Influence of Desulfurization with Fe2O3 on the Reduction of Nickel Converter Slag

Generally in the nickel converter slag, metals are mainly in the form of sulfides, which are difficult to separate from slag. Although metal oxides in the slag, such as NiO, CoO, and Cu2O, are easily reduced into metal using carbon, the presence of sulfur inhibits the reduction reaction. In this study, the addition of Fe2O3 to nickel converter slag produced desulfurized slag, which enhanced the carbothermal reduction process. Increasing the desulfurization rate promoted the conversion of sulfides into oxides in slag, which significantly increased the activity of NiO, Cu2O, and Fe2O3. However, the residual sulfur content had no significant effect on the activity of FeO and CoO, due to the high initial FeO content and cobalt existing mainly in the form of oxides. The optimum addition of Fe2O3 was 15.0 g per 100 g nickel slag, while the desulfurization ratio was 36.84% and the rates of nikel, cobalt and copper recovery were 95.33%, 77.73%, and 73.83%, respectively.


Introduction
The process of oxidizing low-nickel matte using flowing air in a horizontal converter results in a large amount of slag containing nickel, cobalt, copper, and other metals. A subsequent process is used to clean the nickel slag in order to extract some of the nickel, cobalt, and copper [1][2][3]. The molten slag is poured into the melting pool of an electric furnace; then, a matte with high sulfur content is produced by adding coal and pyrite [4,5]. However, the treatment cycle is long and gives poor dilution results of residue copper up to 0.5-1.0% [6,7]. According to previous studies [8], the main components of nickel converter slag is FeO and SiO 2 , where Ni mainly exists as sulfides and some distributes in fayalite or composite oxides. Cobalt is mainly dispersed in fayalite and most of the copper exists as sulfides [9,10]. Therefore, in the nickel converter slag, the sulfides of Ni and copper result in the loss of Ni and Cu [11,12].
It is of great importance to separate the matte from the slag for the recovery of valuable metals and some key performance parameters determines the separation of matte and slag, including density, viscosity and interfacial tension. Various studies have been carried out to investigate the matte and slag's interfacial tension and density in alloy-slag system [13][14][15][16][17][18][19][20]. For molten slag systems, the density values of FeO x -SiO 2 -Al 2 O 3 [13], FeO x -SiO 2 -MgO [13] and FeO x -SiO 2 -CaO-Al 2 O 3 [14] are in the range of 3.68~3.69 g/cm 3 , 3.61~3.69 g/cm 3 and 2.794~2.836 g/cm 3 respectively. For molten matte systems, the density values of Fe-Ni-S [13], Ni-Cu-S [13] and Fe-Cu-Ni-S [15] are in the range of 3.82~5.18 g/cm 3 , 5.18~5.25 g/cm 3 and 3.92~5.59 g/cm 3 respectively. For molten metal systems, the density values of converter slag is dispersive and uniform. In this study, the remaining moisture was removed by drying the sample slag for 6 h at 110 • C. The slag was processed into powder and the ratio of powder that being less than 100 mesh is larger than 95%. Table 1 lists the chemical composition of nickel converter slag. The chemical compositions of the nickel converter slag are 3.51 wt.% Ni, 0.60 wt.% Co, 0.92 wt.% Cu, 1.53 wt.% S, 43.2 wt.% Total Fe, 52.8 wt.% FeO, 29.2 wt.% SiO 2 , 0.11 wt.% CaO, 0.83 wt.% MgO, and 0.48 wt.% Al 2 O 3 ; the ferric oxide (Fe 2 O 3 > 99.0 wt.%) was applied as the oxidizer; and carbon powder (C > 99.85 wt.%) was applied as the reducing agent. Figure 1 presents the thermal experiment flat platform used in this study, including a tubular furnace (Zhongkebeiyi, Beijing, China) with Ar protective gas. The sample was placed in an alumina crucible (ϕ40 mm × 90 mm), and held in an alumina safety crucible (ϕ60 mm × 150 mm) to protect the furnace tube from inner crucible failure. The furnace temperature was increased to 1000 • C at a rate of 10 • C/min. Then, the crucible containing the sample was placed in the tubular furnace at 1000 • C. The furnace was resealed and the furnace temperature was further increased from 1000 to 1300 • C at a rate of 5 • C/min; then the temperature was maintained at 1300 • C for desulfurization or reduction. The slag and metal obtained from the experiments were analyzed in Testing Center of University of Science and Technology Beijing (USTB).

