R.T. Jones, D.A. Hayman, and G.M. Denton

Pyrometallurgy Division, Mintek, Private Bag X3015, Randburg, 2125, South Africa


Valuable metals, such as cobalt, can be recovered from slags by treating these waste materials with a carbonaceous reducing agent in a DC-arc furnace. This technology is applicable to the recovery of nickel, cobalt, copper, and platinum group metals from furnace or converter slags emanating from copper or nickel smelters treating sulphide concentrates. Pilot-plant testwork at Mintek has demonstrated recoveries of 98% for nickel and over 80% for cobalt, at power levels of up to 600 kW. Results of computer simulations and experimental tests are presented.


Slags produced by non-ferrous smelters usually contain significant quantities of valuable metals, such as cobalt, nickel, and copper, present both in an entrained metallic or sulphide form, and in a dissolved oxidized form. These slags can emanate from either furnaces (e.g. six-in-line, reverberatory, or flash) or from converters; they may be internal recycle streams or final waste products. Large quantities of valuable metals are present in the huge dumps that have built up over many years of operation of nickel and copper smelters. Slags are sometimes treated by slow cooling, milling, and flotation. This approach is satisfactory when the metals in the slag are in either the sulphide or metallic form, but is not suitable for the recovery of oxidized metals. Cobalt in particular, and nickel to a lesser extent, are found in an oxidized form, and for these slags, treatment in an electric furnace operated under reducing conditions is necessary.

Conventional slag cleaning furnaces (typically AC submerged-arc) rely largely on a gravity settling mechanism, whereby the entrained sulphide and metallic droplets are simultaneously collected. Sometimes, a quantity of matte is added to the slag, to enhance the coalescence of entrained matte droplets. However, conditions are not usually sufficiently reducing to recover much of the cobalt. Cobalt recoveries may be as low as 20 per cent. The most effective means for the recovery of metals involves the addition of a reductant (such as carbon) to capture some of the metals present in an oxidized form. Processes have previously been described in which carbothermic reduction has been carried out in electric furnaces (1-3).

The aim of a slag cleaning process is to maximize the recovery of valuable metals (such as Co, Ni, and Cu) in an alloy with the lowest possible iron content. The amount of metallic iron produced should be kept to a minimum, as the more iron present in the resulting matte or alloy, the greater the cost of the subsequent hydrometallurgical separation of the valuable metals, and the resulting disposal of the iron residues. Because of the similarities in the reduction behaviour of cobalt and iron, some loss of cobalt is inevitable while separating the iron from the nickel and copper.

Mintek has been working on the recovery of cobalt, and the associated valuable metals, from slags, since 1988, using DC-arc furnace technology to effect selective carbothermic reduction of the oxides of cobalt, nickel, copper, (and zinc, where present), while retaining the maximum possible quantity of iron as oxide in the slag.

Flowsheets for slag cleaning

Many non-ferrous smelters employ a process whereby the concentrates are fed to a furnace which produces a matte (for further treatment) and a slag (which is dumped). The furnace matte is treated in a converter (often of the Peirce-Smith type) to remove most of the remaining iron and sulphur. This resulting 'white matte' or alloy is then treated hydrometallurgically. The converter slag is usually recycled to the furnace. Because of the highly oxidizing conditions in the converter, much of the cobalt is oxidized. The turbulent conditions cause entrainment of valuable metals as well. Of all the streams in a flowsheet of this sort, the converter slag is richest in cobalt. As shown in Figure 1, it is possible to divert this liquid stream of converter slag for slag cleaning, allowing most of the valuable metals to be reclaimed. The impoverished slag can still be recycled to the furnace (with fairly minimal disruption to the existing process, and with the benefit of reduced quantities of magnetite which otherwise builds up in the furnace), or can be dumped (breaking the recycle entirely, necessitating some changes to the operation of the furnace).

Figure 1 - Flowsheet for cleaning of converter slag

A second possibility is to leave the converter slag recycle stream alone, and focus on the treatment of the furnace slag which is the point at which the waste materials finally leave the process. It is also possible to treat material from existing slag dumps at the same time, as the dumped material is usually similar to the furnace slag currently being produced. This option is shown in Figure 2. It is, of course, also possible to use a hybrid of these approaches.

Figure 2 - Flowsheet for cleaning of furnace slag (also applicable to the cleaning of flash-furnace slag, without a converter present)


A wide range of slags are amenable to slag cleaning. These slags differ according to the ores which have been processed, as well as according to the type of process, and whether the slags have arisen from furnaces or from converters. Most of the slags of interest are rich in iron oxide and silica, and many have a bulk composition approximating that of fayalite. Fortunately, similar principles apply to the treatment of all of these slags, although the actual results will differ according to the composition of the slag. For the sake of illustration, two representative slags are shown in Table I, differing primarily in the level of cobalt contained in the slag.

