with both oxidic and metal phase independent melt circulation. The rows of lances 33 and 34 or other appropriate gas sparging devices are beyond the plane of section in the sectional elevation shown. Lance row 33 admits lift-gas into the two-phase flow region that establishes forced circulation of the top layer oxidic melt 16. Lance row 34 extends into the lower metal phase region to provide the driving force for forced circulation of the bottom layer of molten steel 15. In both cases, 33 and 34 show a row containing three lances and typically the requirement is between two and four lances for smaller width arms, say below 5 m in width, and correspondingly larger numbers if wider arms are employed for increased smelting throughputs. The lift gas employed is preferably the hydrogen-enriched fuel gas 12, already compressed by the turbocompressor 11 in preparation for export or in-plant usage to fire the gas turbine combustors of a combined cycle power plant. The solidus composition titanium oxycarbide lining 18 is shown crosshatched in this diagram. Because of the absence of high gas velocities and thus inherently lower gas phase mass transfer coefficients involved, the roof area 35 is of baked carbon or graphite construction, backed with carbon-based insulation, which may if deemed necessary be faced with oxycarbide, but calculations reveal that baked carbon would have an acceptable life span in this region. Elsewhere, in the natural gas top-blown region, extending virtually the whole length of the furnace, oxycarbide facing of the ceiling 22 is mandatory. A single row of top blowing lances 32 is shown as discharging natural gas directly into the narrow gas space 23 above the surface of the oxidic melt 16. The associated gas jets emerging from the hot face of the ceiling 22 are effectively in what is termed the potential core jet region so that hot gas entrainment is minimal and very little opportunity for methane decomposition arises in advance of the jets impinging on the melt surface under precisely controlled non- splash conditions. The structural steelwork and steel joist girder 25 system for supporting the refractory suspended flat arch extends upwards towards the removable lid 30 covering the length of the furnace to ensure gas tightness.

Referring now to Fig. 8, again this diagram is specific to TBRl and in certain cases optionally to TBR2. It relates to the non-top blown area at the right hand end of the furnace shown in Fig. 7, where they are no high gas velocities and the roof 35 is of baked carbon or graphite construction. The solidus composition titanium oxycarbide linings 18 of the cavity-wall type of construction are not crosshatched but are shown schematically, free of expansion skids, boiler tubes, etc. The oxycarbide linings 18 of the cavity walls illustrated envelop all solid faces in contact with oxidic melt 16 and liquid steel 15, including the hearths, sides, end walls, central partitions 36 as well as overflow weirs 37 and underflow weirs 38. There is no liquid phase cross flow in (a) but in (b) and (c) the principal liquid flow in the central regions is cross flow between the principal arms of the melt circulation loop. This cross flow is in the direction from left to right as viewed in the diagram and is oxidic melt 16 in (b) and liquid steel 15 in (c). Hydrogen-enriched fuel gas 12 is admitted via gas sparging devices or alternatively as illustrated by lances 33 and 34 from above into the gas/liquid two-phase flow regions 39 and 40 to create the driving force for melt circulation of each of the individual liquid phases.

The two-phase regions 39 and 40 are equivalent to a less dense arm of a U-type configuration resembling a manometer. Gas injection into one side of the U, if a homogeneous gas/liquid mixed phase is formed, lowers the

density on one side of the U so that the two-phase gas/liquid rises and eventually overflows, if there is inadequate height available to reach a new equilibrium position. Now, if the static leg of the U is connected to a supply of the liquid phase and the whole system forms a closed loop, then liquid circulation will continue indefinitely as long as the two-phase flow region is maintained stable by gas injection. The static region 41 of the hypothetical U configuration can be viewed as a calming region in advance of the active two-phase flow region on the other side of the U. Region 41 is especially useful in the case of the lower liquid metal phase melt circulation loop as it provides access for withdrawing continuously the product unrefined liquid steel 15 from TBRl using the aforementioned vacuum-assisted electrically conductive heating siphon arrangement.

As the example given later in the description illustrates, considerable process benefits are derived in terms of removing certain metallic impurities by increasing per unit mass of ilmenite contained in the feed, the absolute mass of liquid steel employed as extractant for these impurity elements. In this connection it is worth noting that reverting refined liquid steel back into the melt circulation loop does not present a problem, because of the heavier density of steel ensures that it will report to the circulating lower phase, irrespective of where it is added from the top. However, if direct straightforward access to the product liquid steel 15 were not available at 41, special alternative arrangements would have needed to be provided.

Referring now to Fig. 9, the oxidic melt 16 is force circulated around a closed loop encased throughout with a solidus composition titanium oxycarbide lining 18 using cavity-wall type of construction with ancillaries, such as steam raising boiler tubes removed from view for clarity. The lower liquid steel 15 is mainly obscured but this is also force circulated independently of the oxide melt circulation emphasised in this diagram. There is no particular merit in attempting countercurrent contacting of the oxidic melt 16 and the liquid steel 15 as they are both fully back-mixed. On the contrary, to preserve interfacial stability, it is preferable for both liquid phases to be moving co-currently, so that individual bulk phase mixing is promoted at higher turbulence levels by this means without incurring a penalty in terms of Helmholtz interfacial instability.

To induce forced circulation of the oxidic melt 16, a row of three upwards removable vertical lances 33 admits the preferred lift-gas into the active sump 39 for gas lifting the oxidic melt 16 across the overflow weir 37 and thereafter the melt flows by gravity along the length of the furnace split into two by the central partition 36 and around the closed loop path highlighted by the large block arrows and then back to the underflow weir 38. To the right of the aforementioned flow path are various components related to the analogous arrangement for forced circulation of the liquid steel, which will be described in Fig 10.

Now referring to Fig. 10, the unrefined liquid steel 15 constituting the bottom layer of the two liquid phases in TBRl is circulated around a closed loop encased throughout with solidus composition titanium oxycarbide lining 18 using cavity wall type of construction with ancillaries such as steam tubes removed from view for clarity. To induce forced circulation of the steel melt, a row three upwards removable vertical lances 34 admits the preferred strip gas into the active sump 40 for gas lifting the liquid steel 15 across the overflow weir 37 positioned so that it is above the level of the top surface of the oxidic melt 16, thus causing the liquid steel to shower through the oxidic melt layer as it sinks through to the circulating bottom layer of liquid steel 15. Clearly, the pumping head requirements for circulating the liquid steel 15 are somewhat greater than that required for circulating the oxidic

melt 16, so it is necessary to provide additional submergence in the liquid steel sump to facilitate this. A rough guide used by those skilled in "air-lift" technology is that the submergence should be at least twice the head to be pumped, so this is duly taken into account. Once the liquid steel 15 reaches the bottom it continues its circulation by gravity along the length of the furnace split into two by the central partition 36 around the closed loop path highlighted by the large block arrows and then back to the underflow weir 38 to report again into the active sump in advance of its assimilation into the two-phase flow region 40 created by the row of lances 38 or other appropriate means for lift-gas sparging commensurate with dispersing the lift-gas as uniformly as possible throughout the liquid steel 15 so that its density is very considerably reduced and efficient pump action established.