Experiments
Materials 2020, 13, x FOR PEER REVIEW 3 of 11 that being less than 100 mesh is larger than 95%.  Figure 1. presents the thermal experiment flat platform used in this study, including a tubular furnace (Zhongkebeiyi, Beijing, China) with Ar protective gas. The sample was placed in an alumina crucible (φ40 mm × 90 mm), and held in an alumina safety crucible (φ60 mm × 150 mm) to protect the furnace tube from inner crucible failure. The furnace temperature was increased to 1000 °C at a rate of 10 °C/min. Then, the crucible containing the sample was placed in the tubular furnace at 1000 °C. The furnace was resealed and the furnace temperature was further increased from 1000 to 1300 °C at a rate of 5 °C/min; then the temperature was maintained at 1300 °C for desulfurization or reduction. The slag and metal obtained from the experiments were analyzed in Testing Center of University of Science and Technology Beijing (USTB).  Table 1 shows the experimental schemes for the desulphurization of initial nickel converter slag with Fe2O3. The Fe2O3 (0-40.0 g) and the nickel converter slag (100.0 g) were mixed in the desired ratios and these mixtures were held at 1300 °C for 30 min. Then, the molten slag was cooled in flowing Ar. Half of the cooled desulfurized slag was applied in the subsequent reduction experiments, while the remainder was used for analysis by a SPECTRO X-ray Fluorescence Spectrometer, SPECTRO Analytical Instruments, Kleve, Germany.   Table 1 shows the experimental schemes for the desulphurization of initial nickel converter slag with Fe 2 O 3 . The Fe 2 O 3 (0-40.0 g) and the nickel converter slag (100.0 g) were mixed in the desired ratios and these mixtures were held at 1300 • C for 30 min. Then, the molten slag was cooled in flowing Ar. Half of the cooled desulfurized slag was applied in the subsequent reduction experiments, while the remainder was used for analysis by a SPECTRO X-ray Fluorescence Spectrometer, SPECTRO Analytical Instruments, Kleve, Germany. Table 2 shows the experimental schemes for the reduction of the desulfurized slag with carbon powder. In this study, the carbon powder was added with the C/O ratio being at 0.2. The desulfurized slag (~50-70 g) and the carbon powder were mixed, and the mixture was heat treated at 1300 • C for 60 min. After reduction and separation, the sample was cooled in flowing argon. The metal formed in the reduction experiments were sent for analysis.  The FactSage 7.0 thermodynamics software (7.0, FactSage, GTT-Technologies, Herzogenrath, Germany) was used to characterize the chemical equilibrium of the reactions during the desulfurization process, and the activities of FeO, Fe 2 O 3 , NiO, CoO, and Cu 2 O during this process were calculated. According to previous research [37], the ratios of Ni, Co, and Cu existing as sulfide were 51.64%, 15.33%, and 90.08%, respectively. It was assumed that Equation (1) the ratios of Ni, Co, and Cu existing as sulfide were, respectively, 50%, 15%, and 90% and Equation (2) all other sulfur in the slag existed as FeS. The input data used for simulations with the Equilib module of FactSage were 2.23 wt.% NiO, 2.39 wt.% Ni 3 S 2 , 0.65 wt.% CoO, 0.14 wt.% CoS, 0.10 wt.% Cu 2 O, 1.04 wt.% Cu 2 S, 3.00 wt.% Fe 2 O 3 , 51.40 wt.% FeO, 1.76 wt.% FeS, 29.20 wt.% SiO 2 , and 8.09 wt.% Al 2 O 3 (The total slag weight was 100 g, and other substances in the slag were supplemented using Al 2 O 3 ).

Liquidus Temperature of Desulfurized Slag
In the nickel converter slag, the total content of FeO x and SiO 2 was larger than 80% and the liquidus temperature of slag was determined using FeO-Fe 2 O 3 -SiO 2 phase diagram [38][39][40][41]. Although the increase in the slag liquidus temperature had almost no influence on the desulfurization and reduction progress, it is helpful to separate the metal from the slag with the slag with a lower liquidus temperature [42]. According to the previous study [37], it can be known that the addition of Fe 2 O 3 influences the liquidus temperature of the slag significantly. The liquidus temperature of slag can be reduced when the Fe 2 O 3 addition being less than 6 wt.% and that would be increased when the Fe 2 O 3 addition in the range of 6-40 wt.%. According to the composition of total Fe and FeO in desulfurized slag in Table 2, the content of FeO, Fe 2 O 3 and SiO 2 can be calculated. The composition of the desulfurized slag in the ternary phase diagram of FeO-Fe 2 O 3 -SiO 2 is shown in Figure 2. With the addition of Fe 2 O 3 , the composition of slag moves towards the region with high liquidus temperature.

