TSpace Research Repository tspace.library.utoronto.ca

TG/DTA study on the carbon monoxide and graphite thermal reduction of a high-grade

iron nickel oxide residue with the presence of siliceous gangue

Dawei Yu, Mingqian Zhu, Torstein A. Utigard and Mansoor Barati

Version Post-print/Accepted Manuscript

Citation (published version)

Yu, D., Zhu, M., Utigard, T.A. and Barati, M., 2014. TG/DTA study on the carbon monoxide and graphite thermal reduction of a high-grade iron nickel oxide residue with the presence of siliceous gangue. Thermochimica Acta, 575, pp.1-11. DOI: 10.1016/j.tca.2013.10.015

Copyright/License This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0

International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

How to cite TSpace items

Always cite the published version, so the author(s) will receive recognition through services that track citation counts, e.g. Scopus. If you need to cite the page number of the author manuscript from TSpace

because you cannot access the published version, then cite the TSpace version in addition to the published version using the permanent URI (handle) found on the record page.

This article was made openly accessible by U of T Faculty. Please tell us how this access benefits you. Your story matters.

1

TG/DTA Study on the Carbon Monoxide and Graphite Thermal Reduction of a High-Grade Iron Nickel 1

Oxide Residue with the Presence of Siliceous Gangue 2

Dawei Yu*, Mingqian Zhu, Torstein A. Utigard, Mansoor Barati 3

University of Toronto, Department of Materials Science and Engineering, 184 College Street, Toronto, 4

Ontario, M5S3E4, Canada 5

*Corresponding author. Tel.: (+1)416 978 0912. E-mail Address: dawei.yu@mail.utoronto.ca (D. Yu) 6

7

Abstract 8

The reduction behavior of a leach residue containing hematite (Fe2O3), nickel ferrite (NiFe2O4) and a 9

small amount of siliceous gangue was studied using thermogravimetric (TG) and differential thermal 10

analysis (DTA) techniques. Carbon monoxide and graphite were used as reductants and a wide range of 11

temperatures were covered. Both non-isothermal reduction (15 ºC/min) and isothermal reduction tests 12

with continuous measurement of the CO and CO2 evolution were performed to evaluate the temperature 13

dependence of the reduction kinetics. In order to further elucidate the reduction mechanisms, the reduced 14

products were subject to various analyses, i.e. SEM/EDS, XRD, and EPMA. Results show that the 15

reduction rate increased with the increase of temperature when using CO as the reductant. Change of the 16

rate-controlling mechanism from the gaseous diffusion in the pores of the sample bed to the diffusion in 17

the metallic product layer was found to have taken place in the isothermal tests, resulting in the variation 18

of apparent activation energies from 6.7 kJ/mol to 41.1 kJ/mol with the progress of reduction. The 19

siliceous gangue melted at 1200 ºC, inhibiting the reduction of the oxides. Effective reduction by graphite 20

only occurred above around 1000 ºC. Reduction of the molten silicate took place above 1200 ºC resulting 21

in the formation of Fe-Ni-Si alloys. 22

23

Keywords: TGA; DTA; Reduction; Hematite; Nickel ferrite; Silicate 24

25

1. Introduction 26

An integrated pyro- and hydro-metallurgical approach to treat nickel concentrates through sulfation 27

roasting followed by leaching, is being investigated [1, 2] as an alternative processing route with potential 28

environmental benefits. In this process, the majority of the non-ferrous valuable metals are leached with 29

water after sulfation roasting, leaving iron in the leach residue as oxides. However, approximately 20% of 30

the nickel reports to the leach residue due to the inevitable formation of nickel ferrite (NiFe2O4) during 31

sulfation roasting [3]. This leach residue, which is mainly composed of hematite (Fe2O3), nickel ferrite, 32

and siliceous gangue, can be a high-grade source of Ni (10 wt%) and Fe (50 wt%) for the production of 33

ferronickel alloy. As a result, the reduction of the Ni- and Fe-bearing leach residue to produce ferronickel 34

alloy becomes an integral part of the sulfation roasting route and was studied. 35

2

Kinetic study on the reduction of this residue with hydrogen as the reductant has been conducted and 36

reported elsewhere [3]. In this work, the feasibility of using carbon monoxide and graphite as reductants 37

to reduce the leach residue and its reaction kinetics were studied by means of TGA and DTA. The 38

heterogeneous reduction kinetics and mechanism could be affected by various factors, such as the 39

reduction temperature [4], the gas composition [5], the particle size [6], the original structure [7], the size 40