Referring now to Fig. 11, conditions may be established in the process design so there is optionally no liquid steel melt in either TBR2 or in the melt circulation loop associated with continuous vacuum refining (CVR) of the titanium oxycarbide melt fed continuously using the aforementioned siphon arrangements from TBR2 into the fore hearth or sump of the CVR melt circulation loop. In vacuum steel degassing, steam jet injector systems have been developed for exhausting large volumes of gases from degassing vessels typically down to 0.5 to 1 mbar pressure. In particular, oxygen top blown decarburisation in association with circulating flow steel vacuum degassing is now established practice in batch steel refining and multiple steam jet ejectors backed up with water ring pumps are state-of-the-art with, for example, 1800 kg/h dry air equivalent pumping capacity at 1 mbar pressure. This is the technology on which CVR is dependent. Furthermore, as in the comparable steel degassing case, it is considered worthwhile to accommodate a degree of top blowing with reducing gas of the titanium oxycarbide melt, to ensure that gas phase mass transfer does not limit reaction rates in forced circulation "swimming pool" reactors. These reactors resemble in many respects the corresponding near atmospheric pressure TBR2, except that the narrow gap between the melt surface and the overlying roof is no longer necessary and is replaced by a gas freeboard between the top of the melt and the roof in the region of 1 to 3 m so that gas phase pressure drop does not become an issue. The liquid oxycarbide enters and leaves the CVR melt circulation loop through barometric legs continuously and in this respect the invention resembles the commercial vacuum dezincing (VDZ) technology developed by the zinc industry for removing zinc from the circulating "condenser" lead of a zinc blast furnace. As already stated, the essentials of CVR in a pair of side-by-side furnace arms, electrically conductively heated with forced melt circulation of such magnitude, that the CVR reactor is fully back-mixed and near equilibrium conditions established for utilisation of the solidus composition titanium oxycarbide lining throughout the whole circuit, mirrors what has already been described in relation to TBRl and TBR2.

The CVR melt circulation loop across which the sectional elevation of Fig. 11 is taken, contains identical arms projected to be typically 50 metres or so in overall length, so that concerns already expressed about differential thermal expansion apply equally well to the CVR arrangement. Each of the melt circulation arms could optionally be made vacuum tight with crossovers at both ends to interconnect the arms to form a closed loop comprised of two principal side-by-side furnaces. Alternatively, as shown in Fig 11, the whole loop can be accommodated within a single pressure vessel 42. There is a parallel here to established practice in the electric power industry, where high pressure fluidised bed coal combustors are enclosed within spherical or cylindrical pressure vessels to eliminate concerns about pressure tightness of the various individual components that make up a fluidised bed

system. In the United States, for example, utilisation of such prefabricated pressure vessels is limited by the maximum diameter transportable on inland waterways, which is considered to be around 12 metres.

Thermal expansion and contraction of the titanium oxycarbide hearth lining 18 may be taken care of by skid mounts 19 bearing on heat resistant metal alloy sheathing 20 on the cooler feces of the oxycarbide lining in association with steam boiler tubes 31, etc. Alternatively, more sophisticated means maybe adopted. Differential expansion and contraction associated with the suspended refractory roof arches with hot faces around 2000 ⁰ C fabricated from solidus composition titanium oxycarbide elements 22 suspended from structural steelwork comprising steel joist girders 25, etc., are accommodated by the aforementioned system employing fusible alloy or liquid metal 27 inside launders or troughs 28 on each side of the structure spanning the overall width, inside which steel pontoons 26 extend the full length of the furnace. The pontoons 26 either fully support the structural steel and its associated suspended refractory arch, when sufficient depth of liquid metal 27 is contained within the troughs 28 during start-up or shutdown, or when liquid metal is drained or pumped out to pre-determined extent, a controlled pressure is exerted on the ceramic fibre board 29 to effect a gas tight seal, once differential movement has ceased. As before, access is provided by the removable lid 30 and elsewhere there is a fixed steel enclosure 21.

The gas freeboard 44 is of such height that the gas phase pressure drop is only a relatively small fraction of the total operating pressure, which for current steel vacuum degassing systems is typically around 0.5 to 1 mbar. There is extensive gas evolution in the CVR loop associated with chemical reactions involving titanium carbide as an oxide reductant with the thermodynamic activity of the titanium carbide typically being in the region of 0.1 to 0.2 in the fully back-mixed melt circulation loop, which by definition reflects the final composition of the product melt 43. This refined oxycarbide melt is removed preferably continuously from the vacuum chamber back to atmospheric pressure using the aforementioned electrical conductive heating principles not by a siphon, but this time by a barometric leg of the melt sheathed in its protective lining of solidus composition titanium oxycarbide into an atmospheric pressure sump or tundish, from which a siphon can then be used for either continuous or intermittent passage of the melt to the next in-line processing stage, the electrochemical deoxygenation reactor, as per the flow sheet presented in Fig. 1 and 2, to produce titanium metal.

An insight into the thermochemistry is provided in Table 1, which lists the standard free energy and enthalpy changes for the key chemical reactions all at 2000 ⁰ C. In general terms, the more negative is the free energy change δG° than the more the reaction is favoured in the forward direction (i.e. left to right in Table 1). Thus it is quite apparent that iron oxide is the most easily reduced to the metallic state but clearly there is little prospect for reduction of thorium oxide to metal. It should also be borne in mind that the values of δG° listed are for reactants and products in their standard states. Obviously a very minor component, such as uranium oxide, will be so diluted that its thermodynamic activity is extremely small and thus is unlikely to be reduced. The equations for magnesium and calcium also warrant comment. Both these metals are well above their normal boiling points at 2000 ⁰ C, but it is conceivable that at direct reduction followed by rapid absorption into the relatively large amount of molten iron first formed could possibly take Ca and Mg into solution. In actual fact, dilute solutions of magnesium in molten iron have very large positive deviations from ideality (activity coefficient about 90) so this is unlikely to occur with Mg.

Table 1

Table 1 also gives background to the refining of impurities out of the principal titanium oxide phase but it must be appreciated that besides oxide reduction to metal, formation of carbides must also be taken into account at the same time. These aspects are best dealt with using readily available computer software. In this context, the inventor has made extensive use of Outokumpu's HSC4 Program to elucidate favourable conditions and then used mathematical modelling to predict the ultimate performance.

For purposes of illustration only, the end results of computer evaluation of several preferred embodiments will now be summarised for carbothermic smelting of the same typical ilmenite concentrate of the following mass percentage composition:

TiO ₂ 50.4; FeO 34.1; Fe ₂ O ₃ 12.1; Al ₂ O ₃ 0.55; SiO ₂ 0.76; MgO 0.76; MnO 0.61;

P ₂ O ₅ COl; CaO 0.04; Nb ₂ O ₅ 0.11; V ₂ O ₅ 0.28; Cr ₂ O ₃ 0.09.

