Thermodynamic and Experimental Study on Efficient Extraction of Valuable Metals from Polymetallic Nodules

: Polymetallic nodules are promising resources for the extraction of valuable metals such as copper, nickel, and cobalt, as well as manganese alloys. To achieve e ﬃ cient extraction of useful metals from the emerging resource, high-temperature carbothermic reduction of nodules was investigated by optimizing the reductant addition, slag and alloy systems. Thermochemical software FactSage was used to predict the liquidus temperature of the slag system, which is not sensitive to FeO, CaO and Al 2 O 3 , but decreases signiﬁcantly with decreasing MnO / SiO 2 mass ratio. The experiments were designed to reduce the oxides of Cu, Co and Ni completely, and reduce FeO x partially depending on the amount of graphite addition while leaving the residual slag for further processing into ferromanganese and / or silicomanganese alloys. Co, Cu and Ni concentrations in the alloy decreased with increasing graphite addition. The optimal reduction condition was reached by adding 4 wt% graphite at the MnO / SiO 2 mass ratio of 1.6 in slag. The most e ﬀ ective metal-slag separation was achieved at 1350 ◦ C, which enables the smelting reduction to be carried out in various furnaces.


Introduction
The manganese and iron concretions on the sea bottom formed of concentric layers around a core are known as manganese nodules or polymetallic nodules. Polymetallic nodules contain many valuable elements such as Cu, Ni and Co, besides Mn and Fe. Trace amounts of Mo, V, and rare earth elements (REEs) are also present in the nodules. As the land-based resources for Mn, Cu, Ni and Co are depleting very fast, the exploration and exploitation of manganese nodules are essential to meet the future demands of Cu, Ni, Co and Mn in the world [1]. In recent years, the exploitation of manganese nodules from the deep sea has once again become a political and economic topic [2,3]. One of the biggest and economically most important areas, where polymetallic nodules can be found, is situated between the Clarion and Clipperton Zones in the tropical northern Pacific Ocean. The nodules could serve as an enormous reserve for these metals since they are found in relative abundance. The nodules in the Clarion Clipperton Zone alone constitute an estimated reserve of up to 220 million tons of copper, up to 260 million tons of nickel, up to 50 million tons of cobalt, and up to 5 billion tons of manganese [2,3]. These estimated amounts currently surpass universal land-based reserves in mines for Ni, Cu, Co and Mn. Recently, the increased interest in electrical energy storage and rapid growth in the market for electric vehicles will make the polymetallic nodules promising resources to produce component metals of batteries.
Polymetallic nodules are composed of complex Mn-oxides and Fe-oxyhydroxide minerals with mineral particles smaller than one µm. Ni, Cu, Co and Mo are found in the manganese-or iron-oxide matrix [4]. Therefore, beneficiation with classic techniques (density separation, floatation, etc.) to The microstructure and elemental mapping of a local area of an example nodule were determined by scanning electron microscopy (SEM) with energy dispersive X-ray spectrometry (EDS), which is shown in Figure 1. The nodule consists of concentrically banded zones of micro-layers around a nucleus. The microstructure of the shell is dense, while it is looser in the core of the nodule scanned. The maps of the distribution of each element across the local area show the distribution of Mn and Fe is in striation. Si is incorporated with Al, Na, K and Ca as impurity minerals, likely detrital deep-sea ooze [1]. Valuable metal elements such as Cu, Co and Ni in low concentrations are uniformly distributed in the nodule matrix. Ba is found to be associated with S, which is most likely considered to be precipitated barite (BaSO 4 ). Minerals 2020, 10, x FOR PEER REVIEW 3 of 15