[8] and shape [9] of the oxides ore pellets, and the sintering conditions [10], etc. Structural changes during 41

the reduction process has been investigated extensively [11] and they also imposes a significant impact on 42

the reduction rate [12, 13]. The changes in porosity, specific surface area and the pore size distribution 43

have been reported to strongly depend on the reduction temperature [12, 14]. 44

Fig. 1 displays the stability diagram of iron oxides in CO/CO2 atmosphere, as well as CO partial pressure 45

in the CO-CO2-C equilibrium, which can be represented by the Boudouard reaction (Eq. (1)) [15]. This 46

figure is drawn using thermodynamic data from a thermodynamic software package [16]. It illustrates that 47

depending on the reduction temperature, the reduction of Fe2O3 to metallic Fe could adopt a two-step 48

(<550 ºC) or three-step mechanism (>550 ºC) with the increase of the partial pressure of CO, admitting 49

that in an actual Fe2O3 reduction process, the reduction mechanism could be largely different from the 50

thermodynamic prediction and be more complicated due to the contributing factors mentioned above. It 51

has generally been accepted that the indirect reduction of iron oxides with CO dominates the reduction in 52

a blast furnace as well as in the direct reduction processes producing sponge iron, the solid-solid direct 53

reduction playing only a minor role [17]. And conclusion has been made that for those reduction 54

processes with carbon or carbon monoxide as the reductant, the reversible Boudouard reaction could 55

display a strong controlling influence on the reaction kinetics and mechanism [18-20]. However, there are 56

studies [21-23] in which NiO was reduced by pyrolytic and natural graphite showing that the direct 57

reduction played a major role, and the reaction occurred mostly at the edge surface than at the cleavage 58

surface of the graphite due to the much higher number of active carbon sites on the former where 59

reduction took place preferably. This suggests that the nature of the carbon source (graphite, charcoal, 60

coke, etc.) could also have a significant impact on the kinetics and mechanism of reduction. 61

C(s) + CO2(g) = 2CO(g) (1) 62

63

2. Experimental 64

2.1 Materials 65

The leach residue for investigation is primarily composed of hematite (Fe2O3 with varying concentrations 66

of Ni as impurities), nickel ferrite (NiFe2O4) and siliceous gangue. Table 1 shows the chemical and 67

mineralogical compositions of the material. Particle size analysis shows 100% below 100 µm with a 68

major portion in the 5–40 µm range. Fig. 2 displays the backscattered electron (BSE) image of the leach 69

residue [3]. It comprises three types of particles: (1) hematite particles with its characteristic porous 70

columnar microstructure (e.g. particle a in Fig. 2); (2) ring-shaped particles containing nickel ferrite and 71

hematite (e.g. particle b in Fig. 2; Ni3S2 resides within some of the ring-shaped particles shown as bright 72

areas); (3) siliceous gangue containing Fe, Mg, Al (e.g. particle c in Fig. 2). 73

2.2 TG/DTA study 74

3

All reduction experiments were conducted in a Setaram TG-DTA 92 unit (SETARAM Inc., Newark, CA). 75

For the reduction with CO, non-isothermal reduction tests at 15 ºC/min from room temperature to 1400 ºC 76

were performed in both TGA and DTA modes in 1 L/min pure CO gas (CO2<100 ppm, O2<100 ppm). 77

Leach residue of typically 50 mg was loaded into an alumina crucible (I.D. 7.2 mm) for TGA runs, 78

whereas the sample size was 15 mg for DTA runs using alumina crucibles (I.D. 4.1 mm). Calcined Al2O3 79

powder of 15 mg was used as reference for DTA runs. The gaseous reduction product CO2 was 80

continuously swept away by the flow of CO, and the offgas was continuously analyzed for its CO2 81

concentration using a gas analyzer (ABB EL3020). Isothermal reduction tests in the temperature range 82

from 400 ºC to 1300 ºC were also carried out using CO. For each isothermal test, the sample was raised at 83

15 ºC/min to the target temperature in 1 L/min argon, followed by the reduction using 1 L/min CO gas. 84

Two series of reduction tests were also performed using graphite as the reductant. The graphite powder 85

was produced by grinding a purified (ash content <50 ppm) graphite rod and the fraction below 25 µm 86

was used. The powder and the leach residue were well blended with the weight ratio of 2:5 to provide 87

excess carbon for complete reduction. 70 mg of the mixture (50 mg residue + 20 mg graphite) was used 88

for each TGA run, whereas the sample size for the DTA run was 21 mg (15 mg residue + 6 mg graphite). 89

Non-isothermal reduction of the mixture was conducted in both TGA and DTA modes at 15 ºC/min to 90