Table I - Compositions of Typical Slags to be Cleaned
Al2O3, %
CaO, %
Co, %
Cu, %
FeO, %
MgO, %
Ni, %
S, %
SiO2, %
High Co slag
Low Co slag

Note that, for the purposes of these calculations, all of the Co, Cu, and Ni in the feed slag was assumed to be in the oxidized form, i.e. present as CoO (1.272% or 0.254%), Cu2O (0.563%), and NiO (2.546%).

When carbon is added to the slag, the various metallic elements reduce to different extents, at a given level of carbon addition. This behaviour allows a reasonable degree of separation to take place during smelting. The intention in this part of the process is to separate the valuable non-ferrous metals from the iron and the gangue constituents present in the slag. Figure 3 illustrates the differences in reducing behaviour between nickel, cobalt, and iron. The desirable area of operation is clearly somewhere in the region where the recovery of cobalt is high, and the recovery of iron to the alloy is still reasonably low. Note that, in actual practice, there is less than 100 per cent carbon utilization, and the carbon addition would need to be somewhat higher than that shown here, because of burn-off of some of the reductant.

Figure 3 - Recovery of elements to the alloy, as a function of the quantity of carbon added, at 1500°C

The calculations for Figures 3, 4, and 5 were carried out using Mintek's Pyrosim computer software (4) for the calculation of steady-state mass and energy balances for pyrometallurgical processes. These simulations were based on the assumption of chemical equilibrium between gas, slag, and alloy. The equilibrium composition was calculated using free-energy minimization, together with the Ideal Mixing of Complex Components solution modelling technique. An entrainment of 2 per cent of the matte in the resulting slag was assumed, as this was found to agree well with experimental results from DC-arc furnace testwork. Of course, there is a spread of experimental data around the simulated values, as the model does not take into account subtle differences in the mode of operation of the furnace. It should be noted that a good mineralogical description of the slag feed is required in order to provide an accurate estimate of the energy requirements of the process. (Pyrosim can also be used in non-predictive mode, using the empirical Pyrobal model, which allows the user to specify, for example, the actual experimentally obtained percentages of cobalt and nickel in the slag, or the distribution of iron between metal and slag. This is very useful for generating a completely consistent mass balance from incomplete or conflicting experimental data, where at least some information is known to be accurate.)

The equilibrium behaviour shown in Figure 3 is not very sensitive to temperature, and curves plotted for temperatures 100°C colder or hotter are virtually indistinguishable from those presented. Clearly, the temperature has effects other than on the chemical thermodynamics of the system. It has been reported elsewhere (6) that a 100°C increase in temperature may decrease the cobalt and nickel solubility in slag by as much as three times, at a constant partial pressure of oxygen, and at a temperature around 1300°C. However, the solubility is even more strongly affected by the reducing nature of conditions in the furnace (i.e. the partial pressure of oxygen in the system).

The most striking feature of this separation process is the variation of cobalt recovery according to the iron content in the alloy. This behaviour is shown in Figure 4. It can be seen that the recovery of cobalt (in percentage terms) is highest (albeit not by very much) in the case of the slag with the highest initial cobalt content. This is in line with experimental findings (2,7) that recoveries are dependent on initial slag composition, with higher grades leading to better recoveries. If we accept the evidence presented elsewhere (7) that metallic Fe is the effective (intermediary) reductant in the process, it may be more correct to say that the recovery of valuable metals is related to a combined function of the iron and non-ferrous metal contents of the initial slag.

Figure 4 - Recovery of Co to the alloy, as a function of the amount of Fe present in the alloy, at two levels of Co (1.0% and 0.2%) in the slag fed to the slag-cleaning furnace

However, as shown in Figure 5, the residual cobalt content in the cleaned slag, as a function of the iron content of the alloy, is very sensitive to the initial cobalt content.

Figure 5 - Residual Co content in cleaned slag as a function of the Fe content in the alloy

Alloy melting point

The major constituents of the alloys produced melt at rather high temperatures, as shown in Table II. The melting point of the alloy increases sharply with iron content, and melting temperatures in the range from 1300 to 1420°C have been reported (2). This is one of the important determinants of a suitable operating temperature for the furnace. An operating temperature of 1500 to 1550°C has been selected for the process, in accordance with that used for a similar process (2).

