Mechanism of Calcium Sulphate on the Aggregation and Growth of Ferronickel Particles in the Self-Reduction of Saprolitic Nickel Laterite Ore

: Saprolitic nickel laterite is characterized by relatively low iron and nickel contents. Iron and nickel oxides are reduced to form ﬁne ferronickel particles that disperse and embed in silicates in the reduction process, limiting the application of magnetic separation to extract ferronickel. Additives are applied to promote the aggregation and growth of ferronickel particles, then the large ferronickel particles will be separated by ﬁne grinding and recovered via magnetic separation. Calcium sulphate is considered to be capable of increasing the size of ferronickel particles considerably. Due to the decomposition of calcium sulphate in the reduction process, the mechanism of calcium sulphate on the aggregation and growth of ferronickel particles should be conducted studied in-depth. The current work explores the e ﬀ ects of calcium sulphate, elemental sulphur, and calcium oxide on the formation of ferronickel particles in a saprolitic nickel laterite ore. The results showed that the formation of an Fe-FeS eutectic and the mineral structure transformation contributed by calcium oxide were all conducive to the mass transfer of ferronickel particles in gangue, ferronickel particles aggregated and grew up at the boundary between the hole and the gangue. The self-reduction, ﬁne grinding, and magnetic separation of nickel laterite ore in the presence of three types of additive were examined. Nickel laterite ore with 7.88 wt% coal, 12 wt% calcium sulphate reduced at 1200 ◦ C for 30 min, a ferronickel concentrate of Ni 8.08 wt%, and Fe 79.98 wt% was obtained at a nickel and iron recovery of 92.6% and 79.9%, respectively. to sulphur and calcium oxide in the reducing atmosphere. The low-melting-point Fe-FeS eutectic formed by the reaction of sulphur with iron oxides


Introduction
The annual global demand for nickel has increased tremendously due to its numerous applications, especially advantageous for the production of stainless steel, alloys, batteries, and catalysts [1][2][3][4][5][6]. In the last decade, with the continuous depletion of nickel sulfide, increasing attention has been paid to nickel laterite ore [7][8][9][10][11][12][13]. Nickel laterite deposits are classified into two distinct types: limonitic deposits found in the surface layer while the bottom layer is referred to as saprolitic deposits [14]. For saprolitic laterite ore, nickel and iron are difficult to concentrate by physical beneficiation due to the low nickel and iron content [15] and isomorphic goethite or serpentine crystalline structure [9]. Self-reduction of  Anthracite was used as the reductant in this study and ground to a size of -100 mesh for use. The industrial analysis of the anthracite showed that it contained 82.76 wt% fixed carbon, 5.28 wt% volatiles, 6.82 wt% ash, 5.14 wt% moisture, and 0.60 wt% total sulphur. Calcium sulphur (CaSO4), sulphur (S) and calcium oxide (CaO) used in the study are of chemical grade.

Experimental Methods
Ground nickel laterite ore, coal powder, and additives were uniformly mixed in a certain proportion. The coal addition was 7.88 wt%, calcium sulphate additions in the range of 0-16 wt%, sulphur additions in the range of 0-3.96 wt%, and calcium additions ranging from 0-6.6 wt%, the required dosage of all additives with a balanced amount of nickel laterite ore sample. A total mixtures mass of 25 g was wetted with an appropriate amount of water and subsequently compacted at 6000 psi into a cylindrical sample with a diameter of 20 mm and an approximate height of 30 mm. Samples were dried at 110 °C before high-temperature reduction experiments.
The reduction experiments were conducted in a 25 kW high-temperature carbon tube furnace. Samples were charged into a corundum crucible and held for 30 min after the temperature reached the specified operating temperature of 1200 °C under a nitrogen atmosphere. Reduced samples were quenched by quickly immersing them in water and then dried at 110 °C for subsequent detection and magnetic separation.
The reduced sample with a total mass of 15 g was ground by a centrifugal grinding machine for 12 min. The slurry was separated into magnetic product and nonmagnetic product using wet magnetic separation with a magnetic field intensity of 0.1 T. Finally, the obtained magnetic product was ferronickel concentrate. The compositions of ferronickel concentrate were tested and the corresponding nickel and iron recoveries were calculated per Equation (1), and the compositions of slag were calculated mass balance calculation in Equation (2): Anthracite was used as the reductant in this study and ground to a size of −100 mesh for use. The industrial analysis of the anthracite showed that it contained 82.76 wt% fixed carbon, 5.28 wt% volatiles, 6.82 wt% ash, 5.14 wt% moisture, and 0.60 wt% total sulphur. Calcium sulphur (CaSO 4 ), sulphur (S) and calcium oxide (CaO) used in the study are of chemical grade.