1500 ºC in 1 L/min argon flow. The offgas was continuously analyzed for its CO and CO2 content. The 91

isothermal reduction tests were performed for a temperature range of 800–1400 ºC in the following 92

manner: The sample mixture was loaded in an alumina crucible, and was first suspended from the micro-93

balance well above the hot zone of the TGA chamber. The chamber was then heated to the target 94

temperature at the rate of 15 ºC/min while being purged with argon (1 L/min), to prevent air ingress. After 95

the target temperature was reached, the sample was quickly lowered into the chamber with continued 96

purge of 1 L/min argon. Because of the small sample size used, the sample was instantly heated to the 97

target temperature by radiation received from the chamber and the reduction was initiated. 98

2.3 Analytical methods 99

Reduced products from the experiments were mounted into epoxy, grinded and polished for the 100

observation of their microstructures and for phase identifications using scanning electron microscope 101

(SEM, JEOL JSM-840) which is complemented by an EDS detector. Elemental analysis of the phases in 102

the reduced samples was conducted using an electron probe X-ray microanalyzer (EPMA, Camea SX50). 103

Samples were also analyzed by X-ray powder diffractometer (XRD, Philips PW2273/20) using Cu-Kα 104

radiation for their qualitative mineralogical composition. 105

106

3. Results and discussion 107

3.1 Reduction with CO 108

3.1.1 Non-isothermal reduction 109

The leach residue was firstly reduced with CO in both TGA and DTA modes, the results of which are 110

plotted in Fig. 3 as a function of temperature. As can be seen, the reduction is negligible below 400 ºC. 111

Most reduction occurred in the temperature range 400–800 ºC. In this temperature range, two stages of 112

reduction could be identified. The first stage reduction is denoted by the initial quick mass loss between 113

4

400 and 520 ºC, corresponding to two partly overlapped CO2 peaks at 424 ºC and 455 ºC. There is an 114

endotherm at 405 ºC shown on the DTA curve (heat flow curve). This endotherm should represent the 115

first stage reduction reaction. However, there is a temperature discrepancy for this reduction measured by 116

TGA and DTA. This discrepancy is caused by the geometrical difference of the crucibles used for TGA 117

and DTA as well as the porous nature of the sample which tends to delay heat transfer. This discrepancy 118

was also observed and discussed in previous studies [3, 24]. The second reduction stage occurred between 119

520 and 800 ºC which is characterized by a quicker mass loss to reach -23.5 wt% with a corresponding 120

large CO2 peak and an endotherm at 593 ºC. 121

In order to identify what reduction reactions occurred in these two stages, samples were collected from 122

intermediate temperatures of 520 ºC and 800 ºC and were subject to XRD analysis. Their XRD patterns 123

are plotted in Fig. 4 along with the pattern for the original leach residue for comparison. As can be seen, 124

the phases identified in the original leach residue are hematite and nickel ferrite. Hematite disappeared 125

after the sample was heated to 520 ºC with the formation of substantial amount of magnetite (Fe3O4). The 126

formation of an alloy phase Fe10.8Ni is also observed. It becomes evident that one of the reactions in the 127

first stage is the reduction of hematite by CO to magnetite, which could be represented by Eq. (2). By 128

calculation, Eq. (2) could only cause 1.5 wt% mass loss. This suggests that the first small CO2 peak at 424 129

ºC is most likely caused by this reduction reaction. The second reaction which is represented by Eq. (3) is 130

responsible for the formation of the alloy Fe10.8Ni and the appearance of the second CO2 peak at 455 ºC. 131

Due to the presence of Ni as impurities of varying concentrations in the original hematite phase, reduction 132

via Eq. (2) would result in the formation of an iron-nickel oxide (NixFe3-xO4) instead of magnetite (Fe3O4). 133

As a consequence, the reduction product of Eq. (3) is an Fe-rich alloy phase (Fe10.8Ni) instead of pure Fe. 134

Iron oxide reduction is not complete at 520 ºC as suggested by the presence of substantial amounts of 135

magnetite at this temperature, Fig. 4. Relatively broad peaks for magnetite phase, which partly overlap 136

with those for NiFe2O4, can be seen at 520 ºC in Fig. 4. This is due to the presence of various amounts of 137