Table II - Melting Points of Pure Elements, °C


The experimental work on the slag-cleaning process began on the laboratory scale, and was extended to pilot scale on Mintek's DC transferred-arc furnaces (8). In addition to numerous 100 kVA (60 kW) supporting batch tests, five campaigns (of 50 to 100 hours each) have been carried out at the 200 kVA (150 kW) scale, and a campaign (treating about 20 t of slag) has also been undertaken on the 3.2 MVA (600 kW) furnace. Future work is planned for the 5.6 MVA (1-3 MW) furnace facility.

The DC-arc furnace has a single electrode positioned above the molten bath; the molten metal in the furnace forms part (the anode) of the electrical circuit. The furnace comprises a refractory-lined cylindrical steel shell, and a water-cooled roof lined with an alumina refractory. The outer side walls of the furnace are spray-cooled with water, to protect the refractories, and to promote the formation of a freeze lining within the vessel. The roof contains the central entry port for the graphite electrode and up to three equi-spaced side feed ports. The return electrode, or anode, consists of multiple steel rods built into the hearth refractories and connected at their lower end to a steel plate which, via radially extending arms, is linked to the furnace shell, and further to the anode cable. A schematic diagram of this arrangement is shown in Figure 6.

Figure 6 - Schematic diagram of a DC transferred-arc furnace

Molten slag may be fed directly to the 5.6 MVA furnace from a pre-melting furnace. The solid-feed system for the 3.2 MVA furnace comprises a batching plant and a final controlled feeding system. The batching plant consists of feed hoppers mounted on load cells, vibratory feeders positioned under the hoppers, an enclosed belt conveyor, a bucket elevator, and a pneumatically actuated flap valve to direct the feed to one of two final feed hoppers. The final feeding system is made up of separate centre and side feeding arrangements. The centre feeder uses a screw feeder discharging into a telescopic pipe attached to the hollow graphite electrode. The side feeders are vibratory, and discharge into feed chutes.

The gas-cleaning system consists of a water-cooled off-gas pipe, a refractory-lined combustion chamber, water-cooled ducting, a forced-draft gas cooler, a reverse-pulse bag filter, a fan, and a stack. The condensed fume and dust, which accumulates in the lower conical section of the bag plant, is discharged via a rotary valve into a collecting drum. This dust would, of course, be recycled back to the furnace in an industrial situation.


Copper converter slag - 100 kVA

During 1988 and 1989, tests were carried out using a converter slag having a composition of Co: 0.46%, Cu: 3.25%, Fetotal: 52%, and Ni: 0.42%. A 100 kVA furnace (operating at 20 to 25 kW) was used. In some tests, solid slag and coal were fed together, while in others the slag was first melted then the coal added afterwards. Coal (at 53% fixed carbon) additions varied between 4, 6, 8, 10, and 12 per cent of the mass of the slag. The mode of addition of the coal, and the reduction periods of the tests were also varied (around 30 minutes). The results of the tests are shown in Table III and Figure 7.

Table III - Results of 100 kVA Tests on Copper Converter Slag
Fe in alloy, %
Co recovery, %
Cu recovery, %
Ni recovery, %
8% coal, solid slag
8-12% coal, molten slag
10% coal, molten slag

Figure 7 - Results of 100 kVA test on copper converter slag

The use of fine coal (less than 1.5 mm) did not seem to have an effect on the degree of reduction. A representative composition of the alloy produced was Co: 6%, Cu: 32%, Fe: 53%, and Ni: 6%.

Nickel-copper converter slag 'A'- 100 kVA and 200 kVA

Using converter slag from a nickel-copper plant, tests were carried out, in 1990, on a 100 kVA furnace (operated at 30 kW) and on a 200 kVA furnace (operated at 85 kW), at low additions of reductant (in order to minimize the reduction of iron). The slag had a composition of Co: 0.45%, Cu: 3%, Fetotal: 47%, Ni: 3.5%, and S: 3%. These tests examined four different methods of operation, in an attempt to optimize the selective reduction of the slag. These methods included smelting of composite pellets of milled slag and graphite, adding selected quantities of crushed coal to already molten slag, co-feeding crushed cold slag and coal, and pneumatic injection of pulverized coal into the molten slag. On the 100 kVA furnace, the alloy produced typically comprised Co: 1.7%, Cu: 14%, Fe: 48%, Ni: 16%, and S: 14%. On the 200 kVA furnace, an alloy of Co: 2%, Cu: 15%, Fe: 44%, Ni: 22%, and S: 10% was produced. At this level of reduction, 91% of the iron was retained in the slag phase, while only 50% of the cobalt, 63% of the copper, and 83% of the nickel were recovered to the alloy. It was found that injection of pulverized coal greatly improved the reduction, and hence the recovery, of nickel and cobalt oxides from the slag. The results of these tests are shown in Figure 8, where the scatter of results needs to be seen in the context of the various methods employed. Under good conditions, at 7% carbon addition, on the 100 kVA furnace, calculated recoveries of Co: 81%, Cu: 78%, and Ni: 97% were obtained, while retaining 80% of the iron in the slag.