Experimental Methods
Ground nickel laterite ore, coal powder, and additives were uniformly mixed in a certain proportion. The coal addition was 7.88 wt%, calcium sulphate additions in the range of 0-16 wt%, sulphur additions in the range of 0-3.96 wt%, and calcium additions ranging from 0-6.6 wt%, the required dosage of all additives with a balanced amount of nickel laterite ore sample. A total mixtures mass of 25 g was wetted with an appropriate amount of water and subsequently compacted at 6000 psi into a cylindrical sample with a diameter of 20 mm and an approximate height of 30 mm. Samples were dried at 110 • C before high-temperature reduction experiments.
The reduction experiments were conducted in a 25 kW high-temperature carbon tube furnace. Samples were charged into a corundum crucible and held for 30 min after the temperature reached the specified operating temperature of 1200 • C under a nitrogen atmosphere. Reduced samples were quenched by quickly immersing them in water and then dried at 110 • C for subsequent detection and magnetic separation.
The reduced sample with a total mass of 15 g was ground by a centrifugal grinding machine for 12 min. The slurry was separated into magnetic product and nonmagnetic product using wet magnetic separation with a magnetic field intensity of 0.1 T. Finally, the obtained magnetic product was ferronickel concentrate. The compositions of ferronickel concentrate were tested and the corresponding nickel and iron recoveries were calculated per Equation (1), and the compositions of slag were calculated mass balance calculation in Equation (2): where η is the nickel or iron recovery, λ is the content of each element in slag, m l is the mass of the ferronickel concentrate, m 0 is the mass of the slurry, β is the content of each element in ferronickel concentrate, and α is the content of each element in slurry.

Analytical Methods
The reduced samples were ground into fine powder for XRF and XRD analysis. The chemical compositions were determined by XRF (Thermo Scientific, ARL 9900, Waltham, MA, USA). The approach uses X-rays to excite the sample, which in turn causes the sample to emit secondary X-rays These secondary (or fluorescent) X-rays are characteristic of the element from which they were emitted, and so can be indicative of the elemental composition of the sample. A Bruker X'Pert Pro powder diffraction instrument (Bruker, Karlsruhe, Germany) (with X'Celerator detector, Cu Kα X-radiation, and 2θ scans from 10 to 90 deg) was used to analyse the crystalline phases in the samples. The phase compositions and microstructures were examined using field emission scanning electron microscopy (SEM) (Nova 400 Nano SEM; FEI, Hillsboro, OR, USA) at a setting of 20 kV and coupled with EDS (INCA, IE 350 Penta FET X-3, Oxford Instruments, Oxford, UK). All the samples were coated with Au for SEM-EDS analysis.

Determination of Particle Size of Ferronickel Particles
Particle size analysis was performed using the software package ImageJ to metallic particles from 20 regions per sample, with each region observed using Scanning Electron Microscopy (SEM). The profile areas S reported are calculated from the remaining population of particles on a per-sample basis. Because of the irregular shapes of ferronickel particles, the shape of particles was assumed to be spherical [40]. Therefore, the equivalent size d of a single ferronickel particle was calculated using Equation (3): The equivalent sizes of ferronickel particles in the 20 continuous view fields were summarized and statistically analysed. All ferronickel particles were divided into several intervals at a step length of 2 µm according to the equivalent size. The mathematical average equivalent size d i of the ferronickel particles within each interval was calculated. Therefore, the average equivalent size d of ferronickel particles in 20 view fields was calculated using Equation (4): where S i is the total profile area of ferronickel particles within a certain interval, and S T is the total profile area of all ferronickel particles in the 20 view fields.