Example 1

Smelting Reduction loop at 1540 ⁰ C, based on carbon addition (1.5 kmol C/kmol TiO ₂ ) and molten iron circulation with solid raft of sintered oxidic material projected by factional drag forces out of the smelting reduction loop onto the surface of the first of two downstream titanium oxycarbide melt circulation loops with carbon addition

(0.97 kmol C/kmol TiO ₂ ) in the first at atmospheric pressure followed by Continuous Vacuum Refining (CVR) in the second. The first oxycarbide loop at 2080 ⁰ C and CVR conditions are 1920 ⁰ C at 1 mbar pressure.

Assuming equilibrium is established in all three loops, the overall recovery of titanium oxides is 98.9% and the equivalent TiO ₂ purity is 99.57%.

In ascending order, the impurity oxide levels in the final reduced titanium oxide product are computed as:

CrO ₂ , MnO, SiO ₂ < 1 ppm; FeO < 5 ppm; MgO 18 ppm; CaO 220 ppm;

Al ₂ O ₃ 860 ppm; NbO + NbO ₂ 1310 ppm; VO 1920 ppm.

Example 2

Single titanium oxycarbide melt circulation loop for Smelting Reduction Reactor (SRR) based on carbon addition (2.6 kmol C/kmol TiO ₂ ) at 2080 ⁰ C and at atmospheric pressure. Product of SRR overflown or siphoned to the forehearth or tundish at atmospheric pressure with barometric leg withdrawal continuously to a single titanium oxycarbide melt circulation loop (CVR) operating under a vacuum of 1 mbar at 194O ⁰ C. Continuous discharge from CVR via a barometric leg to a tundish at atmospheric pressure to feed either or both an electrochemical deoxygenation reactor for titanium metal production and/or in-line oxidation, if so required, for high-grade titanium dioxide product. In ascending order, impurity levels in final reduced titanium oxidic melt, assuming equilibrium established in both loops, are computed as follows:

CrO ₂ , MnO, SiO ₂ < 1 ppm; MgO 7 ppm; FeO 8 ppm; Al ₂ O ₃ 100 ppm; CaO 186 ppm; (NbO + NbO ₂ ) 337 ppm; VO 345 ppm Equivalent "TiO ₂ " Purity = 99.91%; Overall Recovery "TiO ₂ equivalent" = 92.41%

Example 1 indicates that if the initial melt circulation loop to which preheated ilmenite concentrates are added, is operated with molten iron as the carrier medium at say 1540 ⁰ C so that the unmelted steel shell region is operable, then the oxides of chromium, manganese and silicon can be reduced down to less than 1 ppm and iron down to around 5 ppm after continuous vacuum refining (CVR). On the other hand, aluminium, niobium and vanadium levels range from 860 to 1920 ppm. Of greatest concern, if white pigment is the desired ultimate product, is the relatively high level of vanadium, previously referred to as especially deleterious, which is around 0.19% in this example.

In Example 2 the initial processing loop is operated with oxidic melt circulation at the higher temperature of 2080 ⁰ C rather than 1540 ⁰ C and this reduces the vanadium oxide and niobium oxide down to levels around 1350 ppm. To make further reductions in the V and Nb oxides it is necessary to invoke a procedure involving reversion of molten iron after refining back to the SRR melt circulation loop as the preferred processing option. This extracts both V and Nb oxides from the molten oxide phase by reaction at the slag/metal interface and then diffusion into the molten iron phase, where V in particular has a large negative deviation from ideality and is thus retained at very low thermodynamic activity. The reaction between 3SIbO and TiC is somewhat more favourable but at high dilution Nb forms an almost ideal solution in molten iron, so the effectiveness of extraction from the slag into the molten iron is comparable for both oxides.

The aforementioned slag/metal interaction is an essential component of the present invention and steps to promote the mass transfer aspects have been especially incorporated to facilitate this. Also the thermodynamic aspects are enhanced very considerably at the ultra high temperatures implicit in the invention and are just not available to the same extent with current industrial practice for ilmenite smelting. The same considerations are also very advantageous in lowering the aluminium oxide content by the analogous mechanism. At the same time, it must be

appreciated that reversion of refined molten iron/steel back into the melt circulation loop also extracts additional titanium into the circulating metal phase with a corresponding decrease in ultimate Ti oxide or Ti metal product recovery. Just how critical reduction in vanadium oxide content really is in terms of enhanced pigment quality is probably only known to the pigment manufacturers themselves. However, to illustrate the approach being advocated in the invention, a possibly somewhat excessive amount of molten iron is reverted in Example 3, in which 10 lαnol Fe(μ/kmol TiO ₂ in feed is considered at the expense of marked decrease in ultimate Ti recovery to product (92.5% to 81.5%).

Example 3 is also used to reinforce another vitally important aspect. If high purity is the objective, it is essential to reduce sulphur levels associated with inputs down to an absolute minimum. This applies to the feed ilmenite concentrate and also of course to the reductant. As stated earlier, natural gas is available commercially with very low sulphur levels (single digit ppm range) and if high purity is the over-riding consideration, it is then difficult for coal-based and most other forms of commercially available carbon to compete in purity terms, unless given very special treatment at added expense. It is imperative that the actual level of sulphur in ilmenite be reduced prior to charging by preheating preferably already low sulphur ilmenite concentrates under controlled conditions to oxidise sulphide sulphur to sulphur dioxide in the region of 1200 ⁰ C. Preferably the sulphur must be reduced to around 0.005 mass % in order to sustain production of exceptionally high purity products.

In fundamental terms, the aforementioned sulphur problem is due to calcium sulphide in particular being extremely stable in a reducing environment, so the problem is mitigated if the calcium content of the ilmenite concentrate is itself minimal. In both Examples 1 and 2, the indicated levels of around 200 ppm "calcium oxide" equivalent are very largely due to the anticipated contamination of the reduced melt with dissolved calcium sulphide associated with a feed to the SRR containing 0.05% S. Example 3 considers that the sulphur is reduced down to below 0.005% sulphur in the thermal treatment before preheated concentrates enter the circuit.

Example 3

Case (a)

Ilmenite concentrates of the aforementioned typical composition and with a sulphur content of 0.05 wt% are heated in a fluid bed to 1200 ⁰ C to drive off sulphide sulphur down to 25 ppm prior to charging the preheated concentrates to TBRl. Some 10 kmol refined liquid steel per kmol TiO ₂ is reverted to TBRl in order to enhance the removal of vanadium, niobium and aluminium in particular and also to generally assist in further removal of other reducible oxides. TBRl is operated at 1 bar pressure and 1870 ⁰ C with 2.45 kmol CH ₄ per kmol TiO ₂ being top blown under non-splash conditions onto the oxidic melt surface. It is assumed that both the oxide melt and the liquid steel are fully back-mixed, induced by melt circulation using gas lifting with the aforementioned hydrogen enriched "fuel gas" and the linings throughout are of the appropriate solidus composition titanium oxycarbide.

A further 0.60 kmol CH ₄ per kmol TiO ₂ is top blown onto the surface of the melt in TBR2, which is lined with its own appropriate solidus composition titanium oxycarbide. There is no liquid steel phase in TBR2, which is at 1 bar pressure and a temperature of 2150 ⁰ C. The principal function of TBR2 is to carburise the melt in advance of continuous vacuum refining (CVR). In this example, CVR is conducted in the third melt circulation loop at 1

mbar pressure and 2060 ⁰ C. Again there is no liquid steel phase in this melt circulation loop and the reactor is lined throughout with its individual appropriate solidus composition titanium oxycarbide.