Equilibrium of Sulfide and Oxide in Slag
Although Fe2O3 has weaker oxidizing performance than O2, the added Fe2O3 can react with the sulfides (Ni3S2, CoS, Cu2S, and FeS) in the slag. The reactions are as follows: oxides. This agreed with previous reports [11,15]. The ratio of Fe 3+ /Fe 2+ is mainly influenced by the amount of added Fe2O3, while p SO2 is mainly influenced by the rate of SO2 release during the reaction. Figure 3 shows that the ratio of Fe 3+ /Fe 2+ in the slag measured experimentally increased with

Equilibrium of Sulfide and Oxide in Slag
Although Fe 2 O 3 has weaker oxidizing performance than O 2 , the added Fe 2 O 3 can react with the sulfides (Ni 3 S 2 , CoS, Cu 2 S, and FeS) in the slag. The reactions are as follows: This agreed with previous reports [11,15]. The ratio of Fe 3+ /Fe 2+ is mainly influenced by the amount of added Fe 2 O 3 , while p SO 2 is mainly influenced by the rate of SO 2 release during the reaction. Figure 3 shows that the ratio of Fe 3+ /Fe 2+ in the slag measured experimentally increased with increasing Fe 2 O 3 content, consistent with the theoretical calculations. In this paper, the equilibrium value of sulfur in slag has a balance relationship with SO 2 concentration in the gas phase. When Fe 2 O 3 was used for desulfurization reaction in the tubular furnace, the release rate of SO 2 is different under different conditions, but the flow rate of argon is not constant all the time, which will cause a great difference in SO 2 concentration in the gas phase. Therefore, results with different SO 2 partial pressure were calculated by Factsage calculation as shown in Figure 3. What's more, When more Fe 2 O 3 was added (30 g, 40 g), more Fe 2 O 3 will not participate in the reaction of sulfide oxidation, and the utilization rate of oxidant is low, so that the Fe 3+ remaining in slag is higher, causing the curve to deviate from the calculated value of Factsage calculation.
Materials 2020, 13, x FOR PEER REVIEW 6 of 11 increasing Fe2O3 content, consistent with the theoretical calculations. In this paper, the equilibrium value of sulfur in slag has a balance relationship with SO2 concentration in the gas phase. When Fe2O3 was used for desulfurization reaction in the tubular furnace, the release rate of SO2 is different under different conditions, but the flow rate of argon is not constant all the time, which will cause a great difference in SO2 concentration in the gas phase. Therefore, results with different SO2 partial pressure were calculated by Factsage calculation as shown in Figure 3. What's more, When more Fe2O3 was added (30 g, 40 g), more Fe2O3 will not participate in the reaction of sulfide oxidation, and the utilization rate of oxidant is low, so that the Fe 3+ remaining in slag is higher, causing the curve to deviate from the calculated value of Factsage calculation.

Desulphurization Rate
In order to compare the influence of Fe2O3 addition on sulfur removal, the desulfurization rate was defined as: where φ is the desulfurization rate (%), m is the mass of the slag after the desulfurization (g), m0 is the mass of initial nickel converter slag (g), w(S) is the mass fraction of sulfur in the desulfurized slag (%), and w0(S) is the mass fraction of sulfur in the initial nickel converter slag (%). Figure 4 shows the relationship between the added Fe2O3 and the desulfurization rate for both experimental and theoretical results. When a small amount of Fe2O3 was added (0-5 wt.%), the desulfurization rate was relatively low after reaction. In this case, the initial Fe2O3 content in the slag was low, and most of added Fe2O3 was dissolved into the slag instead of participating in the desulfurization reaction. When 10-20 wt.% Fe2O3 was added, the slag was strongly oxidized, where the sulfur in the slag was oxidized to SO2. Hence, the desulfurization rate increased significantly with the addition of Fe2O3, as shown in Figure 4. With further addition of Fe2O3 (20-40 wt.%), the rate of increase of the desulfurization rate decreased. Under these conditions, the sample could not reach equilibrium within the limited time of the experiment due to the high viscosity of the slag. Therefore, the desulfurization rate increased slowly and inadequate contact between S 2− and Fe2O3 in the slag was the main limiting factor for the desulfurization reaction [43]. In addition, the SO2 concentration of the product in the reaction process will affect the sulfur content of the equilibrium slag, while the SO2 concentration in the gas phase will be different due to the different SO2 release rate in the oxidation desulfurization reaction during the experiment process, which may affect the different desulfurization process. The Factsage calculation can quantify and compare the equilibrium value under different SO2 partial pressure with the test results.