Self-Reduction of Saprolitic Laterite Ore
The thermodynamic analysis of the carbothermic reduction roasting of nickeliferous limonitic laterite ore has been performed by Pickles [41]. The major nickel-bearing phase in the limonitic ores is goethite, for a saprolitic laterite ore, the nickel is widely disseminated in solid solution in the host rock [10], nickel oxide-containing chlorite and goethite undergoes a dehydroxylation reaction and form nickel oxide and iron oxide. Thus, ideal behavior has been assumed in one or more of the phases, so the reactions for the self-reduction of nickel and oxides from nickel laterite ore would be as follows: FeO + C = Fe + CO(g) NiO Since the self-reduction reactions produce carbon monoxide, the gaseous reduction of nickel and iron oxides would be: FeO + CO(g) = Fe + CO 2 (g) NiO + CO(g) = Ni + CO 2 (g) (18) In the metal phase, a solid solution of ferronickel alloy would form.

Effect of on the Particle Size and Distribution of Ferronickel Particles
The anthracite dosage was fixed at 7.88 wt% which was calculated based on the C/O ratio, nickel laterite ore without additive was reduced at 1200 • C for 30 min to the sufficient reduction of nickel and iron oxides. The micrograph of the reduced nickel laterite ore is shown in Figure 2. As shown in Figure 2, the nickel and iron oxides were reduced to form fine ferronickel particles that dispersed and embedded in silicates.