Assuming equilibrium is reached in each of the three back-mixed melt circulation loops, the recovery of titanium oxide principally as titanium oxycarbide and its impurity levels are as follows: - "TiO ₂ " equivalent recovery equals 81.5% and the equivalent TiO ₂ purity is 99.98% Impurity levels in ascending order are computed as:-

CrO ₂ , MnO, SiO ₂ , M ₈ O, Al ₂ O ₃ < 1 ppm; FeO 1.1 ppm; VO 48 ppm; (NbO + NbO ₂ ) 59 ppm; CaO 114 ppm.

****************

At first sight, it may appear that a "TiO ₂ " recovery of 81.5% in this example is an unacceptable price to pay in pursuit of enhancing the purity of the already value-added products of this invention. However, it must be pointed out that two processing steps each with recoveries of 90% by present and emerging technologies, producing synthetic rutile or upgraded slag as feedstock for chloride route pigment manufacture, would themselves equate to an overall TiO ₂ recovery of 81% and, of course, product purity levels of 99.98% are unheard of in such products.

Example 3 Case (b)

Virtually the same as Case (a) except that the sulphur content is reduced to 2 ppm before the preheated ilmenite concentrates are fed into the first melt circulation loop TBRl. Refined liquid steel at 10 kmol Fe/kmol TiO ₂ is retained the same as in Case (a). Top blowing is retained at 2.45 kmol CH/kmol TiO ₂ in TBRl and 0.6 kmol CHLi/kmol TiO ₂ in TBR2 and again conditions are chosen such that a liquid steel phase is only present in TBRl, where it is circulated independently of the oxidic melt TBRl operates at 1870 ⁰ C and 1 bar pressure; TBR2 is 2115 ⁰ C and 1 bar pressure with CVR at 200O ⁰ C and 1 mbar pressure, in this example.

Assuming equilibrium is reached in each of the three back-mixed melt circulation loops, the recovery of titanium oxide principally as titanium oxycarbide and its impurity levels are as follows: -

"TiO ₂ " equivalent recovery equals 81.7% and the equivalent TiO ₂ purity is 99.98%

Impurity levels in ascending order are computed as>

CrO ₂ , MnO, SiO ₂ < 1 ppm; M ₈ O 1.4 ppm; AI ₂ O ₃ 1.9 ppm; FeO 2.1 ppm; VO 25 ppm; (NbO + NbO ₂ ) 35 ppm;

CaO 82 ppm.

Case (c)

Parallels Case (b) except that the amount of liquid steel reverted to TBRl is reduced from 10 to 3 kmol Fe/kmol TiO ₂ . The sulphur level of the preheated concentrate remains at the exceedingly low figure of 2 ppm. The CH ₄ required is reduced from 2.45 to 2.15 kmol CEL _t /kmol TiO ₂ but the temperature of the first loop TBRl is retained at 1870 ⁰ C. TBR2 is now 2.40 at 1 bar pressure requiring 0.65 kmol/CELi/kmol TiO ₂ and CVR is undertaken at 2030 ⁰ C and 1 mbar pressure. It will be observed that various temperature levels are indicated so that the melt in each loop is very close to the liquidus temperature. Having fixed the TBRl temperature at 1870 ⁰ C, this effectively identifies the amount of CH ₄ addition required to maintain the appropriate composition at the liquidus temperature. The various CH ₄ and temperature requirements vary in accordance with the mandatory condition

that operation close to the liquidus temperature must be attained. Accordingly, the figures in Case (c) and Case (b) are not identical and, for example, the operating temperature for CVR is some 15 ⁰ C hotter than in Case (b).

Assuming equilibrium is reached in each of the three back-mixed melt circulation loops, the recovery of titanium oxide principally as titanium oxycarbide and its impurity levels are as follows:-

"TiO ₂ " equivalent recovery equals 93.1% and the equivalent TiO ₂ purity is 99.97% Impurity levels in ascending order are computed as:-

CrO ₂ , MnO, SiO ₂ , M ₈ O, FeO < 1 ppm; Al ₂ O ₃ 6.1 ppm; CaO 35 ppm; VO 142 ppm; (NbO + NbO ₂ ) 155 ppm.

****************

All the examples so far discussed could be adapted to titanium metal production, but there is uncertainly about the purity levels actually required for pigment manufacture. Data are available in the patent literature on this aspect, but one cannot be sure that the information is authorative or is currently the view of the major pigment manufacturers. For example, U.S. 6,548,039, in describing a hydrometallurgical route to pigment states that preferably the impurity content of colouring impurities (such as Fe, Cr, Ni, V ₇ etc.) to the amount of Ti in solution is such that the hydrolysis product contains no more that 20 ppm of these impurities. U.S.Patent 6,375,923, again on a hydrometallurgical route, reports a final titanium dioxide that contains only 6 ppm Fe and states that the invention described does not require an extra processing step to meet market requirements. More recently, U.S. Patent Application 0060051267 relating to purification of titanium tetrachloride, states that common metal chloride impurities in crude titanium tetrachloride include chlorides and complex chlorides of Al, Nb, Ta and V. It is asserted that "these metal chloride impurities are not susceptible to removal by distillation because of the proximity of their boiling points to that of titanium tetrachloride or their solubility in the titanium tetrachloride. They have a detrimental impact on downstream processes." Further quantitative information unfortunately is not given. Examples 4 and 5 are now presented to demonstrate that the aforementioned impurities can be reduced to single digit ppm levels by in-line high temperature refining during direct smelting of ilmenite concentrates and four nines (99.99%) TiO ₂ equivalent is attainable in the present invention.

Example 4

Ilmenite concentrates of the typical composition and with a sulphur content of 0.05 wt% are heated in a fluid bed to 120O ⁰ C to drive sulphur down to < 10 ppm prior to charging to TBRl. Natural gas-based smelting reduction with both oxidic melt and liquid steel melt independently circulated in TBRl and TBR2. 3 kmol refined liquid steel per kmol TiO ₂ is reverted to TBRl in order to enhance the removal of V, Nb and Al. TBRl operated at 1 bar pressure and 1870 ⁰ C with 2.15 kmol CH ₄ per kmol TiO ₂ top blown under non-splash conditions. Further 0.65 kmol CH ₄ per kmol TiO ₂ top blown onto the surface of the melt in TBR2. 2 kmol refined liquid steel per kmol TiO ₂ reverted to TBR2, which is at 1 bar pressure and 2100 ⁰ C. Principal function of TBR2 is to carburise the melt in advance of continuous vacuum refining (CVR) and to further reduce VO and (NbO + NbO ₂ ) by extraction into reverted liquid steel circulation. CVR conducted in the third melt circulation loop at 1 mbar pressure and 1990 ⁰ C. There is no liquid steel phase in this third melt circulation loop.