Effect of Flux on the Liquidus Temperature of Resulted Slags
Thermodynamically, it is possible to reduce the oxides of nickel, copper, cobalt, and iron to metals while leaving the manganese oxide over a wide range of temperatures. The selective reduction of NiO, CuO, CoO and Fe2O3 over MnO by carbon reductant can be simplified in Reactions (1)-(4) with a ratio of CO and CO2 gases between critical limits. The iron oxide can be fully or partially reduced in the smelting process to adjust the slag composition and to control the alloy grade. For the sake of convenience, the stoichiometric amount of carbon needed to reduce the oxides is 2.4 g per 100 g ore, assuming only CO is generated during carbon gasification instead of a mixer of CO/CO2.
In similar commercial scaled operations, some common oxides such as CaO, SiO2, and Al2O3 are normally added as a flux to adjust the slag composition and keep the smelting slag in a fully molten state at a suitable temperature. The thermodynamic software package FactSage 7.2 [32] with the suitable optimized compounds and solution databases was used to predict the liquidus temperatures of the MnO-SiO2-"FeO"-Al2O3-MgO-Na2O-CaO-K2O system at iron saturation condition based on the normalization of major components of Table 1 and assumption of full reduction of NiO, CuO and CoO. The composition of the slag used for the FactSage calculation is given in Table 2.
The FactSage databases including FactPS (for pure compounds), FToxide (for solid and liquid oxide solutions), and FTmisc (for liquid alloys), were used. The effects of the MnO/SiO2 ratio and CaO, Al2O3, and "FeO" on the liquidus temperatures of the slag are shown in Figure 2. It can be seen from Figure 2a that the liquidus temperature increases with increasing CaO concentration in the monoxide primary phase field at different MnO/SiO2 ratios but decreases in the olivine primary phase

Effect of Flux on the Liquidus Temperature of Resulted Slags
Thermodynamically, it is possible to reduce the oxides of nickel, copper, cobalt, and iron to metals while leaving the manganese oxide over a wide range of temperatures. The selective reduction of NiO, CuO, CoO and Fe 2 O 3 over MnO by carbon reductant can be simplified in Reactions (1)-(4) with a ratio of CO and CO 2 gases between critical limits. The iron oxide can be fully or partially reduced in the smelting process to adjust the slag composition and to control the alloy grade. For the sake of convenience, the stoichiometric amount of carbon needed to reduce the oxides is 2.4 g per 100 g ore, assuming only CO is generated during carbon gasification instead of a mixer of CO/CO 2 . CoO In similar commercial scaled operations, some common oxides such as CaO, SiO 2, and Al 2 O 3 are normally added as a flux to adjust the slag composition and keep the smelting slag in a fully molten state at a suitable temperature. The thermodynamic software package FactSage 7.2 [32] with the suitable optimized compounds and solution databases was used to predict the liquidus temperatures of the MnO-SiO 2 -"FeO"-Al 2 O 3 -MgO-Na 2 O-CaO-K 2 O system at iron saturation condition based on the normalization of major components of Table 1 and assumption of full reduction of NiO, CuO and CoO. The composition of the slag used for the FactSage calculation is given in Table 2.
The FactSage databases including FactPS (for pure compounds), FToxide (for solid and liquid oxide solutions), and FTmisc (for liquid alloys), were used. The effects of the MnO/SiO 2 ratio and CaO, Al 2 O 3 , and "FeO" on the liquidus temperatures of the slag are shown in Figure 2. It can be seen from Figure 2a that the liquidus temperature increases with increasing CaO concentration in the monoxide Minerals 2020, 10, 360 4 of 14 primary phase field at different MnO/SiO 2 ratios but decreases in the olivine primary phase field at the MnO/SiO 2 ratio of 1.0. Figure 2b shows the Al 2 O 3 concentration has a little effect on the liquidus temperature at the MnO/SiO 2 ratio of 3.0. The liquidus temperature increases slightly with increasing Al 2 O 3 concentration at the MnO/SiO 2 ratio of 2.0 in the monoxide primary phase field but decreases slightly at the MnO/SiO 2 ratio of 1.0 in the olivine primary phase field. It shows in Figure 2c that the FeO concentration has little effect on the liquidus temperature at different MnO/SiO 2 ratios, which indicates the liquidus temperature of slag will not change significantly when FeO in slag is partially or fully reduced.
It is clear from Figure 2 that the MnO/SiO 2 ratio has a significant impact on the primary phase field and liquidus temperature. SiO 2 can be added as a flux to control the liquidus temperature of the slag. The effect of SiO 2 or MnO/SiO 2 on the liquidus temperature is shown in Figure 3. It can be seen that the autogenic residue without SiO2 flux has a liquidus temperature of around 1570 • C, which requires an operation temperature of above 1600 • C to keep the slag fully liquid. By adding SiO 2 flux, the primary phase field changes from mono-oxide to olivine 2(Ca, Mn)O·SiO 2 and rhodonite (Ca, Mn)O·SiO 2 . A large operating window with the liquidus temperature below 1300 • C can be obtained with 10 to 50 wt% SiO 2 addition.
Minerals 2020, 10, x FOR PEER REVIEW 4 of 15 field at the MnO/SiO2 ratio of 1.0. Figure 2b shows the Al2O3 concentration has a little effect on the liquidus temperature at the MnO/SiO2 ratio of 3.0. The liquidus temperature increases slightly with increasing Al2O3 concentration at the MnO/SiO2 ratio of 2.0 in the monoxide primary phase field but decreases slightly at the MnO/SiO2 ratio of 1.0 in the olivine primary phase field. It shows in Figure  2c that the FeO concentration has little effect on the liquidus temperature at different MnO/SiO2 ratios, which indicates the liquidus temperature of slag will not change significantly when FeO in slag is partially or fully reduced. It is clear from Figure 2 that the MnO/SiO2 ratio has a significant impact on the primary phase field and liquidus temperature. SiO2 can be added as a flux to control the liquidus temperature of the slag. The effect of SiO2 or MnO/SiO2 on the liquidus temperature is shown in Figure 3. It can be seen that the autogenic residue without SiO2 flux has a liquidus temperature of around 1570 °C, which requires an operation temperature of above 1600 °C to keep the slag fully liquid. By adding SiO2 flux, the primary phase field changes from mono-oxide to olivine 2(Ca, Mn)O·SiO2 and rhodonite (Ca, Mn)O·SiO2. A large operating window with the liquidus temperature below 1300 °C can be obtained with 10 to 50 wt% SiO2 addition.