Ni in this phase (NixFe3-xO4) which caused the distortion of the lattice structure to varying degrees, 138

thereby shifting and/or broadening the XRD peaks. The completion of iron oxides reduction is marked by 139

the disappearance of the magnetite peak at 800ºC, indicating that magnetite reduction proceeds in both 140

reduction stages (400–520 ºC, 520–800 ºC). This is probably caused by the presence of two morphologies 141

of the original hematite phase which gives rise to the difference in their reduction kinetics. The first type 142

has a rather porous columnar structure (e.g. Particle a in Fig. 2), which has easy access to the reduction 143

gas. The second type is relatively dense and co-exists with nickel ferrite in the ring-shaped particles (e.g. 144

Particle b in Fig. 2). The reduction of the latter particles should be slower because of the dense nature of 145

solid phase(s). As a result, it is tentatively suggested that the reduction of the porous magnetite occurred 146

at around 455 ºC, and the denser magnetite in the ring-shaped particles was reduced in between 520 and 147

800 ºC. 148

3Fe2O3(s) + CO(g) = 2Fe3O4(s) + CO2(g) (2) 149

Fe3O4(s) + 4CO(g) = 3Fe(s) + 4CO2(g) (3) 150

In the second reduction stage (520–800 ºC), nickel ferrite is also reduced, as indicated by the 151

disappearance of nickel ferrite and the formation of an alloy phase Fe0.64Ni0.36 in Fig. 4. The reduction of 152

nickel ferrite could be represented by Eq. (4). 153

NiFe2O4(s) + 4CO(g) = Fe2Ni(alloy) + 4CO2(g) (4) 154

5

Further increase in temperature above 800 ºC caused little mass loss before 1200 ºC. On the contrary, a 155

slight mass gain is observed in this temperature range, which is believed to be caused by the carbon 156

deposition onto the sample and the crucible by the reverse Boudouard reaction (Eq. (1)). The gradual drift 157

of the base of the CO2 curve in Fig. 3 also indicates the occurrence of the reverse Boudouard reaction. 158

The base of the heat flow curve in Fig. 3 has drifted, which is caused by the physical changes of the 159

sample during the course of reduction, i.e. decrease in the heat capacity of the sample and/or increase in 160

its thermal conduction rate partly due to the shrink of the sample bed. 161

Further mass loss occurred after 1200 ºC, following an endothermic peak. A previous study [3] showed 162

that the silicate materials in the sample melted at around 1200 ºC, suggesting that the reduction of the Fe 163

from the molten silicates (Eq. (5)) should be responsible for this mass loss. The reduced product Fe 164

should be in the form of alloy by merging with the existed alloy phases. 165

Fe2SiO4(l) + 2CO(g) = 2Fe(alloy) + SiO2(s) + 2CO2(g) (5) 166

3.1.2 Isothermal reduction 167

Isothermal tests were performed to evaluate the reduction kinetics at various temperatures in the range 168

400–1300 ºC. The mass changes were plotted in Fig. 5. As can be seen, the reduction rate increased with 169

the increase of temperature from 400 to 1100 ºC indicated by the greater slope of the mass change curve 170

at higher temperatures. At 1100 ºC, the reduction was complete within around 3 min. The occurrence of 171

the reverse Boudouard reaction (Eq. (1)) is seen in the isothermal reduction runs in the temperature range 172

800–1100 ºC which is revealed by the slight mass gain after the reduction is complete or near complete. 173

Two distinct reduction stages could be clearly identified for the isothermal run at 1200 ºC. At this 174

temperature, the reduction rate is high and comparable to those at 1000 and 1100 ºC in the first 45 175

seconds. A sudden drop in the reduction rate took place at 45 seconds, after which the reduction is slow 176

and only comparable to that at 400 ºC. As discussed earlier that the silicates material in the sample begins 177

to melt at around 1200 ºC. This sudden drop in the reduction rate is believed to be caused by the melting 178

of the silicates in the sample. In the first stage where the reduction is fast, the sample is rather porous due 179

to the presence of large volume percentage of oxides as solid particles even though the silicates may have 180

melted. Substantial melting of silicates at around 45 seconds results in the collapse of the porous structure, 181

as well as pore blockage by liquid, both of which lead to significant reduction in the reactive surface area. 182

The diffusion through the molten silicates can soon become the rate controlling step, hence the sudden 183

drop in the reduction rate. This phenomena was also observed when using hydrogen as the reductant in 184

the previous study [3]. This sequence of events resembles the reduction of iron ore in the blast furnace 185

when fayalitic liquid is formed as a coating layer on the ore surface, severely limiting the reduction rate in 186

a phenomenon known as slag blocking [25]. At 1300 ºC, the quick reduction stage is even shorter. But 187

due to the higher diffusion rate in the silicate melt, the reduction is faster compared with the reduction at 188

1200 ºC in the second reduction stage. 189

The kinetics of the isothermal reduction was evaluated by analyzing the mass change curves in Fig. 5. Eq. 190