"TiO ₂ " equivalent recovery equals 84.1% and the equivalent (TiO ₂ + ZrO ₂ ) purity is 99.99%. Impurity levels in ascending order are computed as:

CrO _2, MnO, SiO ₂ , Al ₂ O ₃ < 1 ppm; M ₈ O 1.9 ppm; FeO 2.4 ppm; VO 5.4 ppm; (NbO + NbO ₂ ) 9.4 ppm; CaO 85 ppm. For pigment purposes, so-called "coloureds" in elemental form: Mn, Cr < 1 ppm; Fe 1.9 ppm; V 4.1 ppm.

Example 5

Carbon-based with two smelting redaction loops (SRRl and SRR2) with both oxidic melt and liquid steel melt circulation in each loop and with refined liquid steel reversion to both SRRl and SRR2, followed by CVR. As in (b) very low sulphur in concentrate feed to SRRl operating at 197O ⁰ C with 2.45 kmol C/kmol TiO ₂ and 3 lαnol refined liquid steel per kmol TiO ₂ reversion. Further 0.5 kmol C/kmol TiO ₂ added to SRR2 at 2160 ⁰ C with 2 kmol refined liquid steel per kmol TiO ₂ reversion. CVR at 1 mbar and 2020 ⁰ C. Equivalent (TiO ₂ + ZrO ₂ ) purity is 99.99%; Overall Recovery Of(TiO ₂ + ZrO ₂ ) = 82.50%.

Impurity levels in ascending order are computed as: CrO ₂ , MnO, SiO ₂ , Al ₂ O ₃ < 1 ppm; FeO 1.9 ppm; M ₈ O 2.2; VO 5.4 ppm; (NbO + NbO ₂ ) 11.80 ppm; CaO 42.1 ppm. For pigment purposes, so-called "coloureds" in elemental form: Mn, Cr < 1 ppm; Fe 1.9 ppm; V 4.1 ppm.

Experimental studies are not available on the kinetics of the interaction between methane and oxycarbide melts, but a relevant study of the corresponding interaction of methane top jetted onto the surface of molten iron at 1600 ⁰ C has been reported in the literature (K. Sekino, T. Nagasaka and R. J. Fruehan, Met &Mat Transactions B, Vol. 26B, April 1995, 317-324). Calculations have been made using the Fruehan et al. chemical reaction rate data extrapolated from 1600 ⁰ C to temperatures around 1900 ⁰ C on the likely chemical sub-processes involved in the initial adsorption of methane, coupled with studies undertaken by the inventor and co-workers for mass transfer on top-blowing liquid surfaces under both liquid phase (T. Jones and N. A. Warner: Pyrometallurgy '87, Instn. Min. &Metall., London, 1987, 605) and gas phase (X. Dai and N. A. Warner: Trans. I. Chem. E., Research and Design, 70, A6, 1992, 585) controlled conditions. In considering this methane/melt interaction, it is very important to ensure the avoidance of what the inventor has recently referred to in several publications as the "volcano under an active state of eruption" effect first identified by Bessemer. Conditions must be carefully selected so the sub-surface generation of carbon monoxide does not occur. Also from the viewpoint of carbon deposition and accretion formation within the reactor, it is vital that a methane utilisation efficiency greater than 99% can reasonably be anticipated. The prerequisite for both is for quiescent chemical reaction occurring right on the gas-liquid interface as the principal mechanism involved. The calculations indicate that methane utilisation efficiency greater than 99% is in fact likely to be achievable.

Methane decomposition and the oxycarbide reduction together are highly endothermic as indeed are the reactions with carbon as the reductant. With methane the initial cracking of the molecule to elemental carbon and hydrogen add to overall endothermicity. To mitigate the deleterious quenching effect that these chemical reactions have on the maintenance of close proximity to temperature and composition uniformity, which is mandatory in the present invention, special steps clearly need to be taken. Electrical conductive heating is essential in the present invention

to supply the energy needs. The furnaces involved are thus strictly speaking electric furnaces, but not submerged or open arc of the type used presently throughout pyrometallurgical industry.

With natural gas as the reductant the preferred embodiment avoids high intensity top blowing, as for example is characteristic in LD steelmaking. Instead it is preferable to spread out the top-blowing to cover the whole available surface of gas/melt interface, which in the present case, means both arms of the first melt circulation loop, the top-blown reactor (TBRl), with each jet staying comfortably below the onset of splashing and then rely on melt circulation and the turbulence of the melt to ensure good mixing in the bulk of the liquid phase.

Although the invention is not specifically targeted at very large smelting capacities, these days with emphasis emerging on carbon dioxide capture and storage (CCS) in response to concerns about climate change, it seems inevitable that as the century progresses, pyrometallurgical operations of the future will need to be conducted at large scale and perhaps centralised to realise economies of scale necessary to ease the burden of CCS ₅ air separation for oxygen production and perhaps in-plant electric power generation at sufficiently large scale that modern advanced power systems employing combined cycle can be utilised. The invention facilitates carbon and natural gas-based ihnenite processing to virtually zero gas emission in association with CCS in the future if or when this becomes necessary.

A noteworthy beneficial attribute for smaller levels of production, if natural gas is used as reductant, is potential export of hydrogen-enriched fuel gas to a nearby or possibly across the fence power station and this is most definitely a preferred embodiment. Ihnenite smelting with natural gas effectively results in carbon consumption at the expense of hydrogen leaving as off-gas what is effectively a medium LCV fuel gas, as compared to the very much larger LCV of natural gas itself. It is not difficult to see the attraction of this "by-product", if emission- trading schemes become universal as they already are in the European Union. Also, carbon capture and removal in the longer term for climate change reasons may be facilitated by recycling carbon dioxide to the combustors of gas turbines and the like to control exomermicity, if pure oxygen rather than air is used. Clearly, availability of a medium LCV gas is highly desirable in such a scenario.

A further consequence of using natural gas is that special steps need to be taken to avoid premature thermal decomposition of methane, which would result in carbon deposition and accretion build-up, making the whole concept unworkable in practice. It is well known that heating natural gas below 350 ⁰ C does not result in equipment malfunction and up to this temperature can be tolerated indefinitely. A totally different picture emerges in the present case, if innovative steps are not put securely in place to preclude what is potentially a recipe for catastrophic failure. In essence, it is essential in the present invention to prevent contact of methane with solid surfaces that are hotter than 350 ⁰ C. Initially, a number of solutions to the problem merited closer evaluatioiL, which led ultimately to identification of the preferred option for this embodiment of the invention.

To put the matter in its correct perspective, it must be borne in mind that, as already stated, the preferred embodiment sees as mandatory provision of low intensity top-blowing over the entire available area of gas/liquid interface to prevent reaction endotheπnicity from disrupting the over-riding requirement of proximity to temperature uniformity to ensure stability of the oxycarbide linings.

On heat loss grounds, an elementary calculation rules out maintaining a cooled roof at a temperature of around 35O ⁰ C above the melt and passing the large number of nozzles or top-blow lances through this relatively cool environment before the methane emerges at high velocity into the gas space above the melt followed then by minimal clearance between the roof and the melt surface to substantially limit enlrainment of hot gas into the methane jet. In practice this would mean a clearance of say about 10 cm between the ceiling and the melt surface, which in itself is achievable in a melt circulation system, given the ability to control melt levels accurately. However, to illustrate the ramifications of this geometrical arrangement, it is first necessary to put forward some initial estimates of plant requirements.