Effect of Carbon Addition on the Reduction Degree
The effect of carbon addition on the extension of the reduction was examined by FactSage 7.2 with the SiO2 addition of 10 wt% of ore weight at 1350 °C. The recovery rate and alloy composition as a function of carbon addition are shown in Figure 4. It can be seen from Figure 4 that the NiO and CuO are first reduced and Ni and Cu concentrations in the alloy reached a maximum. With increasing carbon addition, FeOx and CoO are reduced and their concentrations in the alloy gradually increase. At 1.2 wt% carbon addition, all NiO, CuO and CoO are reduced and the alloy composition is 50Ni-12Cu-6Co-32Fe. Further addition of carbon will continuously reduce FeOx and increase Fe in the alloy. Accordingly, the concentrations of Ni, Co and Cu will decrease as the total alloy increases. At carbon 2.5 wt%, all FeOx is reduced and Fe in the alloy reaches maximum. MnO starts to be reduced and carbon starts to dissolve into the alloy. Further addition of carbon will increase Mn and carbon concentrations in the alloy, and decrease the concentrations of other components in the alloy. Over 6.5 wt% carbon addition, the alloy composition will be constant, which is considered as the critical

Effect of Carbon Addition on the Reduction Degree
The effect of carbon addition on the extension of the reduction was examined by FactSage 7.2 with the SiO 2 addition of 10 wt% of ore weight at 1350 • C. The recovery rate and alloy composition as a function of carbon addition are shown in Figure 4. It can be seen from Figure 4 that the NiO and CuO are first reduced and Ni and Cu concentrations in the alloy reached a maximum. With increasing carbon addition, FeO x and CoO are reduced and their concentrations in the alloy gradually increase. At 1.2 wt% carbon addition, all NiO, CuO and CoO are reduced and the alloy composition is 50Ni-12Cu-6Co-32Fe. Further addition of carbon will continuously reduce FeO x and increase Fe in the alloy. Accordingly, the concentrations of Ni, Co and Cu will decrease as the total alloy increases. At carbon 2.5 wt%, all FeO x is reduced and Fe in the alloy reaches maximum. MnO starts to be reduced and carbon starts to dissolve into the alloy. Further addition of carbon will increase Mn and carbon concentrations in the alloy, and decrease the concentrations of other components in the alloy. Over 6.5 wt% carbon addition, the alloy composition will be constant, which is considered as the critical amount of graphite. Thus, adding graphite greater than 6.5 wt% would not change the composition of alloy significantly.
Minerals 2020, 10, x FOR PEER REVIEW 6 of 15 amount of graphite. Thus, adding graphite greater than 6.5 wt% would not change the composition of alloy significantly.