(6) [26] was adopted for this purpose, in which α is the extent of reduction and t is time. The left side of 191

the equation represents the rate of reduction, which is affected by two variables, i.e. temperature (T) and 192

extent of reduction (α). Arrhenius equation (Eq. (7)) was used to describe the temperature dependence. 193

The dependence of reduction rate on the extent of reduction is represented by the kinetic model f(α), the 194

form of which is determined by its rate controlling mechanism. For an isothermal program, the integral 195

6

form of the kinetic model g(α) has the relationship with temperature (T) and time (t) which is expressed 196

as Eq. (8) [26]. Isoconversional (model-free) method was firstly employed to calculate the apparent 197

activation energies without determining the kinetic models. Eq. (9) can be easily derived from Eq. (8), in 198

which tα,i represents the time required to reach certain extent of reduction (α) at temperature (Ti). 199

Apparent activation energy at certain extent of reduction can be calculated from the slope of the plot lntα,i 200

v.s. 1/Ti. The mass change curves for the isothermal reduction tests at 700–1100 ºC were evaluated by 201

applying Eq. (9) at various extents of reduction (α, 0.2–0.9). Evaluation was not performed for α<0.2 202

because of the relatively large error arising from the ambiguity in determining the starting time of the 203

reduction from the TGA curves. Apparent activation energies were calculated and plotted against α in Fig. 204

6. The plot of lntα,i v.s. 1/Ti is also shown in Fig. 6 as an inset. As can be seen, the apparent activation 205

energy increases from 6.7 kJ/mol to 41.1 kJ/mol with the progress of the reduction. This large variation of 206

apparent activation energy suggests that these Ea are composite values resulting from the tangled interplay 207

of different steps/processes. These steps/processes could be species-dependent chemical steps, gaseous 208

reactants/products diffusion through a solid product/reactant, adsorption–desorption of gaseous 209

products/reactants on the surface of the reacting solid, etc. [27, 28]. The apparent activation energies are 210

relatively low, indicating the rate controlling mechanism is very likely diffusion control rather than 211

chemical control. In order to determine whether the diffusion of CO through the pores of the sample bed 212

is the rate-controlling factor, two series of TGA experiments were performed. In the first series of 213

experiments, 50 mg leach residue was isothermally reduced at 800 ºC with varying flowrates of CO. The 214

results are plotted in Fig. 7. As can be seen, three mass change curves overlap, indicating that the transfer 215

of CO to the surface of sample bed at 1 L/min was high enough and did not adversely affect the reduction 216

rate. Variation of sample size was then evaluated in the second series of isothermal experiments at 800 ºC, 217

the results of which are shown in Fig. 8. Higher reduction rate can be observed with smaller sample size 218

especially within the first minute of reduction, indicated by the varying slopes of the mass change curves. 219

This suggests that the reduction did not take place uniformly in the sample bed, which resulted from the 220

limited mass transfer (gaseous transfer) in the pores of the sample bed during reduction. The effect of this 221

rate controlling mechanism gradually diminished with the progress of reduction, as suggested by the 222

parallel of the curves after approximately 1 minute. 223

)f(k(T)dt

d

(6) 224

)RT

E-exp(Ak(T) a (7) 225

t*)RT

Eexp(*A)dt

RT

Eexp(*A

)f(

dα)g( a

t

0

aα

0

(8) 226

i

a

RT

E

A

g

)(lnlnt i, (9) 227

Model fitting method was also used to further determine the rate controlling mechanism. Linear trends are 228

obtained between the reduction rate and f(α) for the isothermal tests in the temperature range 500–1100 ºC 229

by employing the 2D diffusion model, which is represented by Eq. (10) [26]. These linear trends are 230

7

exhibited in Fig. 9 as an inset. For each individual test, 2D diffusion model applies only when α is higher 231

than a certain value, which is shown in Fig. 9. This is because in the lower range of α, reduction rate is 232

predominantly controlled by the gaseous diffusion in the pores which is discussed earlier. Since the 2D 233

diffusion model represents the diffusion in the solid product layer [29], this indicates that in the higher 234

range of α where 2D diffusion model applies, diffusion of reactive species through metallic alloys in 235

individual particles controls the reduction rate. Arrhenius plot is also shown in Fig. 9 yielding the 236

apparent activation energy of 31.8 kJ/mol based on the slope of the line. This corresponds to a value when 237