It is assumed that some 500,000 tonnes per annum of TiO ₂ or equivalent in Ti metal are being produced from a Iypical ilmenite concentrate. Accordingly, about one million tonnes per annum (1 Mtpa) of ilmenite is being smelted. For top-blowing with say 2.5 kmol CBL|/kmol TiO ₂ , it can be calculated that to avoid splashing with 1 cm diameter jets placed 10 cm above the melt would require some 1880 jets and if these are placed to form an array with a triangular pitch of 0.5 m, the total bath area is 407 m ^A . This means that with both arms of TBRl melt circulation loop being top blown, each arm would need to have an area available for top blowing 6 m wide by 51 m in length. In other words, some 407 m ² of cooled surface at 350 ⁰ C would be receiving about 338 MW of direct radiation from the bath. This figure is to be compared with the theπnochemical enthalpy requirement in the region of 260 MW. Clearly, a cooled roof is an untenable proposition.

Next is outlined the essentials of a scheme, which has its roots in the cooling of 1500 ⁰ C class gas turbine rotors and stators, where combustor gases enter the turbine. For this demanding situation, technology has been developed, which uses closed loop steam cooling for steam reheating duties in association with combined cycle power generation. The same approach is identified as the means for cooling individual top blowing lances in the present invention as they pass through a roof enclosure about 0.3 - 0.5 m in thickness at an average temperature of say 1800°C before the temperature is moderated somewhat.

Flat refractory suspended arches or suspended ceilings with multiple support systems are commercially available and are used extensively in industry. Clearly, the present very high temperature requirements are beyond the range of current designs, but if prefabricated shapes of titanium oxycarbide of the appropriate solidus composition were deployed on the hot face of the ceiling backed with graphitic or baked-carbon thermal insulation and then conventional refractories as the temperature level decreases, this satisfies the specific needs of the present invention. A preferred embodiment would entail closed loop steam cooling of steel lances placed concentrically in and surrounded by say 2 or 3 radiation shields within the very hot 0.3 - 0.5 m immediately above the hot face, and oxycarbide materials would be used to fabricate the cylindrical elements so required. Then as temperature decreases, there is a progressive changeover from oxycarbide radiation shields to alumina or other high temperature ceramics. Alternatively, a design based on graphite or baked-carbon rigid insulation would appear feasible, especially if the roof area is purged with a non-oxidising H ₂ /CO gas mixture derived from the hydrogen enriched "fuel gas" previously referred to with its already low carbon dioxide and water vapour contents reduced even further before being recycled as purge gas.

By either of the afore-mentioned means, natural gas can be top-jetted onto the oxide melt surface throughout both arms of TBRl so that thermal decomposition does not take place and carbon deposition accretion problems do not arise. The material of solidus composition constituting the flat ceiling some 10 cm or so above the melt is in theory close to thermodynamic equilibrium and would last indefinitely provided bulk gas velocities are not excessive and erosive wear then a problem.

As there is no merit in attempting gas/liquid counterflow in the present invention, gas removal at relatively low velocities can be accomplished throughout the length of both arms on each side through a number gas off-takes with the hot faces fabricated from material of the solidus composition. In effect the bulk gas and the liquid phases are flowing in a cross-flow pattern with bulk gas flow in both directions from the centre line, so questions concerning interfacial stability even with the relatively small clearance between the melt and the flat ceiling of about 10 cm or so, do not arise.

The discussion in this description mainly deals with natural gas usage, but it is important to reiterate that the preferred reductant will be a matter for experts through the world to decide on depending on prevailing local conditions. However, for Europe it does seem that natural gas is the frontranner provided of course the supply is secure and is available at the right price.

If market demand exists and the return on investment is greater with metal as opposed to in-line pigment manufacture, then all the oxycarbide melt after partial oxidation to reduce the oxycarbide concentration would proceed in-line to electrochemical deoxygenation. This is the scenario, which will now be evaluated, but it is not the lowest energy option to titanium. TiO is by far the lowest energy precursor for titanium metal production by electrolytic deoxygenation. Equally, the greater the carbide concentration of the melt in conjunction with TiO provides the lowest energy consumption for direct in-line chlorination, if this is considered by the industry to be the preferred route to pigment manufacture. Thus in both cases, oxidation of the melt to reduce the carbide concentration is in energy terms counterproductive. However, the practical reality would be that as titanium metal is produced, the melt would enter the two-phase region, ultimately leaving a mud of TiC dispersed in oxycarbide. Clearly, if only a relatively small demand for metal exists, only enough melt would be electrolysed to satisfy that demand and the residual melt would then proceed to autogenous CO ₂ /O ₂ oxidation to produce 99.99% purity TiO ₂ . Alternatively, if pigment manufacturers were to utilise the vast amount of no-how that already exists in the industry, then direct in-line chlorination without prior oxidation would perhaps be the best way forward.

Consider now the oxidic melt in Example 2 being subjected to thermally balanced oxidation with a CO ₂ /gas mixture at 1940 ⁰ C in advance of electrochemical deoxygenation at 1860 ⁰ C, so that greater than 97% of the contained titanium is recovered as metal, before the two-phase liquid/solid region is entered.

According to Cardarelli (F. Cardarelli: "A Method for Electrowinning of Titanium Metal or Alloy from Titanium Oxide Containing Compound in the Liquid State" Quebec Iron & Titanium Inc., PCT Patent Application, International Publication No. WO 03/046258 A3, June 2003). 99.9% pure TiO ₂ , electrolysed with pure CaF ₂ as electrolyte and a carbon anode at 1860 ⁰ C yields titanium at a "Faradaic efficiency" of 95% and an energy efficiency of 62% as ingot metal with a purity in excess of 99.9%. Assuming electrolysis under reversible

conditions (i.e. infinitely low current density) at 186O ⁰ C in conjunction with the HSC4 computer program (A. Roine: Outokumpu HSC Chemistry ^® 4.0, Outokumpu Research Oy, Finland, 1999). and then applying the two efficiency figures, the author's calculations confirm the specific energy consumption is less than 7.0 kWh kg^Ti.

For the particular melt under discussion, after partial TiC removal at 1940 ⁰ C, the equilibrium composition in mol % is: TiO 20.74%; Ti ₂ O ₃ 32.0%; Ti ₃ O ₅ 32.09%; TiO ₂ 14.75%; TiC 0.42%. Ti metal production with this melt as feed going right through to completion would consume an estimated 4.91 kWh kg ^" *Ti of electricity, compared with 6.98 IcWh kg ^-1 Ti evaluated by the author for the Cardarelli conditions.

The above figures are in marked contrast to the figures given by Norgate et al. (T. E. Norgate, V. Rajakumar and S. Trang: AusIMM Green Processing Conference, Fremantle, W. A. 10-12 May 2004, 105-112) for the projected electrolytic cell power consumption for the FFC Cambridge Process of 15.19 IcWh kg ^"J Ti, assuming the latter with further technological development will improve the cell reaction efficiency from the energy efficiency of 10% reported in the literature up to 50% as proposed by Norgate et al. Bearing in mind that a further 1 kWh kg ^"x Ti is necessary for one stage vacuum arc remelting (VAR), this brings the projected FFC electricity consumption to 26.2 kWh kg ^": Ti in the metal production stages.