Figure 8 - Results of 100 kVA and 200 kVA tests on nickel-copper converter slag

Nickel-copper converter slag 'B' - 200 kVA

This slag was generated in a plant utilizing a conventional six-in-line furnace and Peirce-Smith converter configuration. The composition of the bulk slag was Co: 1.25%, Cu: 1.0%, Fetotal: 49%, Ni: 3.6%, and SiO2: 30%. During testwork carried out in 1993, the furnace was operated at power levels of 100 to 170 kW and tapping temperatures of 1400 to 1500°C. The alloys produced comprised Co: 4.5 to 5.5%, Cu: 5.5 to 8.5%, Ni: 25 to 35%, Fe: 35 to 50%, and S: 8 to 10%. From the starting level of 1.25% cobalt in the feed slag, it was possible to produce a discard slag with typical values of 0.22 to 0.29% cobalt. Results from the campaign are summarized in Figure 9.

Figure 9 - Results of 200 kVA nickel-copper converter slag campaign

Co-feeding of the coal and sequential feeding (addition of coal to a molten slag bath) gave similar recoveries to the alloy of 74 and 80% respectively. While cobalt recoveries were similar, sequential feeding (the preferred route for an industrial plant where molten slag is available) produces a better grade of alloy, with an iron content in the region of 10 to 15% lower.

The most abundant phase present in the solid converter slag is Fe2SiO4 (olivine), followed by Fe3O4 (spinel). The cobalt is associated primarily with the olivine, whereas the nickel is distributed between the olivine and the spinel. Copper was present only in entrained sulphides. The analysis of highly reduced slags has shown that it is possible to remove virtually all of the cobalt and nickel from the olivine in the slag.

Nickel-copper converter slag 'B' - 3.2 MVA

Large-scale testwork on converter slag was conducted, during 1994, on a 3.2 MVA DC-arc furnace operating at a power level of 600 kW. The sequential feeding of reductant was used as the preferred mode of feeding. Operating temperatures were in the region of 1300 to 1600°C, and neither the temperature of the bath before the reduction period nor the tapping temperature seemed to have a pronounced effect on the cobalt recovery.

The average electrode consumption during the campaign was 2.6 kg/MWh, while the dust loss was low, at 1% of the mass of the feed. The alloy produced comprised Co: 7.8%, Cu: 3.8%, Ni: 26.4%, Fe: 56.9%, and S: 2.1%. The cobalt levels achieved in the discard slag were between 0.15 and 0.33%.

A coal addition of 9% was required to achieve cobalt recoveries of at least 80%. Increasing the coal addition did not significantly increase the recovery of cobalt. Increasing the batch mass of slag from 500 kg to 1000 kg and increasing the reduction period by 75% resulted in increases in cobalt recovery from 71 to 86%, and from 70 to 82%, for coal additions of 9 and 11% respectively. The main factor affecting cobalt recovery appears to be the time allowed for the reduction to take place. At this scale of operation, a duration of two hours was required to achieve cobalt recoveries greater than 80%.

Nickel-copper furnace and converter slag -200 kVA

During 1995, a campaign was undertaken on the 200 kVA furnace with the intention of combining furnace slag with the converter slag previously treated. When treating furnace slag containing 0.22% Co on its own, the maximum cobalt recovery that could be obtained was 66%, with a cobalt value of 0.08% in the discard slag. As in previous testwork, the cobalt in the discard slag when treating converter slag on its own was still in the region of 0.22%. However, by combining increasing amounts of furnace slag with converter slag, values approaching 0.08% Co in the discard slag could still be achieved. This resulted in cobalt recoveries approaching 90% being attained, while the average recovery was in the region of 85%.

The fume produced was of the order of 1% of the mass of the slag fed. The electrode consumption was 2 kg/MWh.

Copper reverberatory furnace slag - 200 kVA

During 1995, a campaign was conducted on Mintek's 200 kVA DC-arc furnace to recover cobalt from copper reverberatory furnace slag (initially containing Co:0.77% and Cu:1.3%). Alloys containing Co: 6-7%, Cu: 9-11%, Fe: 77-78%, and S: 2-3% were produced, leaving slags containing Co: 0.08-0.16% and Cu: 0.18-0.29%. Coal additions of 4 per cent of the mass of the slag fed, and residence times of one to two hours, gave cobalt recoveries ranging from 77 to 91%. Tapping temperatures of 1490-1560°C were attained. As shown in Figure 9, the cobalt level in the slag (and therefore the cobalt recovery) varied according to the residence time in the furnace. A retention time of 2 hours was required to achieve cobalt recoveries in the region of 90%.