Effect of on the Particle Size and Distribution of Ferronickel Particles
The anthracite dosage was fixed at 7.88 wt% which was calculated based on the C/O ratio, nickel laterite ore without additive was reduced at 1200 °C for 30 min to the sufficient reduction of nickel and iron oxides. The micrograph of the reduced nickel laterite ore is shown in Figure 2. As shown in Figure 2, the nickel and iron oxides were reduced to form fine ferronickel particles that dispersed and embedded in silicates.  The effects of the various calcium, sulphur, and calcium oxide additions on the size and distribution of ferronickel particles are demonstrated in the micrographs in Figures 3-5. These micrographs depict representative areas from samples reduced at 1200 • C for 30 min. As seen in Figure 3, a trend of ferronickel particle size increased with increasing calcium sulphate addition is evident. The number of ferronickel particles decreased, and larger particles aggregated at the boundary between the hole and the gangue due to the aggregation and growth of fine ferronickel particles.
In the case of Figure 4, sulphur addition resulted in similar ferronickel particle size and distribution as in the absence of additive case. Sulphur significantly decreases the surface tension of metal particles, and thus the elastic resistance of metal particles agglomeration decreases [33]. As a saprolitic laterite ore with a great mass of gangue different from that of the limonitic laterite ore, the migration of ferronickel particles in gangue is an important factor for the aggregation and growth of fine ferronickel particles, and the low reduction temperature was not conducive to ferronickel particle migration.
Metals 2020, 10, 423 6 of 17 micrographs depict representative areas from samples reduced at 1200 °C for 30 min. As seen in Figure 3, a trend of ferronickel particle size increased with increasing calcium sulphate addition is evident. The number of ferronickel particles decreased, and larger particles aggregated at the boundary between the hole and the gangue due to the aggregation and growth of fine ferronickel particles. In the case of Figure 4, sulphur addition resulted in similar ferronickel particle size and distribution as in the absence of additive case. Sulphur significantly decreases the surface tension of metal particles, and thus the elastic resistance of metal particles agglomeration decreases [33]. As a saprolitic laterite ore with a great mass of gangue different from that of the limonitic laterite ore, the migration of ferronickel particles in gangue is an important factor for the aggregation and growth of fine ferronickel particles, and the low reduction temperature was not conducive to ferronickel particle migration. As seen in the micrographs in Figure 5, there is an increase in particle size resulting from the increased addition of calcium oxide, with few calcium oxide additions demonstrating a slight influence on the size and distribution of ferronickel particles. Relatively large ferronickel particles were observed at the boundary between the hole and the gangue with the calcium oxide dosage up to 6.6 wt%. Calcium oxide reacted with silica to form a low melting point mineral [19], which promoted ferronickel particle migration in the gangue.
The profile area of ferronickel particles was obtained using ImageJ software, and the average particle size was calculated according to Equations (2) and (3). The results are shown in Figure 6. Ferronickel particles in the absence of additive considered here were on average of 0.975 µm, significantly smaller than the ferronickel particles formed in the presence of additives. The average size of ferronickel particles exceeded 20 µm with increasing calcium sulphate dosage to 12 wt%. With respect to sulphur and calcium oxide addition, increasing the addition led to the growth of ferronickel particles, but the average size was less than 5 µm. As the dosage of sulphur or calcium oxide addition was the sulphur or calcium oxide content of calcium sulphate, the increase in particle growth resulting from the increased addition of calcium sulphate is significantly greater than that observed for a similar increase in the mass of sulphur or calcium oxide. This suggests that the mechanism of calcium sulphate to aid the growth of the ferronickel particles was not the result of either sulphur or calcium oxide acting unilaterally.  As seen in the micrographs in Figure 5, there is an increase in particle size resulting from the increased addition of calcium oxide, with few calcium oxide additions demonstrating a slight influence on the size and distribution of ferronickel particles. Relatively large ferronickel particles were observed at the boundary between the hole and the gangue with the calcium oxide dosage up to 6.6 wt%. Calcium oxide reacted with silica to form a low melting point mineral [19], which promoted ferronickel particle migration in the gangue. The profile area of ferronickel particles was obtained using ImageJ software, and the average particle size was calculated according to Equations (2) and (3). The results are shown in Figure 6. Ferronickel particles in the absence of additive considered here were on average of 0.975 μm, significantly smaller than the ferronickel particles formed in the presence of additives. The average size of ferronickel particles exceeded 20 μm with increasing calcium sulphate dosage to 12 wt%. With respect to sulphur and calcium oxide addition, increasing the addition led to the growth of ferronickel particles, but the average size was less than 5 μm. As the dosage of sulphur or calcium oxide addition was the sulphur or calcium oxide content of calcium sulphate, the increase in particle growth resulting from the increased addition of calcium sulphate is significantly greater than that observed for a similar increase in the mass of sulphur or calcium oxide. This suggests that the mechanism of calcium sulphate to aid the growth of the ferronickel particles was not the result of either sulphur or calcium oxide acting unilaterally. The profile area of ferronickel particles was used to represent the distribution of ferronickel, and the distribution of ferronickel with different particle sizes is presented in Figure 7. In the absence of additive, approximately 85% of ferronickel particles have a particle size less than 2 μm, and no particles above 6 μm were observed. The presence of calcium sulphate in the ore results in a shift The profile area of ferronickel particles was used to represent the distribution of ferronickel, and the distribution of ferronickel with different particle sizes is presented in Figure 7. In the absence of additive, approximately 85% of ferronickel particles have a particle size less than 2 µm, and no particles above 6 µm were observed. The presence of calcium sulphate in the ore results in a shift towards larger particles. More than 70% of ferronickel formed ferronickel particles larger than 10 µm with calcium sulphate addition up to 16 wt%. With the addition of merely sulphur or calcium oxide, most ferronickel particles were smaller than 4 µm. Figure 6. The average ferronickel particle size from samples with 7.88 wt% coal in the absence and presence of additive, reduced at a reduction temperature of 1200 °C for 30 min.
The profile area of ferronickel particles was used to represent the distribution of ferronickel, and the distribution of ferronickel with different particle sizes is presented in Figure 7. In the absence of additive, approximately 85% of ferronickel particles have a particle size less than 2 μm, and no particles above 6 μm were observed. The presence of calcium sulphate in the ore results in a shift towards larger particles. More than 70% of ferronickel formed ferronickel particles larger than 10 μm with calcium sulphate addition up to 16 wt%. With the addition of merely sulphur or calcium oxide, most ferronickel particles were smaller than 4 μm.