Experimental Procedure
A vertical tube furnace with lanthanum chromate heating elements was used for the experiments. For each experimental run, 10 g of nodules were crushed into powder, and well mixed with flux and reductant in an agate mortar then pressed into briquette. Pure graphite powder (99.99 wt%) was used as the reductant. The flux SiO2 and reductant graphite were added as per ore weight. The addition of SiO2 was adjusted based on the MnO/SiO2 ratios and the liquidus temperatures showing in Figure 3. The experimental conditions are shown in Table 3. The pelletized sample was then placed into an alumina crucible (ID: 18 mm, H: 25 mm). The sample was heated and pre-reduced under an Ar atmosphere at 1000 °C for 60 min. Then, each pre-reduced sample was heated immediately to the smelting temperature and held there for 60 min. A Pt-30% Rh/Pt-6% Rh thermocouple placed in an alumina sheath was located adjacent to the sample to monitor the temperature. The temperature was controlled to an accuracy of ±2 K. The samples were directly quenched into the water after the smelting reduction. After drying, they were mounted in resin and polished for metallographic analysis. The microstructures were examined by scanning electron microscopy coupled with x-ray energy-dispersive spectroscopy analysis (SEM-EDS). Compositions of the liquid and solid phases were measured by electron probe X-ray microanalysis (EPMA). The EPMA operation was conducted at an accelerating voltage of 15 kV and a probe current of 15 nA. The standards used for analysis were spinel (MgAl2O4) for Al and Mg, wollastonite (CaSiO3) for Ca and Si, chalcopyrite (FeCuS2) for Cu, rutile (TiO2) for Ti, chromite (FeCrO4) for Cr, albite (NaAlSi3O8) for Na, orthoclase (KAlSi3O8) for K, olivine (Ni2SiO4) for Ni, spessartine (Mn3Al2Si3O12) for Mn, hematite (Fe2O3) for Fe, and pure Co for Co. The ZAF correction procedure, which is a series of multiplicative factors that account for the effects of atomic number (Z-stopping power, back-scattering factor and X-ray production power), absorption (A) and fluorescence (F), built-in the electron-probe software was applied automatically to the matrix correction. The average accuracy of the EPMA measurements is within ±1 wt%. The alloy droplets were collected to be measured by Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES).

Experimental Procedure
A vertical tube furnace with lanthanum chromate heating elements was used for the experiments. For each experimental run, 10 g of nodules were crushed into powder, and well mixed with flux and reductant in an agate mortar then pressed into briquette. Pure graphite powder (99.99 wt%) was used as the reductant. The flux SiO 2 and reductant graphite were added as per ore weight. The addition of SiO 2 was adjusted based on the MnO/SiO 2 ratios and the liquidus temperatures showing in Figure 3. The experimental conditions are shown in Table 3. The pelletized sample was then placed into an alumina crucible (ID: 18 mm, H: 25 mm). The sample was heated and pre-reduced under an Ar atmosphere at 1000 • C for 60 min. Then, each pre-reduced sample was heated immediately to the smelting temperature and held there for 60 min. A Pt-30% Rh/Pt-6% Rh thermocouple placed in an alumina sheath was located adjacent to the sample to monitor the temperature. The temperature was controlled to an accuracy of ±2 K. The samples were directly quenched into the water after the smelting reduction. After drying, they were mounted in resin and polished for metallographic analysis. The microstructures were examined by scanning electron microscopy coupled with X-ray energy-dispersive spectroscopy analysis (SEM-EDS). Compositions of the liquid and solid phases were measured by electron probe X-ray microanalysis (EPMA). The EPMA operation was conducted at an accelerating voltage of 15 kV and a probe current of 15  for Fe, and pure Co for Co. The ZAF correction procedure, which is a series of multiplicative factors that account for the effects of atomic number (Z-stopping power, back-scattering factor and X-ray production power), absorption (A) and fluorescence (F), built-in the electron-probe software was applied automatically to the matrix correction. The average accuracy of the EPMA measurements is within ±1 wt%. The alloy droplets were collected to be measured by Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES).