Figure 10 - Kinetics of slag cleaning

During this campaign, the dust loss was 0.7 per cent of the mass of feed slag. The electrode consumption was 1.0 kg/MWh. The slag resistivity was calculated to be 1.1 cm. Measurements were also made of the arc characteristics. The melting point of the alloy produced in the furnace was 1340°C, as determined by differential thermal analysis.

Some of the alloy was upgraded by blowing with oxygen in a top-blown rotary converter (TBRC), preferentially oxidizing the iron, and thereby concentrating the cobalt and copper in the alloy. The iron content in the alloy was lowered from 76 to 25%, with the effect that the alloy was concentrated up to 30% cobalt and 40% copper. The sulphur level in the resulting alloy was 0.8%. This means that 75% of the cobalt remained in the alloy, while 90% of the iron was removed to the slag. The slag from the TBRC would, of course, be recycled back to the DC-arc furnace, in order to prevent the blown cobalt from being lost.

In the case of copper reverberatory furnace slag, mineralogical studies showed that the cobalt is present as CoO. Copper in the slag is mainly attributed to the presence of copper-rich sulphides. The cobalt oxide, and, to a lesser extent, the copper oxide associated with the silicate / oxide phases, is reduced by Fe from the alloy to form metallic Co (and Cu), resulting in the formation of FeO in the slag. Given that this reaction occurs between the metal bath and the overlying slag, the exchange of Co and Cu with Fe will take place only at the slag/metal interface. Improved recoveries of valuable metals can be achieved by allowing greater quantities of slag to come into contact with the alloy (by mild stirring, for example), and increasing the length of the contact time between slag and metal. The CoO in the slag is associated primarily with Fe2SiO4, and analysis by scanning electron microscopy showed some Fe2SiO4 particles with no detectable Co or Cu, thus demonstrating that it is, in principle, possible to remove all the Co and Cu from this phase.

The low concentrations of Co found in the sulphide phases of the alloy indicate that the possible reaction of CoO + FeS CoS + FeO is not significant for the cobalt in this system. (This reaction might, of course, be more significant in systems where the sulphur content of the alloy approaches that of a matte.)

Copper sulphide and copper metal (with its inherent cobalt content), because of their higher densities and immiscibility with the slag, settle from the molten slag into the metal bath. Improving recoveries via this mechanism would involve decreasing the viscosity of the slag (by increased temperature or flux additions), and / or increasing the settling time to allow smaller droplets to fall.


The question might well be asked as to why a DC-arc furnace should perform better than an AC slag-resistance furnace for this type of process. There are a number of reasons for this.

Metallurgical flexibility because of independent power supply

Since DC-arc furnaces operate under open arc conditions with the electrode positioned above the bath, they do not rely very much on the resistivity of the slag in order to supply energy to the furnace bath. This renders the energy supply nearly independent of slag composition, which allows the slag chemistry to be optimized for the best recovery of valuable metals (instead of for the required electrical characteristics). When this type of process is operated in an AC slag-resistance furnace, the degree of reduction cannot readily be controlled, because the electrodes are immersed in the slag, and alloys with high iron contents (relative to the amount of cobalt reduced) result (2,3).

Temperature control

As the current flowing in a DC-arc furnace has to travel through the entire depth of the liquid bath (as opposed to merely between the electrodes of an AC furnace), the temperature distribution is more likely to be relatively even. This is very important for determining factors such as slag viscosity (which is very temperature-sensitive) and density, which play a significant role in allowing the efficient settling of the droplets of alloy. Furthermore, it is possible to achieve a specified power input, almost without regard to the temperature of the slag. This is in marked contrast to the case of an AC furnace, where, as the slag gets hotter, the conductivity increases, thereby limiting the amount of energy which can be dissipated in the slag by the mechanism of resistive heating. This fact limits the temperature that can be obtained in a submerged-electrode or slag-resistance AC furnace. Furthermore, the high iron oxide content, of iron-silicate converter slags in particular, results in high electrical conductivity (9) which does not permit effective energy generation in the melt when using a slag-resistance furnace.

Stable operation

The inherent stability of a DC arc offers the potential for improved operational control. In an AC furnace, conditions under one electrode affect the currents in the other two electrodes. In the slag-cleaning process considered here, the furnace would be operated with a layer of electrically conductive coke or coal covering the slag. As the electrodes of the AC furnace penetrate this layer, there exists the possibility of the current flowing between the electrodes on the surface of the slag. This inter-electrode conduction not only reduces the energy dissipated in the slag layer, but would also result in difficulty in controlling electrode penetration of the slag.