Mechanism of Calcium Sulphate on the Aggregation and Growth of Ferronickel Particles
X-ray diffraction analysis (XRD) was carried out to demonstrate the differences in mineral composition with the addition of different additives are shown in Figure 8. Combining the XRD pattern of the original ore ( Figure 1) and samples reduced at 1200 °C for comparison, all the goethite

Mechanism of Calcium Sulphate on the Aggregation and Growth of Ferronickel Particles
X-ray diffraction analysis (XRD) was carried out to demonstrate the differences in mineral composition with the addition of different additives are shown in Figure 8. Combining the XRD pattern of the original ore ( Figure 1) and samples reduced at 1200 • C for comparison, all the goethite phase disappeared and phase transformation occurs to Fe 2 O 3 and ferronickel alloy [13]. Clinoenstatite (MgSiO 3 ) and forsterite ((Mg,Fe) 2 SiO 4 ) are resulted from the dehydroxylation of Fe-substituted lizardite [(Mg,Fe) 3 Si 2 O 5 (OH) 4 ] and the recrystallization of silicate minerals [21][22][23]. The new mineralogical compositions of pigeonite (Ca(Mg,Fe)Si 2 O 6 ) and troilite (FeS) in the reduced nickel laterite ore with a dosage of 16wt% calcium sulphate appeared. Calcium sulphate decomposed to calcium oxide and sulphur in a reducing atmosphere [35].
Calcium oxide reacted with silicate to form pigeonite, troilite generated by the reaction of S 2 with iron oxides or iron [25,34]. As a similar mechanism, pigeonite was found in the reduced sample in the presence of calcium oxide and troilite appeared in the reduced sample with an additive of sulphur.
The microstructures and elemental maps of samples with calcium sulphate, sulphur and calcium oxide addition reduced at 1200 • C were analysed by SEM coupled with EDS are shown in Figures 9-11. The corresponding EDS results of the mineral composition are listed in Tables 2-4. calcium oxide and sulphur in a reducing atmosphere [35].
2CaSO4+6CO=2CaO+S2+6CO2 (19) Calcium oxide reacted with silicate to form pigeonite, troilite generated by the reaction of S2 with iron oxides or iron [25,34]. As a similar mechanism, pigeonite was found in the reduced sample in the presence of calcium oxide and troilite appeared in the reduced sample with an additive of sulphur. In the case of Figure 9, troilite was present as a greyish white phase, surrounded the ferronickel particles or bridges with different big ferronickel particles, and reacted with iron to form an Fe-FeS eutectic [35]. The Fe-FeS eutectic as a low-melting-point mineral wrapped ferronickel particles and migrated in gangue together, different ferronickel particles aggregated and grew up by bridge in the migration process, then precipitated at the boundary between the hole and the gangue. Chromium and nickel oxides were also sulphurized by S2, NiS then reacted with FeO to form FeS in a CO atmosphere [26]. 2Cr2O3 +3C+3S2=2Cr2S3+3CO2 (20) 2NiO+C+S2=2NiS+CO2 (21) NiS+FeO+CO=Ni+FeS+CO2 (22) Meanwhile, calcium oxide from the decomposition of calcium sulphate in the reduction atmosphere, reacted readily with silicate to form a low-melting-point mineral accelerated the migration of ferronickel particles in gangue and facilitated their aggregation and growth. The mechanism of calcium sulfate in promoting the ferronickel particle growth is similar to that of sodium sulfate described by Jiang [21].  Table 2. EDS analysis of mineral composition in the reduced sample with CaSO4 addition in Figure 9 (at%).    Table 3. EDS analysis of mineral composition in the reduced sample with sulphur addition in Figure  10 (at%). An appropriate amount of calcium oxide contributes to liquid generation and slag viscosity [15]. In the present work, the dosage of calcium oxide addition was 6.6 wt% less than that of the literature [15], and the reduction temperature of 1200 °C was also low. These conditions could not favor the liquid generation. Furthermore, calcium oxide transformed the compact structure of forsterite and the formation of a loose mineral structure was conducive to the migration of ferronickel particles in the gangue.  Table 2. EDS analysis of mineral composition in the reduced sample with CaSO 4 addition in Figure 9 (at%).  Table 4. EDS analysis of mineral composition in the reduced sample with CaO addition in Figure 11 (at%).  Figure 11. Back-scattered electron (BSE) image of nickel laterite ore with an additive of 6.6 wt% CaO by reducing at 1200 °C for 30 min.