α is approximately 0.8 in Fig. 6. The plot for 500 ºC deviated from the linear trend in Fig. 9. This 238

deviation is possibly caused by the partial reduction at low temperatures due to the stepwise reduction 239

mechanism (Eqs. (2–4)), whereas at other higher temperatures the reduction is near completion. From the 240

above discussion, gaseous diffusion through the pores of the sample bed mainly controls the reduction 241

rate at lower range of α, and the rate controlling factor gradually shifts to the diffusion through the layer 242

of alloy products in individual particles at higher range of α. Large variation of activation energies with α 243

in Fig. 6 is a result of the combining effect of the two rate-controlling mechanisms as well as their varying 244

contributions to the overall reduction rate. 245

f(α)=[-ln(1-α)]-1 (10) 246

The reduction products from the isothermal tests were examined by SEM/EDS. Fig. 10 illustrates the 247

microstructures of the porous hematite particles after reduction by CO at various temperatures. The 248

hematite was not reduced at 400 ºC and its columnar porous structure was preserved (top-left). At 500 ºC, 249

the hematite was reduced to form highly porous Fe particles with submicron grains (top-right). The grains 250

grow coarser and further apart when reduced at higher temperatures (bottom-left and bottom-right). Some 251

sintering could be observed in the reduced hematite particles at 900 ºC. 252

The change of morphologies of the ring-shaped particles reduced at various temperatures was also 253

examined and is presented in Fig. 11. Nickel sulfide resided within some of the ring-shaped particles, 254

which can be seen in Fig. 11. The particle at 400 ºC in Fig. 11 was not reduced. And the oxide ring is 255

much denser compared with the hematite particle in Fig. 10. Reduction was obvious when the 256

temperature reached 500 ºC and the ring formed Fe-Ni alloy with micron-sized pores. At 600 and 700 ºC, 257

the alloy rings were still rather porous, and the residual nickel sulfide was partly converted to nickel-rich 258

Ni-Fe alloy. This reaction is tentatively suggested as the reduction of the nickel sulfide to form alloy with 259

carbonyl sulfide (COS) as the gaseous product, which is represented by Eq. (11). Crowe and Utigard [30] 260

have reported this reaction as being responsible for the reduction of Ni, Cu, and Co sulfides by CO. 261

Above 800 ºC in Fig. 11, substantial sintering took place, and the degree of sintering is greater at higher 262

temperatures. For the sample reduced at 900 ºC, some Fe3O4 was found locked in the alloy phase. This 263

was caused by the immediate sintering of the alloys that were formed on the surface of the oxide which 264

limited the reduction rate of the oxide. The residual nickel sulfide was fully reduced to nickel-rich alloy 265

which sintered on the alloy ring. At 1100 ºC, the reduced ring was adequately densified and formed a 266

uniform alloy particle. 267

Ni3S2(s) + 2CO(g) = 3Ni(alloy) + 2COS(g) (11) 268

The silicates melted at 1200 ºC and all pores were removed. The alloy phase formed during this process 269

was distributed in the silicate melt as small particles of various shapes. Two silicates were formed from 270

the exsolution of the melt upon cooling, i.e. pyroxene which is rich in Fe and Al, and olivine which is rich 271

8

in Fe and Mg. The silicates phases formed were found to have higher Fe content than the original silicates 272

in the leach residue before reduction. This suggests that part of the iron oxide in the leach residue must 273

have dissolved into the silicate melt. This is further supported by the fact that a large number of small 274

regular-shaped Fe2O3 particles were exsolved from the silicate melt upon cooling, which could be seen in 275

the BSE image at 1200 ºC in Fig. 11. The partial dissolution of the hematite into the silicate melt which 276

substantially lowered the activity of hematite might be another critical factor that led to the slow reduction 277

rate at 1200 ºC (Fig. 5). In this case, the reduction is better represented by Eq. (5) in which Fe is formed 278

from the reduction of the silicate melt and merges with the existing alloy particles. At 1300 ºC, more 279

alloys were formed and the two silicate phases (pyroxene and olivine) exsolved from the melt formed a 280

laminar structure with no exsolution of Fe2O3 particles. This is probably caused by the greater degree of 281

reduction at higher temperature which resulted in an iron-deficient silicate melt. During reduction, gas 282

bubbles of CO2 were formed, which on solidification left behind features such as those seen in Fig. 12. 283