The recently proposed Plasma Powder Process according to Norgate εt al. has an estimated electricity consumption of 25.2 kWh kg^Ti to which a further 1 kWh kg^Ti has to be added for single stage VAR, giving an overall electricity consumption of 26.2 kWh kg ^"x Ti.

Direct production of molten titanium from a 99.9% pure oxidic melt would not require VAR. Thus the electricity consumption for the proposed process would appear superior to other emerging titanium metal technologies. However, to make a more comprehensive comparison, it is necessary to evaluate the so-called "Process Fuel Equivalent (PFE)" introduced initially by Kellogg. (H. H. Kellogg: JMetals, 26, June 1974, 25-9). PFE values are burdened with the uncertainty concerning the appropriate efficiency for electricity generation from fossil fuels. Norgate et al. consider current black coal electricity generation in Australia and assign a figure of 35%. On this basis, the PFE for the overall carbon-based proposed direct smelting process for ilmenite is 110 GJn ₁ /t Ti.

The corresponding gross energy requirement (GER) figure evaluated by Norgate et al. for the Kroll process is 361 GJB ₁ /t Ti, including the gross energy involved in dredging and mineral processing. The precise figures for the latter components are not given in the Norgate et al. ²² paper, but inspection of their graphical figures would suggest that the combined figure for both is somewhat less than say 6 GJm /t Ti, so a conservative estimate for the GER is 355 GJ _Th /t Ti. Assuming that various other components in the GER so evaluated are relatively small, the PFE value for the Kroll process is in the vicinity of 355 GJn ₁ /t Ti.

According to these preliminary calculations, the proposed technology for continuous smelting of ilmenite directly to titanium metal appears to offer energy consumption less than one third of the current best available technology (Kroll process). This is before claiming a credit for the co-produced liquid steel.

In the early part of the description of the invention, it was mooted that certain ilmenite concentrates contained radioactive impurities, which were unlikely to be removable during the proposed smelting reduction and in-line continuous vacuum refining approach, which is central to the new technology and a complementary chlorine- based route may be needed to be included within the "downstream processing options" of Fig. 1. There is also a possibility that mechanical milling or other appropriate means to achieve particle size requirements for pigment manufacture starting with a solidified melt, may not be viable or may be too energy intensive to obtain cost- effectively the preferred crystal size, which is believed to be about 0.2 micron and thus pigmentary titanium dioxide is normally characterised by crystal sizes in the range 0.1 to 0.5 micron. Therefore, besides the aforementioned radioactivity problem, pigment manufacturers may prefer to rely on crystal growth from the vapour phase, i.e. oxidation of gaseous titanium tetrachloride.

A preferred embodiment is thus to continue the in-line processing by direct carbochlorination of the refined oxycarbide melt immediately after continuous vacuum refining, when the melt is returned to atmospheric pressure. This sub-process is best carried out at virtually the same temperature and melt composition, characteristic of the CVR step. Using the teachings of the present invention, carbochlorination is preferably conducted in a melt circulation loop, which is fully back-mixed so that operation of the melt close to the liquidus temperature enables utilisation of oxycarbide linings of the solidus composition for the hearth, walls and ends of both arms of the melt circulation loop. As before, it is mandatory to minimise temperature and compositional variations throughout this loop, which of course implies relatively large melt circulation ratios without incurring high intensity reactions locally at any stage during the closed loop circulation. Accordingly, the preferred embodiment is to top blow with natural gas under relatively mild non-splash conditions uniformly over the whole area of one arm of the loop and then on the other arm to. add chlorine judiciously so that virtually all the melt surface area is exposed to relatively mild reaction intensity. By these means, the stability of the oxycarbide- unmelted shell constituting the hearth, sidewalls and ends is assured.

Similarly, for the same reasons as discussed earlier in relation to thermal decomposition of methane, the same steps must be taken in the carbonising arm of the carbochlorination melt circulation loop using natural gas to ensure uninterrupted operation. The necessary steps will not be repeated here, except to confirm that the flat suspended arch through which the methane jets emerge into the narrow gas gap between the ceiling and the melt surface will also be lined with oxycarbide of the solidus composition.

On the chlorination arm, the ceiling can be fabricated from graphite, as this is resistant to chlorine, carbon monoxide and the metallic chloride vapours, which constitute the gas phase on this particular arm. A very small gap, probably purged with carbon monoxide or other inert gas, needs to be maintained between the oxycarbide unmelted shell and the graphite upper regions. It is obviously important that the temperature of the melt, being already close to the liquidus temperature in relation to the oxycarbide lining, is not permitted to decrease so that the two-phase solid/liquid region is entered, so careful control of temperature using balanced conductive heating and heat removal to steam raising or other appropriate heat removal means, is required for this purpose.

The same reasoning is applicable when considering the so-called "condenser" for removing thorium chloride in the gases evolved from the chlorination arm of the carbochlorination melt circulation loop. The fundamental

mechanism involved here is gas absorption, which is very likely to be gas-phase mass transfer controlled. In 1953 the present inventor began his long association with pyrometallurgy in the Chemical Engineering Department at the University of New South Wales in a PhD programme, elucidating the mechanism of the so-called "condenser" of the then new Imperial Smelting Furnace (ISF) or lead-zinc blast furnace, in which zinc vapour is removed from furnace gases by absorption into a closed melt circulation loop employing molten lead as the solvent. Exactly the same considerations apply in the present case. Molten potassium chloride is force circulated around a closed loop employing electric motor-driven mechanical rotors to generate the intensive splash conditions, characteristic of an ISF condenser, in order to absorb the gaseous thorium chloride into the fused salt. ISF condenser rotors are of alloy steel construction but for the present case graphite construction would eliminate any concerns about corrosion or contamination of the fiised potassium chloride melt. In a similar application the aluminium industry has been well served by centrifugal pumps constructed of graphite at comparable temperature levels of operation. Commercially proven technology is therefore readily available for all aspects relating to fused salt splash "condenser" proposed in the current invention. Also incorporated are the teachings of the invention and the maintenance of an unmelted shell of solid potassium chloride on the hearth, walls and ends of both arms of the splash "condenser" melt circulation loop, require operation of this loop close to the melting point and removal of heat involved in the quenching of the hot gases from the carbochlorination melt circulation loop by conduction to steam raising tubes on the cooler faces of the solidified melt.

A small portion of the fused salt melt is withdrawn either continuously or intermittently to regenerate the potassium chloride content by electrolytic deposition of thorium (and probably calcium) or other appropriate chemical means before the withdrawn potassium chloride melt side stream is reverted back to the splash "condenser" melt circulation loop. The thorium will be deposited in the solid state on a non-consumable electrode arrangement and can be recovered using well-established procedures.