Analysis of Magnetic Products
Magnetic products were obtained from reduced nickel laterite ore by fine grinding and magnetic separation, the compositions of magnetic products are shown in Table 5. Magnetic separation results of nickel laterite ore reduced without additive are undesirable as the Ni and Fe grades are only 5.59 wt% and 60.23 wt%, and the gangue compositions of SiO2 and MgO are 13.35 wt% and 4.88 wt%, respectively. Generally, when the metallic particles have a particle size of less than 10 μm, they are very difficult to liberate during the grinding process, and very difficult to recover even if liberated [42][43][44]. The nickel and iron grades continuously improved with the increase of calcium sulphate dosage. By comparison with those in the absence of calcium sulphate, the nickel content of ferronickel concentrate increases significantly from 5.59 wt% to 8.52 wt% in the presence of 16 wt% calcium sulphate. As the ferronickel particles obtained in the presence of sulphur or calcium oxide smaller than 10 μm, in this case, there is a less-than-ideal separation between the ferronickel particles and the gangue achieved.  Figure 11. Back-scattered electron (BSE) image of nickel laterite ore with an additive of 6.6 wt% CaO by reducing at 1200 • C for 30 min. Table 3. EDS analysis of mineral composition in the reduced sample with sulphur addition in Figure 10 (at%).  Table 4. EDS analysis of mineral composition in the reduced sample with CaO addition in Figure 11 (at%). In the case of Figure 9, troilite was present as a greyish white phase, surrounded the ferronickel particles or bridges with different big ferronickel particles, and reacted with iron to form an Fe-FeS eutectic [35]. The Fe-FeS eutectic as a low-melting-point mineral wrapped ferronickel particles and migrated in gangue together, different ferronickel particles aggregated and grew up by bridge in the migration process, then precipitated at the boundary between the hole and the gangue. Chromium and nickel oxides were also sulphurized by S 2 , NiS then reacted with FeO to form FeS in a CO atmosphere [26].

Mineral
2NiO + C + S 2 = 2NiS + CO 2 (21) NiS + FeO + CO = Ni + FeS + CO 2 (22) Meanwhile, calcium oxide from the decomposition of calcium sulphate in the reduction atmosphere, reacted readily with silicate to form a low-melting-point mineral accelerated the migration of ferronickel particles in gangue and facilitated their aggregation and growth. The mechanism of calcium sulfate in promoting the ferronickel particle growth is similar to that of sodium sulfate described by Jiang [21].
The case of sulphur is shown in Figure 10. It also appears that a sulphide network surrounds the ferronickel particles. However, ferronickel particles encased in the Fe-FeS eutectic are still embedded in silicate. Forsterite is an island structure silicate with a compact crystalline structure, limiting the migration of ferronickel particles.
An appropriate amount of calcium oxide contributes to liquid generation and slag viscosity [15]. In the present work, the dosage of calcium oxide addition was 6.6 wt% less than that of the literature [15], and the reduction temperature of 1200 • C was also low. These conditions could not favor the liquid generation. Furthermore, calcium oxide transformed the compact structure of forsterite and the formation of a loose mineral structure was conducive to the migration of ferronickel particles in the gangue.