Desulphurization Rate
In order to compare the influence of Fe 2 O 3 addition on sulfur removal, the desulfurization rate was defined as: where ϕ is the desulfurization rate (%), m is the mass of the slag after the desulfurization (g), m 0 is the mass of initial nickel converter slag (g), w(S) is the mass fraction of sulfur in the desulfurized slag (%), and w 0 (S) is the mass fraction of sulfur in the initial nickel converter slag (%). Figure 4 shows the relationship between the added Fe 2 O 3 and the desulfurization rate for both experimental and theoretical results. When a small amount of Fe 2 O 3 was added (0-5 wt.%), the desulfurization rate was relatively low after reaction. In this case, the initial Fe 2 O 3 content in the slag was low, and most of added Fe 2 O 3 was dissolved into the slag instead of participating in the desulfurization reaction. When 10-20 wt.% Fe 2 O 3 was added, the slag was strongly oxidized, where the sulfur in the slag was oxidized to SO 2 . Hence, the desulfurization rate increased significantly with the addition of Fe 2 O 3 , as shown in Figure 4. With further addition of Fe 2 O 3 (20-40 wt.%), the rate of increase of the desulfurization rate decreased. Under these conditions, the sample could not reach equilibrium within the limited time of the experiment due to the high viscosity of the slag. Therefore, the desulfurization rate increased slowly and inadequate contact between S 2− and Fe 2 O 3 in the slag was the main limiting factor for the desulfurization reaction [43]. In addition, the SO 2 concentration of the product in the reaction process will affect the sulfur content of the equilibrium slag, while the SO 2 concentration in the gas phase will be different due to the different SO 2 release rate in the oxidation desulfurization reaction during the experiment process, which may affect the different desulfurization

Activity of Metal Oxide in Slag
Ni, Co, and Cu mainly existed as NiO, CoO, Cu2O, Ni3S2, CoS, and Cu2S in the slag. The gaseous products of oxides reduced by carbon are mainly CO, while the gaseous products of sulfide reduction are mainly CS2 [38]. According to our previous study [37], the Standard Gibbs free energy of relative reactions have been obtianed from thermodynamic modeling using FactSage. Considering the change in the Gibbs free energy, carbon reacts most readily, first with metal oxides and then with sulfides, in the order of Cu2O, NiO, CoO, FeO, Fe2SiO4, CoS, Ni3S2, FeS, and Cu2S. The ability to oxidize the slag has an equilibrium relationship with the sulfur content in the desulfurized slag, which further influences the ratio of metal sulfides to oxides. According to our previous study [37], the predominant reactions were the reduction of metal oxides by carbon and the increases in the activities of FeO, Fe2O3, NiO, CoO, and Cu2O in the slag could promote the reduction reaction.
As shown in Figure 5a, the activities of FeO, Fe2O3, NiO, CoO, and Cu2O in desulfurized slag were exported from FactSage to analyze the influence of desulfurization rate on the activity of metal oxide in the desulfurized slag. With the addition of Fe2O3, the sulfur content decreased and the activities of FeO, Fe2O3, NiO, and Cu2O increased, while the activity of CoO was almost not affected by the increase of Fe2O3 addition.