Upon cooling the melt, the gas bubbles shrink in volume, drawing the still-liquid silicate phase from the 284

already sintered alloy particles. On complete solidification, a spherical void and an alloy-free halo remain 285

at the bubble site. Because the exsolved pyroxene phase has a higher freezing temperature than the olivine 286

phase, the pyroxene solidifies before olivine, creating the unique laminar structure with parallel plates of 287

olivine separated by a gap, Fig. 12. 288

3.2 Reduction with graphite 289

3.2.1 Non-isothermal reduction 290

Fig. 13 shows the results for the reduction of the leach residue with excess amount of graphite powder by 291

continuously heating the mixture of the two at 15 ºC/min to 1500 ºC under argon atmosphere. As can be 292

seen, little reduction occurred below 800 ºC. Higher temperature (>800 ºC) is required to initiate the 293

reduction using graphite as reductant compared with the reduction using CO (>400 ºC). This is possibly 294

due to the different reduction mechanisms involved: initial reduction of the oxides by graphite is solid-295

solid reaction which has very limited reaction sites, with much smaller rate than the gas-solid reaction 296

involving CO. Moreover, even with the same area of reaction sites for two cases, the different reduction 297

mechanisms determine that the effect of temperature on the progress of reduction for both cases must be 298

different. Above 800 ºC in Fig. 13, the reduction progressed gradually, resulting in an increasing mass 299

loss and the formation of mainly CO2 as the gaseous product below 1000 ºC. Massive reduction took 300

place above 1000 ºC peaking at 1050 ºC, which resulted in large emission of CO and CO2. DTA analysis 301

measured a large endotherm at 1020 ºC which corresponds to this stage of reduction. The main reduction 302

reactions occurred in this stage are the reduction of hematite and nickel ferrite, represented by Eqs. (12–303

17). Gas-solid reduction reactions which are represented by Eqs. (14) and (17) are expected to take place 304

due to the formation of CO from Eqs. (12) and (15) as well as from the Boudouard reaction (Eq. (1)). The 305

following scenario can be envisaged. The gas (CO and CO2) formed from within the sample bed needs to 306

diffuse upwards to the surface of the bed before it is swept away by the argon flow. As the CO and CO2 307

formed diffuse upwards through the pores of the sample bed, CO component could reduce oxides on its 308

path (Eqs. (14) and (17)), resulting in an increased partial pressure of CO2. While the CO2 component 309

could be reduced by the graphite particles on its path depending on the equilibrium of the Boudouard 310

reaction (Eq. (1)), resulting in the increased partial pressure of CO. The reversed Boudouard reaction 311

could also take place if the partial pressure of CO is higher than equilibrium. As a result, the relative 312

partial pressure of CO and CO2 in the offgas analyzed (Fig. 13) is a result of the competition among the 313

9

reversible Boudouard reaction (Eq.(1)), reduction reactions by graphite (Eqs. (12), (13), (15), (16)), and 314

reduction reactions by CO (Eqs. (14) and (17)). Due to the complexity of the reaction system, no further 315

effort was made to determine which type of reaction prevailed in the reduction process. 316

The Boudouard reaction, Eq. (1), was also expected to take place. The speculation is made based on the 317

fact that in Fig. 13, the formation of CO2 as the gaseous reduction product predominates below around 318

1000 ºC, above which the formation of CO prevails. This trend is in accordance with the equilibrium of 319

Boudouard reaction (Fig. 1), although the offgas CO/CO2 in Fig. 13 did not reach equilibrium because of 320

the non-equilibrium conditions applied. The reduction slowed down until the temperature reached 1200 321

ºC, above which a second stage quick reduction occurred. It peaked at 1300 ºC and ended at 1340 ºC. This 322

second stage of reduction resulted in the formation of CO as the gaseous product and corresponds to an 323

endotherm peak at 1235 ºC. This is believed to be induced by the melting of the silicates in the sample 324

which caused the reduction of the molten silicate by the graphite, which could be represented by Eq. (18). 325

Fe2O3(s) + 3C(s) = 2Fe(s) + 3CO(g) (12) 326

2Fe2O3(s) + 3C(s) = 4Fe(s) + 3CO2(g) (13) 327

Fe2O3(s) + 3CO(g) = 2Fe(s) + 3CO2(g) (14) 328

NiFe2O4(s) + 4C(s) = NiFe2(alloy) + 4CO(g) (15) 329

NiFe2O4(s) + 2C(s) = NiFe2(alloy) + 2CO2(g) (16) 330

NiFe2O4(s) + 4CO(g) = NiFe2(alloy) + 4CO2(g) (17) 331

Fe2SiO4(l) + 2C(s) = 2Fe(s,l) + SiO2(s) + 2CO(g) (18) 332

3.2.2 Isothermal reduction 333

In order to study the reduction kinetics at different temperatures, isothermal reduction tests were 334

performed in the temperature range 800–1400 ºC with the results shown in Fig. 14. And the reduction 335