Pre-Reduction
Pre-reduction was conducted to calcinate nodules and enable nickel, copper, cobalt and iron oxides to be partly reduced. The microstructures of samples ND2 and ND9 after pre-reduction at 1000 • C for 60 min are shown in Figure 5. In Figure 5a

Effect of Reductant Addition on the Alloys and Slags
The smelting process was conducted at high temperatures to further reduce nickel, copper, cobalt, and iron oxides to form the required alloys and slags. Figure 6 shows the pictures and microstructures of samples ND8, ND9, and ND10 after smelting at 1350 • C with a carbon addition of 2, 4, and 6 wt%, respectively. It can be seen from Figure 6a that in sample ND8, the slag contains spinel crystals and the alloy is not observed. In samples ND9 and ND10, the slag is fully liquid and alloy sphere (around 5 mm in diameter) settled at the bottom of the crucible can be easily separated from the slag, and small alloy droplets are also present in the slag. A small amount of graphite was left after smelting reduction in sample ND10, which indicates the addition of graphite was in excess. Comparing Figure 6a-c, when the addition of the reductant was insufficient, the liquidus temperature of slag was much higher due to the presence of Fe 3+ . Consider the results that no alloy was observed with 2 wt% graphite addition and excess graphite was observed with 6 wt% graphite addition. Therefore, an optimal addition of graphite is around 4 wt% per residues in terms of alloy morphology and full-liquid slag. The effect of the addition of graphite on the reduction extent of copper, nickel, cobalt, iron and manganese oxides will be discussed in the following section.
MnO-FeO phases. Some reactions occurred in the local area near the pore, where olivine and liquid together with rhodonite were formed. Metals were not observed in the samples after pre-reduction due to the insufficient reductant addition. In Figure 5b, alloy droplets were observed by the reduction with more reductant addition. Olivine and liquid phases were formed in the sample ND9. The alloy and liquid compositions are 38.4Ni-36.1Fe-19.7Cu-3.4Co-2.3Mn and 49.7SiO2-13.4MnO-11.8Al2O3-12.9Na2O-5.5K2O-2.5FeO-2.8CaO-0.7MgO-0.6TiO2, respectively.

Effect of Reductant Addition on the Alloys and Slags
The smelting process was conducted at high temperatures to further reduce nickel, copper, cobalt, and iron oxides to form the required alloys and slags. Figure 6 shows the pictures and microstructures of samples ND8, ND9, and ND10 after smelting at 1350 °C with a carbon addition of 2, 4, and 6 wt%, respectively. It can be seen from Figure 6a that in sample ND8, the slag contains spinel crystals and the alloy is not observed. In samples ND9 and ND10, the slag is fully liquid and alloy sphere (around 5 mm in diameter) settled at the bottom of the crucible can be easily separated from the slag, and small alloy droplets are also present in the slag. A small amount of graphite was left after smelting reduction in sample ND10, which indicates the addition of graphite was in excess. Comparing Figure 6a-c, when the addition of the reductant was insufficient, the liquidus temperature of slag was much higher due to the presence of Fe 3+ . Consider the results that no alloy was observed with 2 wt% graphite addition and excess graphite was observed with 6 wt% graphite addition. Therefore, an optimal addition of graphite is around 4 wt% per residues in terms of alloy morphology and full-liquid slag. The effect of the addition of graphite on the reduction extent of copper, nickel, cobalt, iron and manganese oxides will be discussed in the following section.

Effect of Flux Addition
SiO2 as a flux was added to decrease the liquidus temperature of slag, which enables the slag to be a fully-liquid at the smelting temperature. FactSage calculations shown in Figure 3 demonstrated that the liquidus temperature of the slag decreases rapidly with the addition of up to 50 wt% SiO2. Different amounts of SiO2 were added to verify the predictions at a given temperature of 1350 °C. The dependence of SiO2 addition on the formation of slags with the same amount of graphite additive is shown in Figure 7. It is evident from Figure 7a-c that the slags were fully liquid at the experimental temperature with SiO2 addition over 10 wt%. However, as shown in Figure 7d, with the SiO2 addition down to 5 wt%, the liquidus temperature of the slag was above the experimental temperature resulting in the presence of solid. The primary monoxide phase precipitated in the slag, and tiny reduced alloy blebs are distributed in the periphery of the monoxide solids. In addition to the liquidus temperature, the viscosity of slag can influence the settlement of alloys and tapping of slags. The viscosity of slags in the full-liquid state at 1350 °C with the addition of different amounts of SiO2 was calculated by FactSage 7.2. Figure 8 shows the viscosity of slag increases with increasing SiO2. The addition of SiO2 needs to be minimized to keep the slags in a full liquid state with a relatively low viscosity. Overall, SiO2 addition can be optimized by several factors including the liquidus temperature of slag for operational temperature, the viscosity of slag for separation of alloy and slag tapping, and MnO-grade of the residue slag for further production of ferromanganese or silicomanganese alloys.