Activity of Metal Oxide in Slag
Ni, Co, and Cu mainly existed as NiO, CoO, Cu 2 O, Ni 3 S 2 , CoS, and Cu 2 S in the slag. The gaseous products of oxides reduced by carbon are mainly CO, while the gaseous products of sulfide reduction are mainly CS 2 [38]. According to our previous study [37], the Standard Gibbs free energy of relative reactions have been obtianed from thermodynamic modeling using FactSage. Considering the change in the Gibbs free energy, carbon reacts most readily, first with metal oxides and then with sulfides, in the order of Cu 2 O, NiO, CoO, FeO, Fe 2 SiO 4 , CoS, Ni 3 S 2 , FeS, and Cu 2 S. The ability to oxidize the slag has an equilibrium relationship with the sulfur content in the desulfurized slag, which further influences the ratio of metal sulfides to oxides. According to our previous study [37], the predominant reactions were the reduction of metal oxides by carbon and the increases in the activities of FeO, Fe 2 O 3 , NiO, CoO, and Cu 2 O in the slag could promote the reduction reaction.
As shown in Figure 5a, the activities of FeO, Fe 2 O 3 , NiO, CoO, and Cu 2 O in desulfurized slag were exported from FactSage to analyze the influence of desulfurization rate on the activity of metal oxide in the desulfurized slag. With the addition of Fe 2 O 3 , the sulfur content decreased and the activities of FeO, Fe 2 O 3 , NiO, and Cu 2 O increased, while the activity of CoO was almost not affected by the increase of Fe 2 O 3 addition.
In order to compare the relative changes in activity, the activity of the oxides changes with the amount of Fe 2 O 3 addition was defined as: where a is the activity of oxides in Experiment No. 1-7; a 1 is the activity of the oxides after desulfurization in Experiment No. 1.
With the addition of Fe 2 O 3 , the increase of activity in Cu 2 O, NiO, and Fe 2 O 3 were significantly, while the increase of activity in FeO and CoO were not. Although Fe 2 O 3 was added, the most notable increase in activity was for Cu 2 O, rather than Fe 2 O 3 or FeO, according to Figure 5b. The added Fe 2 O 3 reacted with sulfides, such as Ni 3 S 2 , CoS, and Cu 2 S, where the sulfur was removed in the form of SO 2 . After sulfur removal from the slag, sulfides in the slag were converted into oxides, increasing the amount (and activity) of metal oxides in the slag. According to the analysis of the existing forms of nickel, cobalt, and copper in nickel slag [44], copper exists mainly in the form of sulfide, which makes sulfur removal have the most significant effect on copper oxide and sulfide in the slag. However, cobalt mainly exists in the form of oxide, and its activity was barely affected by sulfur in the slag. Although most of the added Fe 2 O 3 was oxidized to FeO, the Fe 2+ in the slag increased, while Fe 3+ /Fe 2+ increased simultaneously. Because of the high initial FeO content, the increase in Fe 3+ fraction was larger, i.e., ∆Fe 3+ /Fe 3+ was greater than ∆Fe 2+ /Fe 2+ (∆Fe 3+ and ∆Fe 2+ are the increase in the Fe 3+ and Fe 2+ ion concentrations in the slag, respectively). This resulted in a larger rate of change in activity for Fe 2 O 3 than FeO, as shown in Figure 5b. reactions have been obtianed from thermodynamic modeling using FactSage. Considering the change in the Gibbs free energy, carbon reacts most readily, first with metal oxides and then with sulfides, in the order of Cu2O, NiO, CoO, FeO, Fe2SiO4, CoS, Ni3S2, FeS, and Cu2S. The ability to oxidize the slag has an equilibrium relationship with the sulfur content in the desulfurized slag, which further influences the ratio of metal sulfides to oxides. According to our previous study [37], the predominant reactions were the reduction of metal oxides by carbon and the increases in the activities of FeO, Fe2O3, NiO, CoO, and Cu2O in the slag could promote the reduction reaction.
As shown in Figure 5a, the activities of FeO, Fe2O3, NiO, CoO, and Cu2O in desulfurized slag were exported from FactSage to analyze the influence of desulfurization rate on the activity of metal oxide in the desulfurized slag. With the addition of Fe2O3, the sulfur content decreased and the activities of FeO, Fe2O3, NiO, and Cu2O increased, while the activity of CoO was almost not affected by the increase of Fe2O3 addition.   Figure 6 shows the Ni, Co and Cu recovery rates change with different addition of Fe 2 O 3 . With the Fe 2 O 3 addition increased in the range of 0-15.0 g, the metals recovery rates increased. However, with the addition of Fe 2 O 3 further increasing from 15.0 to 40.0 g, the metals recovery rate declined sharply. According to the phase diagram of FeO-Fe 2 O 3 -SiO 2 , the slag will melt very well at 1500 • C, while the phase diagram of Fe-Ni-Co-Cu-S shows that the decrease of sulfur content will make the liquidus temperature of metal increase significantly [45][46][47][48][49]. With the increase of Fe 2 O 3 addition, the sulfur content in metal decreased, causing the high liquidus temperature of metal and poor separation of metal from slag. Therefore, although the removal of sulfur from slag promotes the reduction of oxides, the liquidus temperature of metal obtained increases simultaneously. The optimum addition of Fe 2 O 3 was 15.0 g, while the desulfurization rate was 36.82% and the Ni, Co and Cu recovery rate were 95.33%, 77.73%, and 73.83%, respectively.