Because of the difficulty of controlling the position of AC electrodes, there are often large imbalances in the power distribution. This results in hot zones, which reduce the freeze-line of solidified slag near the dissipating electrodes, exposing the refractories to slag attack. With a closely controlled single electrode, the freeze-lining on a DC-arc furnace can be better maintained, thus significantly reducing the slag attack on the refractories.

Electrode consumption and maintenance

AC furnaces used for slag cleaning are usually resistively heated three-phase three-electrode furnaces, where the electrodes are in contact with the slag. Unavoidably, the electrodes are attacked by the highly aggressive slag. Furthermore, electrodes carrying AC current suffer from the 'skin effect', where most of the electrode current is forced to a narrow outer band of the electrode, thereby reducing the electrode's current-carrying capacity. As a consequence of this, the electrodes have to be larger, thereby exposing an even greater surface to the reactive slag. The DC-arc furnace used for slag cleaning operates under open-arc conditions, with the electrode tip removed from contact with the aggressive slag. DC electrode current also eliminates the 'skin effect', allowing a greater current loading (A/cm2) of the electrode. This phenomenon, added to the fact that there is only one instead of three electrodes, results in a much smaller electrode erosion area. These design considerations result in a significantly lower electrode consumption, anticipated to be around 1 to 3 kg/MWh. Since the electrodes do not react with the slag, the reductant addition and consumption can be controlled accurately.

Because of the interactive nature of the electrodes in a three-phase furnace, a loss of power on one electrode will inevitably result is a loss of power in the other two. The problem is particularly prevalent when Söderberg-type electrodes are used, because such electrodes are prone to breakages, and also need to be baked in at low current (and hence low power) for extended periods of time after an upset. Losses of around 1% of total plant production are typically due to this type of problem.

AC furnaces with Söderberg-type electrodes are cumbersome to maintain, since electrode casings continually need to be welded to the three electrode columns, and the paste level in the electrodes monitored and maintained. On a DC furnace, one electrode section simply needs to be screwed on to the column at regular intervals (although this renders the furnace unavailable during the short power-off down-time).

Structural benefits

Because it has only one electrode to support, the superstructure of a DC-arc furnace is a lot simpler and cheaper than that required for a three-electrode AC furnace. The transformer and rectifier can be placed away from the superstructure, in a convenient location. Because there is only one electrode, there is room on the roof to accommodate the suitable positioning of feed chutes to allow the feeding of coke (or coal) directly onto the slag entering through the melt inlet. This facilitates good mixing between the slag and reductant. It is simpler to perform maintenance above a DC furnace electrode, since there is less electrode clutter above the roof. The gas seal on a DC furnace is better, as only one electrode seal, instead of three, is used. There is also usually less electrode movement, as the arc is more stable, and the electrode is not interacting with others.

Electrical power supply

The new generation of DC power supplies specifically addresses the problems of harmonics and flicker associated with AC power supplies. For example, the Robicon power supply uses a 24-pulse rectification system with an output chopped at a frequency of 1 kHz, and conforms to the IEEE 519 specification for both voltage and current distortion. This eliminates the need for costly harmonic filters, or for flicker-reducing static VAR compensators. An added benefit of the new-generation DC power supply is the fact that it is designed to provide low-reactance power at a power factor of 0.95 or better throughout the required voltage range, thereby reducing the transformer size and MVA demand, and eliminating the cost of any external power factor correction.

In some DC-arc furnaces operating at high voltages (and hence with long arcs) or high process temperatures, problems have been experienced with uncontrolled or 'stray' arcing between the electrode and the roof or refractories, resulting in damage to the roof (and water leaks in the case of a water-cooled roof). With a suitable roof design, and by maximizing the current / voltage capability of the power supply, the arc length can be reduced, and stray arcing minimized.


There are many aspects to the recovery process, and many of these can be tailored specifically to the particular material being treated.

Upgrading of cobalt-rich iron alloys

It is possible to upgrade the alloys produced in the DC-arc reduction process by the selective oxidation of iron (2,10). Figure 10 shows the flow of material streams in a process that allows the upgrading of cobalt-rich iron alloys. If the DC-arc furnace is run under strongly reducing conditions (to ensure the greatest possible recovery of valuable metals), the alloy will contain a large quantity of iron. However, if this alloy is subjected to a blowing stage (using air or oxygen), much of the iron can be eliminated into a slag phase. Typically, a silica flux is added in order to combine with the resulting iron oxide to form a fayalitic slag (nominally 2FeO.SiO2). This blowing stage can be carried out in a converter of some kind, or even in a closed furnace utilising a submerged top lance. Tests at Mintek have been done using a top-blown rotary converter (TBRC). The slag from the upgrading step would be recycled back to the DC-arc furnace at industrial scale, in order to recover the portion of valuable metals that have been re-oxidized. It is possible to arrange conditions in the two process units such that an alloy containing only 30% iron is produced, even though a cobalt recovery of 80% is achieved.