Effect of Flux Addition
SiO 2 as a flux was added to decrease the liquidus temperature of slag, which enables the slag to be a fully-liquid at the smelting temperature. FactSage calculations shown in Figure 3 demonstrated that the liquidus temperature of the slag decreases rapidly with the addition of up to 50 wt% SiO 2 . Different amounts of SiO 2 were added to verify the predictions at a given temperature of 1350 • C. The dependence of SiO 2 addition on the formation of slags with the same amount of graphite additive is shown in Figure 7. It is evident from Figure 7a-c that the slags were fully liquid at the experimental temperature with SiO 2 addition over 10 wt%. However, as shown in Figure 7d, with the SiO 2 addition down to 5 wt%, the liquidus temperature of the slag was above the experimental temperature resulting in the presence of solid. The primary monoxide phase precipitated in the slag, and tiny reduced alloy blebs are distributed in the periphery of the monoxide solids. In addition to the liquidus temperature, the viscosity of slag can influence the settlement of alloys and tapping of slags. The viscosity of slags in the full-liquid state at 1350 • C with the addition of different amounts of SiO 2 was calculated by FactSage 7.2. Figure 8 shows the viscosity of slag increases with increasing SiO 2 . The addition of SiO 2 needs to be minimized to keep the slags in a full liquid state with a relatively low viscosity. Overall, SiO 2 addition can be optimized by several factors including the liquidus temperature of slag for operational temperature, the viscosity of slag for separation of alloy and slag tapping, and MnO-grade of the residue slag for further production of ferromanganese or silicomanganese alloys.

Effect of Graphite Addition on the Alloy Grade
The liquidus temperature of slags has been predicted by FactSage, and verified by the experiments with a variation of SiO2 flux and carbon addition in the above sections. The liquidus temperature of slags decreases with the SiO2 addition and reduction of Fe 3+ to Fe 2+ in the slag by carbon. The reduction degree of FeO in the slag and reduced Fe in the alloy as a function of carbon are shown in Figure 9, where experimental results of samples ND8 to ND11 with 15 wt% SiO2 addition at 1350 °C are compared with FactSage calculation. The compositions of alloys and phases present in the quenched samples are listed in Table 4. Figure 9 shows FeO concentration in the slag

Effect of Graphite Addition on the Alloy Grade
The liquidus temperature of slags has been predicted by FactSage, and verified by the experiments with a variation of SiO2 flux and carbon addition in the above sections. The liquidus temperature of slags decreases with the SiO2 addition and reduction of Fe 3+ to Fe 2+ in the slag by carbon. The reduction degree of FeO in the slag and reduced Fe in the alloy as a function of carbon are shown in Figure 9, where experimental results of samples ND8 to ND11 with 15 wt% SiO2 addition at 1350 °C are compared with FactSage calculation. The compositions of alloys and phases present in the quenched samples are listed in Table 4. Figure 9 shows FeO concentration in the slag