Analysis of Magnetic Products
Magnetic products were obtained from reduced nickel laterite ore by fine grinding and magnetic separation, the compositions of magnetic products are shown in Table 5. Magnetic separation results of nickel laterite ore reduced without additive are undesirable as the Ni and Fe grades are only 5.59 wt% and 60.23 wt%, and the gangue compositions of SiO 2 and MgO are 13.35 wt% and 4.88 wt%, respectively. Generally, when the metallic particles have a particle size of less than 10 µm, they are very difficult to liberate during the grinding process, and very difficult to recover even if liberated [42][43][44]. The nickel and iron grades continuously improved with the increase of calcium sulphate dosage. By comparison with those in the absence of calcium sulphate, the nickel content of ferronickel concentrate increases significantly from 5.59 wt% to 8.52 wt% in the presence of 16 wt% calcium sulphate. As the ferronickel particles obtained in the presence of sulphur or calcium oxide smaller than 10 µm, in this case, there is a less-than-ideal separation between the ferronickel particles and the gangue achieved.

Analysis of Slags
The compositions of slags were calculated by mass balance calculation and the results are shown in Table 6. With increasing calcium sulphate dosage, the nickel and iron contents of slag decrease. In the presence of sulphur or calcium oxide, the compositions of slags are consistent with that in the absence of additive. The results show that calcium sulphate is beneficial to separate ferronickel from the slag.  Figure 12 presents the nickel and iron recoveries of the ferronickel concentrate obtained in the presence of different additives. The nickel recovery increase with increasing the dosage of additive, the increment of nickel recovery with calcium sulphate is significantly greater than that in the presence of sulphur or calcium oxide. The same phenomenon appears for the recovery of iron. These results suggest that large ferronickel particle favour the efficient recycling of nickel. In the presence of 12 wt% calcium sulphate, the corresponding nickel and iron recoveries are 92.59% and 79.9% respectively.
The compositions of slags were calculated by mass balance calculation and the results are shown in Table 6. With increasing calcium sulphate dosage, the nickel and iron contents of slag decrease. In the presence of sulphur or calcium oxide, the compositions of slags are consistent with that in the absence of additive. The results show that calcium sulphate is beneficial to separate ferronickel from the slag.  Figure 12 presents the nickel and iron recoveries of the ferronickel concentrate obtained in the presence of different additives. The nickel recovery increase with increasing the dosage of additive, the increment of nickel recovery with calcium sulphate is significantly greater than that in the presence of sulphur or calcium oxide. The same phenomenon appears for the recovery of iron. These results suggest that large ferronickel particle favour the efficient recycling of nickel. In the presence of 12 wt% calcium sulphate, the corresponding nickel and iron recoveries are 92.59% and 79.9% respectively.

Conclusions
Nickel laterite ore with an anthracite dosage of 7.88 wt% and a calcium sulphate addition of 16 wt% reduced at 1200 • C for 30 min, a ferronickel concentrate with 8.52 wt% Ni, and 80.75 wt% Fe was obtained by magnetic separation. The corresponding recoveries of Ni and Fe were 94.8% and 82.1%, respectively.
The growth of ferronickel particles during the reduction roasting of a saprolitic laterite ore has been studied as a function of calcium sulphate, sulphur, and calcium oxide additions to the ore. The particle growth resulting from calcium sulphate is significantly greater than that resulting from sulphur or calcium oxide. Calcium sulphate decomposes to sulphur and calcium oxide in the reducing atmosphere. The low-melting-point Fe-FeS eutectic formed by the reaction of sulphur with iron oxides or iron which facilitate the migration of ferronickel particles. Calcium oxide could not contribute to liquid generation at a low temperature, but transforms the compact structure of forsterite, which is conducive to the migration of ferronickel particles in the gangue. Thus, large ferronickel particles precipitate at the boundary between the hole and the gangue. The results demonstrate that fine ferronickel particles in the reduced saprolitic laterite ore aggregate and grow up under the joint action of a liquid-phase growth mechanism driven by the Fe-FeS eutectic and a loose mineral structure formation by calcium oxide. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of Interest:
The authors declare no conflict of interest.