The general philosophy behind the aforementioned approach for dealing with unacceptable radioactive impurity levels in certain ilmenite concentrates is one of concentrating to contain as opposed to diluting to disperse. Environmental agencies worldwide increasingly will need to be convinced that the best possible steps are taken to contain threats to the environment and the issues under discussion must be seen in this context. Some other emerging technologies applaud their removal of thorium during beneficiation and end up with it dispersed in an iron oxide of possible commercial value. However, to claim that this iron oxide material could be utilised in iron and steelmaking is probably flawed. The iron and steel industry is rightly proud of its recyclability credentials, so is unlikely to get involved in radioactive waste disposal issues with strong environmental protection overtones.

Referring now to Fig. 12, this is effectively a flowchart encapsulating the key features of the aforementioned chlorine-based technology seen to be included as one of the "downstream processing options" referred to in Figs. 1 and 2. Fig. 12 shows the two aforementioned melt circulation loops in plan view, one for carbochlorination and the other representing the splash "condenser", i.e. absorber, for removing gaseous thorium tetrachloride from the gas evolved in the chlorination arm of the carbochlorination melt circulation loop. The flow sequence begins at the bottom left-hand corner of Fig. 12 with the underlined "Oxycarbide Melt ex Continuous Vacuum Refining (CVR)" and proceeds initially upwards and then enters territory on the right-hand side of the diagram, which legitimately can be referred to as "Prior Art" in terms of published Patent Applications, Patents as well as actual present day commercial practise, reported in the literature.

Besides electrochemical de-oxygenation of oxidic titanium melts, which has been highlighted so far throughout the description of the present invention, there is a respected body of opinion that sees high temperature electrolysis of titanium tetrachloride dissolved in a molten salt electrolyte in order to produce liquid titanium metal initially as the best way forward. The contributions of Ginatta in US Patent No. 6,074,545 and related material e.g. M. V. Ginatta: "Extractive Metallurgy of Primary Titanium", Light Metal Age, Vol. 62 Pt. 1-2, 48- 51, 2004. More recently a team from General Electric Company (Carter JR. et al.) have amplified the Ginatta approach somewhat in US Patent Application No. US 2005/0145065 Al.

The present invention offers a cost-effective means for supplying continuously the refined titanium tetrachloride either in gaseous or liquid form ready for incorporation into the Ginatta/Carter JR. et al. chlorine-based electrochemical approach to titanium melt production on a large scale. It can thus be seen to be the front end of truly continuous titanium metal production directly in-line from ilmenite or other titaniferous minerals.

Let us know consider a preferred embodiment for processing an ilmenite concentrate with a potential radioactivity problem. For example, the chemical analysis of a weathered Australian Murray Basin ilmenite concentrate in mass percentage is as follows:

TiO ₂ 58.40; Fe ₂ O ₃ 27.95; FeO 2.45; Al ₂ O ₃ 1.98; SiO ₂ 2.74; MgO 1.46; MnO 0.97; P ₂ O ₅ 0.29; CaO 0.11; Nb ₂ O ₅ 0.07; V ₂ O ₅ 0.26; Cr ₂ O ₃ 1.20; ZrO ₂ 0.33; UO ₂ 30 ppm; ThO ₂ 316 ppm.

The equilibrium melt leaving CVR at 1 mbar and 2035 ⁰ C after carbon-based smelting reduction in two melt circulation loops comprised of SRRl at 1970 ⁰ C with 3 kmol refined liquid steel reversion, SRR2 at 2160 ⁰ C with 2 kmol refined liquid steel reversion with carbon additions of 2.15 kmol C/kmol TiO ₂ in SRRl and 0.55 kmol C/kmol TiO ₂ in SRR2, is estimated to contain the following elemental impurities: less than 1 ppm Cr, Mn, Al, Si, Mg; 1.1 ppm Fe; 2.7 ppm V; 6.3 ppm Nb; 46 ppm Ca; 61 ppm U; 648 ppm Th and 4783 ppm Zr, all based on an equivalent "TiO ₂ " recovery of around 83% of the input TiO ₂ in the ilmenite feed. The actual composition entering the flow chart depicted in Fig. 12, in this particular case, if equilibrium were established, would be expected to contain the following principal constituents in mol fraction: TiO = 0.756; Ti ₂ O ₃ = 0.017; TiO ₂ = 0.002 and TiC = 0.220 (sum equals 0.995).

Carbochlorination of this melt at 2035 ⁰ C would consume per initial kmol TiO ₂ in the ilmenite feed concentrates a total of 1.665 kmol Cl ₂ and 0.486 kmol CH ₄ , based on an equilibrium gas phase composition at say 775 ⁰ C (mean temperature for the splash "condenser" melt circulation loop) of almost 100% TiGU(g) as the main chloride molecular species involved with principally CO as the other major gas phase component at this lower temperature level. This assumes that as the gases are cooled no carbon deposition occurs and reactions involving CO and CO ₂ are effectively frozen once HOO ⁰ C is reached.

Using experimental data reported by Russian workers at the Institute of Electrochemistry, Sverdlovsk (M. V. Smirnov and V. YA. Kudyakov: "The Saturation Vapour Pressure and Decomposition Potential of ThCl ₄ Solutions in molten Alkali Chlorides" Electrochimica Acta, Vol. 29 No. 1, 63-68, 1984), conditions have been evaluated for a proposed operating temperature of the splash "condenser" of say 775°C to achieve a target

elimination of thorium chloride in the gas phase down to a level such that less than 5 ppm Th contaminates the "TiO ₂ " equivalent product. To achieve this objective requires stabilisation of the dissolved ThCL _t in the binary potassium chloride melt to be at about 0.5 mol%. This estimate is based on an in-depth evaluation of the likely rate controlling diffusional resistance in the gas absorption process involved. Liquid-phase difrusional resistance will be negligible, so it is appropriate to base calculations on exclusive gaseous diffusion mass transfer control, The equivalent number of overall gas phase transfer units (N _O α) equals 6.55 for the desired target elimination of thorium chloride in the splash "condenser" melt circulation loop. This is very conservative estimate, because it assumes that all the gaseous thorium chloride not absorbed will form a liquid mist initially and then a solid fume as the temperature is reduced below 775 ⁰ C and ultimately it will all report in the TiCl ₄ intermediate product.

If it is deemed necessary to condense out ZrCl ₄ , or more correctly precipitate out solid ZrCl ₄ because it sublimes at atmospheric pressure, in advance OfTiCl ₄ recovery from the TiCl ₄ (g)/CCγ _g ) gas mixture, then both ZrCl ₄ and ThCl ₄ will be filtered out before TiCl ₄ condensation. However, it is debatable whether or not ZrCl ₄ elimination would be necessary, as ultimately at the concentration levels involved in say a TiO ₂ pigment, the zirconia would be present as a very minor component in solid solution in rutile, for example. In this context it is worth noting that some pigment manufacturers purposely add a small mount of zirconia to improve TiO ₂ pigment properties.

Rutile pigment material produced by the chloride and sulphate routes are basically similar and to improve dispersibility, dispersion stability, opacity, gloss and durability, both require finisliing, which includes coating with inorganic compounds by selective precipitation on milled aqueous suspensions of the titanium dioxide material. This invention is not directly concerned with these downstream finishing operations but it is anticipated that the finishing requirements will parallel those of existing pigment manufacture.