Figure 11 - Flowsheet for DC-arc slag cleaning, with an upgrading step, using a top-blown rotary converter (TBRC), for example

Objections to having a converter in the process usually revolve around the loss of sulphur to the atmosphere. Most modern non-ferrous smelters are moving away from Peirce-Smith converters for just that reason. However, there is a vast difference between the conversion of a high-sulphur matte and the blowing of an iron-rich alloy. The small quantity of sulphur present in the alloy (a small fraction of a per cent of the incoming sulphur from the concentrate to the smelter) is moderately tightly bound to the metallic elements, and is not too readily given up to the gas phase. Furthermore, the gas stream is only a waste product from a secondary stream much smaller than the main stream, and the very small quantity of sulphur in the gas could be scrubbed. If this is still seen to be a problem, a stationary enclosed vessel, such as those used by the new generation of submerged-lance systems, should be investigated for their suitability to the task.

Kinetics / mass transfer

The gas injection of solid reductant has been found to increase reaction and mass transfer rates. Stirring (using nitrogen injection, for example) can also be employed in order to improve the mass transfer particularly between slag and reductant.

Lime addition

The addition of CaO (up to a certain point) decreases the cobalt and copper content of slags (9). The activity coefficients of the valuable metal oxides are increased by the addition of CaO. In addition, the viscosity of the slag decreases as the slag becomes more basic (e.g. by the addition of CaO or MgO). This is important for slag-metal reaction kinetics, and for the settling behaviour of metal droplets. The liquidus temperature of the slag also decreases (to a minimum value) with increasing content of CaO.

It has been found (2) that the kinetics of reduction (for copper converter slag) are enhanced by the addition of lime, and good recoveries could be obtained in a one hour reduction period. It should also be mentioned that the addition of silica and magnesia can raise the liquidus temperature, thereby reducing the degree of attack on the furnace refractory lining.

Recovery of platinum group metals

The platinum group metals (PGMs) are often found together with nickel, copper, and cobalt sulphide deposits. Even in small quantities, these can be economically significant. The PGMs follow the nickel, copper, and iron through the pyrometallurgical process, and can be extracted from the hydrometallurgical leach residues for further processing.

Water atomization of alloys

Cobalt-rich iron alloys are virtually unbreakable, which poses a problem of delivery of the alloy to the downstream process units. It is common practice in a number of slag-cleaning processes to add sulphur (in the form of pyrite, concentrate, or matte) to the alloy, in order to make it sufficiently brittle to be able to be successfully milled after granulation. Apart from the inconvenience and expense of having to add this material to the furnace, this sulphur needs to be removed during subsequent hydrometallurgical processing.

Water atomization, involving the 'smashing' of a stream of molten alloy with a high-pressure stream of water, can directly produce fine particles of alloy with a mean diameter of less than 100 µm (even as small as 40 µm). The design of the atomizing system is simplified by not having any tight constraints on the range of particle sizes and shapes of the particles. Small-scale experiments have been carried out successfully on 5 kg batches of alloy. This technology is commercially available up to industrial scale, and appears to be very cost-effective when compared to the option of granulation and milling. This step introduces another level of flexibility into the process; one can now optimize the metallurgy to maximize recovery of the valuable metals, without needing to be too concerned with the physical properties of the alloy.

Very little material handling is required after atomization. A simple screen to separate oversize materials (for recycling to the DC-arc furnace) is about all that is needed. The alloy particles can be pumped as a slurry, then de-watered. Drying is not necessary, as the alloy will next be subjected to a wet process.

Downstream hydrometallurgical processing

The alloy produced in the reduction process can be leached with spent electrolyte from a copper or nickel electrowinning process. Mintek has developed a leaching process that solubilizes the nickel, copper, and cobalt, while rejecting the iron and any sulphur into the solid residue (by decreasing the pH appropriately). The resulting solution can be processed further to separate the cobalt, nickel, and copper from each other via conventional technology. Mintek has also developed direct solvent extraction of cobalt from cobalt-bearing nickel solutions. Nickel and cobalt have been successfully electrowon from the raffinate and strip liquor respectively.


The following furnace design specifications can be provided, once testwork has been carried out on the slag of interest.