Recovery of Nickel, Cobalt, and Copper
Materials 2020, 13, x FOR PEER REVIEW 8 of 11 In order to compare the relative changes in activity, the activity of the oxides changes with the amount of Fe2O3 addition was defined as: where a is the activity of oxides in Experiment No. 1-7; a1 is the activity of the oxides after desulfurization in Experiment No. 1.
With the addition of Fe2O3, the increase of activity in Cu2O, NiO, and Fe2O3 were significantly, while the increase of activity in FeO and CoO were not. Although Fe2O3 was added, the most notable increase in activity was for Cu2O, rather than Fe2O3 or FeO, according to Figure 5b. The added Fe2O3 reacted with sulfides, such as Ni3S2, CoS, and Cu2S, where the sulfur was removed in the form of SO2. After sulfur removal from the slag, sulfides in the slag were converted into oxides, increasing the amount (and activity) of metal oxides in the slag. According to the analysis of the existing forms of nickel, cobalt, and copper in nickel slag [44], copper exists mainly in the form of sulfide, which makes sulfur removal have the most significant effect on copper oxide and sulfide in the slag. However, cobalt mainly exists in the form of oxide, and its activity was barely affected by sulfur in the slag. Although most of the added Fe2O3 was oxidized to FeO, the Fe 2+ in the slag increased, while Fe 3+ /Fe 2+ increased simultaneously. Because of the high initial FeO content, the increase in Fe 3+ fraction was larger, i.e., △Fe 3+ /Fe 3+ was greater than △Fe 2+ /Fe 2+ (△Fe 3+ and △Fe 2+ are the increase in the Fe 3+ and Fe 2+ ion concentrations in the slag, respectively). This resulted in a larger rate of change in activity for Fe2O3 than FeO, as shown in Figure 5b. Figure 6 shows the Ni, Co and Cu recovery rates change with different addition of Fe2O3. With the Fe2O3 addition increased in the range of 0-15.0 g, the metals recovery rates increased. However, with the addition of Fe2O3 further increasing from 15.0 to 40.0 g, the metals recovery rate declined sharply. According to the phase diagram of FeO-Fe2O3-SiO2, the slag will melt very well at 1500 °C, while the phase diagram of Fe-Ni-Co-Cu-S shows that the decrease of sulfur content will make the liquidus temperature of metal increase significantly [45][46][47][48][49]. With the increase of Fe2O3 addition, the sulfur content in metal decreased, causing the high liquidus temperature of metal and poor separation of metal from slag. Therefore, although the removal of sulfur from slag promotes the reduction of oxides, the liquidus temperature of metal obtained increases simultaneously. The optimum addition of Fe2O3 was 15.0 g, while the desulfurization rate was 36.82% and the Ni, Co and Cu recovery rate were 95.33%, 77.73%, and 73.83%, respectively.

Conclusions
The Fe 2 O 3 was used to remove sulfur firstly in the recovery of Ni, Co and Cu from the nickel converter slag as a weak oxidant due to the peroxidation of the molten slag with blowing air in the smelting oxidation-reduction process of nickel converter slag. The theoretical analysis and tubular furnace experiments were carried out to investigate the effect of desulfurization with Fe 2 O 3 on the production process. The main conclusions are as following: (1) the liquidus temperature decreased with the addition of Fe 2 O 3 being in the range of 0-6 wt.%. however, the liquidus temperature began to increase for a Fe 2 O 3 content of 6-40 wt.%. In addition, the desulfurization rate increased with the addition of Fe 2 O 3 , where the growth was the most significant for Fe 2 O 3 contents of 5-20%. (2) According to the FactSage calculations, a higher desulfurization rate of the slag promoted conversion of sulfides into oxides, which significantly increased the activities of Cu 2 O, NiO, and Fe 2 O 3 in the slag. However, activity of CoO and FeO was not significantly affected. (3) Although the removal of sulfur from slag promotes the carbothermic reduction of oxides, the liquidus temperature of metal obtained increases simultaneously. The optimum addition of Fe 2 O 3 was 15.0 g, while the desulfurization rate was 36.82% and the metal recovery rate of Ni, Co, and Cu were 95.33%, 77.73%, and 73.83%, respectively.