Effect of Graphite Addition on the Alloy Grade
The liquidus temperature of slags has been predicted by FactSage, and verified by the experiments with a variation of SiO 2 flux and carbon addition in the above sections. The liquidus temperature of slags decreases with the SiO 2 addition and reduction of Fe 3+ to Fe 2+ in the slag by carbon. The reduction degree of FeO in the slag and reduced Fe in the alloy as a function of carbon are shown in Figure 9, where experimental results of samples ND8 to ND11 with 15 wt% SiO 2 addition at 1350 • C are compared with FactSage calculation. The compositions of alloys and phases present in the quenched samples are listed in Table 4. Figure 9 shows FeO concentration in the slag decreases along with carbon addition, while Fe concentration in the alloy increases for experimental results. A significant difference was found between experimental results and FactSage calculation for the FeO reduction. FactSage calculation indicates FeO in the slag is fully reduced by 3 wt% carbon, while more carbon is needed in the experiments, in which approximately 8 wt% carbon can fully reduce the FeO in the slag. One of the reasons is that a small fraction of MnO was reduced during the carbothermic reduction, which resulted in carbon more than theoretical is needed. Moreover, the metallic oxides could present in multivalent forms (e.g., MeO x ) rather than bivalent form (e.g., MeO) that require additional carbon to reduce them.
Minerals 2020, 10, x FOR PEER REVIEW 11 of 15 decreases along with carbon addition, while Fe concentration in the alloy increases for experimental results. A significant difference was found between experimental results and FactSage calculation for the FeO reduction. FactSage calculation indicates FeO in the slag is fully reduced by 3 wt% carbon, while more carbon is needed in the experiments, in which approximately 8 wt% carbon can fully reduce the FeO in the slag. One of the reasons is that a small fraction of MnO was reduced during the carbothermic reduction, which resulted in carbon more than theoretical is needed. Moreover, the metallic oxides could present in multivalent forms (e.g., MeOx) rather than bivalent form (e.g., MeO) that require additional carbon to reduce them. Full reduction of CuO, NiO, and CoO are desired to allow for high recovery rates, while the iron oxide can be partly/sufficiently reduced depending on the amount of reductant. Recovery of Cu, Ni, and Co is taken as a whole, and the alloy grade is defined as the weight percentage of the sum of Cu, Ni, and Co, (Cu + Ni + Co). The effect of graphite addition on the alloy grade is shown in Figure 10, where the data is plotted from Table 4. The alloy grade decreases with increasing graphite addition. The sequential ease of reduction of metal oxides is (Cu, Ni, Co) > Fe > (Mn, Si). The (Cu + Ni + Co) concentration gradually decreases in the alloy along with the reduction of Fe and Mn oxides. Figure  10 can be divided into three areas, where "Area I" represents the insufficient addition of reductant and no alloys are reduced, "Area III" means excess addition of reductant, beyond that, MnO (or even SiO2) in the slag will be significantly reduced into the alloys. SiO2 and MnO are not required to be reduced at the smelting reduction stage to produce high-grade Fe-Ni-Cu-Co alloy. "Area II" in Figure  10 is thus the working area to optimize the addition of reductant to control the alloy grade and recovery of valuable metals. Within "Area II", iron oxide is gradually reduced into the alloy along with graphite additive. Iron oxide is almost fully reduced with 8 wt% graphite addition. Full reduction of CuO, NiO, and CoO are desired to allow for high recovery rates, while the iron oxide can be partly/sufficiently reduced depending on the amount of reductant. Recovery of Cu, Ni, and Co is taken as a whole, and the alloy grade is defined as the weight percentage of the sum of Cu, Ni, and Co, (Cu + Ni + Co). The effect of graphite addition on the alloy grade is shown in Figure 10, where the data is plotted from Table 4. The alloy grade decreases with increasing graphite addition. The sequential ease of reduction of metal oxides is (Cu, Ni, Co) > Fe > (Mn, Si). The (Cu + Ni + Co) concentration gradually decreases in the alloy along with the reduction of Fe and Mn oxides. Figure 10 can be divided into three areas, where "Area I" represents the insufficient addition of reductant and no alloys are reduced, "Area III" means excess addition of reductant, beyond that, MnO (or even SiO 2 ) in the slag will be significantly reduced into the alloys. SiO 2 and MnO are not required to be reduced at the smelting reduction stage to produce high-grade Fe-Ni-Cu-Co alloy. "Area II" in Figure 10 is thus the working area to optimize the addition of reductant to control the alloy grade and recovery of valuable metals. Within "Area II", iron oxide is gradually reduced into the alloy along with graphite additive. Iron oxide is almost fully reduced with 8 wt% graphite addition.

Conclusions
An efficient process for the smelting reduction of polymetallic nodules at a lower temperature was investigated by the optimization of both slag and alloy systems. Essential findings are summarized as follows: (1) The as-received nodules mainly contain Mn and Fe oxides in striation. Si is incorporated with Al, Na, K and Ca as impurity minerals. Valuable metal elements such as Cu, Co and Ni in low concentrations are uniformly distributed in the nodule matrix. Ba is found to be associated with S, most likely as barite. (2) FactSage software was used to simulate/predict the liquidus temperature of the slag system, which is not sensitive to FeO, CaO and Al 2 O 3 , but decreases most significantly with decreasing MnO/SiO 2 mass ratio. The SiO 2 flux to decrease the liquidus temperature of slag was experimentally verified. (3) The alloy grade was found to decrease with increasing graphite addition. The optimal reduction was achieved by adding a 4 wt% graphite at the MnO/SiO 2 mass ratio of 1.6 in slag. The most effective metal-slag separation was achieved at 1350 • C. This relatively low-temperature smelting process for the utilization of polymetallic nodules is promising for industrial applications.