DC-arc furnace technology has been successfully applied to the recovery of cobalt, nickel, and copper from non-ferrous furnace and converter slags. Pilot-plant testwork at Mintek has demonstrated recoveries of 98% for nickel and over 80% for cobalt, at power levels of up to 600 kW.

Mintek expects to participate in the commissioning of the first commercial units during 1998, and is currently involved with several engineering companies on final feasibility studies. Mintek intends to offer this patented technology (11) for implementation world-wide, and to play an active role in the design, supply, and commissioning of suitable DC-arc furnaces to meet clients' needs.


This paper is published by permission of Mintek. Contributions to this work were made by a number of individuals at Mintek. Particular mention should be made of the pioneering DC-arc smelting testwork of the late L.B. (Bruce) McRae. Thanks are also due to T.R. (Tom) Curr for many useful suggestions, to Dr A.S.E. (Arno) Kleyenstuber, Dr J. (Johan) Nell, A.D. (Alan) McKenzie, and S.D. (Steve) McCullough for mineralogical analyses, to Dr M.J. (Mike) Dry for a description of the downstream hydrometallurgical process, and to E. (Elana) Engelbrecht for the schematic diagram of the furnace. Helpful discussions with consulting metallurgists and other personnel from the industry are gratefully acknowledged.

  1. J.W. Matousek, "The Pyrometallurgical Winning of Cobalt", CIM Bulletin, Vol. 75, No. 848, December 1982, 121-127.

  2. R.M. Whyte, J.R. Orjans, G.B. Harris, and J.A. Thomas, "Development of a Process for the Recovery of Electrolytic Copper and Cobalt from Rokana Converter Slag", Advances in Extractive Metallurgy, Institute of Mining and Metallurgy, London, 1977, 57-68.

  3. S.C.C. Barnett and J.H.E. Jeffes, "Recovery of Nickel from Thompson Smelter Electric Furnace Slag", Transactions of the Institute of Mining and Metallurgy, Section C, Mineral Processes, Extractive Metallurgy, Vol. 86, 1977, 155-157.

  4. R.T. Jones, "Computer Simulation of Pyrometallurgical Processes", APCOM 87, Proceedings of the Twentieth International Symposium on the Application of Mathematics and Computers in the Minerals Industries, Volume 2: Metallurgy, SAIMM, Johannesburg, October 1987, 265-279.

  5. J.W. Hastie and D.W. Bonnell, "A Predictive Phase Equilibrium Model for Multicomponent Oxide Mixtures: Part II. Oxides of Na-K-Ca-Mg-Al-Si", High Temperature Science, Vol. 19, 1985, 275-306.

  6. S.S. Wang, N.H. Santander, and J.M. Toguri, "The Solubility of Nickel and Cobalt in Iron Silicate Slags", Metallurgical Transactions, Vol. 5, January 1974, 261-265.

  7. C.C. Banks and D.A. Harrison, "The Recovery of Non-ferrous Metals from Secondary Copper Smelter Discard Slags", Canadian Metallurgical Quarterly, Vol. 14, No. 2, 1975, 183-190.

  8. R.T. Jones, N.A. Barcza, and T.R. Curr, "Plasma Developments in Africa", Second International Plasma Symposium: World progress in plasma applications, EPRI (Electric Power Research Institute) CMP (Center for Materials Production), Palo Alto, California, 9-11 February 1993.

  9. A. Fontana, M. Ilunga, M. Mwema, L. Segers, K. Twite, and R. Winand, "Properties of Ferrous Silicate Slags Associated with Copper Flash Smelting and Electric Furnace Processes", Extraction Metallurgy '89, Institution of Mining and Metallurgy, London, 10-13 July 1989, 147-164.

  10. I.D. Reznik, "Flowsheets of Cobalt Recovery from Nickel and Copper - nickel Ores in Russia", The Paul E. Queneau International Symposium: Extractive Metallurgy of Copper, Nickel and Cobalt, Volume 1: Fundamental Aspects, The Minerals, Metals & Materials Society, U.S.A., 1993, 1293-1304.

  11. D.A. Hayman and G.M. Denton, "The Recovery of Metal Values from Slags", South African Patent, No. 95/4458, 31 May 1995.

Keywords: cobalt, copper, DC-arc, furnace, nickel, plasma, plasma-arc, process, pyrometallurgy, Pyrosim, recovery, reduction, slag, slag cleaning
International Symposium on Challenges of Process Intensification

35th Annual Conference of Metallurgists, Montreal, Canada, August 1996

Mintek Paper No. 8360


Copyright © 1996-2001 Mintek,,
19 June 2001