products from these isothermal runs were examined by SEM/EDS, EPMA, and XRD to elucidate the 336

reduction mechanisms. As can be seen in Fig. 14, at low temperatures of 800 and 900 ºC, the reduction 337

rate is slow and very limited reduction took place within 30 min. Effective reduction only occurred when 338

the temperature is higher than 1000 ºC. Two quick reduction stages can be observed at this temperature, 339

which took place during 0–2 min and 4–7 min, respectively. This possibly resulted from the sequential 340

reduction of the porous hematite particles and the relatively dense ring-shaped particles which are 341

composed of both hematite and nickel ferrite. The reduction at 1100 ºC was faster and near complete 342

within 3 min. Two stages of reduction were also observed for the isothermal runs at and above 1200 ºC. 343

In this case, the first stage is always the quick reduction of the oxides until approximately 25 wt% mass 344

loss is reached, while the following second stage is the reduction of the molten silicate. It is apparent from 345

Fig. 14 that the molten silicate reduction stage is faster at higher temperatures, indicated by the higher 346

slope of the curves below -25 wt% at higher temperatures. 347

The alloy phases that were formed from these isothermal runs were analyzed by EPMA and the results are 348

presented in the ternary graph in Fig. 15. The alloy phases identified at 900 ºC under the electron probe 349

are all nickel-rich Ni-Fe alloys, which are formed from the reduction of the nickel sulfide. Previous study 350

10

showed that the nickel sulfide (Ni3S2) melted at 813 ºC [24]. As a result, it was in its molten state during 351

isothermal reduction at 900 ºC. The reaction is tentatively represented by Eq. (19). Reduction by CO (Eq. 352

(11)) could also take place if appreciable CO is present in the sample bed. Fig. 16 (a) exhibits a partly 353

reduced nickel sulfide particle forming nickel-rich alloy with a clear boundary between the alloy and 354

sulfide phases. Another alloy particle is shown in Fig. 16 (b) as the product from the reduction of the 355

nickel sulfide at 1000 ºC. It is evident from this particular morphology that both the alloy and the nickel 356

sulfide were in liquid state during reduction, the sulfide covering the surface of the alloy drop. 357

Ni3S2(l) + C(s) = 3Ni(alloy) + CS2(g) (19) 358

With the further increase of temperature, more iron-rich alloys were formed from the reduction of the 359

hematite and nickel ferrite, which can be seen from Fig. 15. Reduction of the silicates started at 1200 ºC 360

denoted by the presence of around 15 at% Si in the alloy phase. The silicates after isothermal reduction 361

were also analyzed by EPMA to reveal the progress of reduction, which is plotted in Fig. 17. The near 362

depletion of the iron from the silicates at 1200 ºC indicates that Fe is also reduced along with Si. The 363

reduction of Fe from the silicate is more complete at 1200 ºC and higher. The SiO2 content of the silicates 364

also dropped with the increase of temperature, indicating a larger extent of reduction at higher 365

temperatures in terms of Si removal. Fig. 18 (a) exhibits the nucleation and growth of Fe–Si alloy 366

particles within the silicate melt. The sizes of the particles range from nanometers to microns. The 367

exterior Fe-Ni alloy gradually absorbs some of these Fe-Si particles, particularly from the edge of the 368

silicate, to form a Fe-Ni-Si phase. Fig. 18 (b) shows a Fe-depleted silicate particle covered with a relative 369

thick layer of Fe-Ni-Si alloy at 1200 ºC. This suggests that the reduction of silicate melt could also 370

proceed by the diffusion of carbon or possibly CO through the alloy. 371

With the increase of the isothermal temperature from 1200 to 1400 ºC, the composition of the alloy 372

particles becomes more uniform, suggested by the converging trend of the plots in Fig. 15. This indicates 373

that alloy particles have a greater tendency to merge and form larger and more uniform particles at higher 374

temperatures, which is supported by the evidence from both the SEM and XRD analysis. Fig. 18 (c) 375

shows the spherical alloy particles produced at 1300 ºC. These particles could be as big as 100 µm in 376

diameter. Fig. 18 (d) shows an alloy particle 175 µm in diameter. Two phases were exsolved from the 377

alloy melt showing different grey levels. The interior brighter phase has more Fe and less Si. XRD 378

analysis was conducted on the samples produced from the isothermal runs at 1000 ºC, 1200 ºC and 1400 379

ºC with the results shown in Fig. 19. At 1000 ºC, the main alloys are Fe0.64Ni0.36 and Fe10.8Ni which were 380

produced from the reduction of the nickel ferrite and hematite, respectively. Due to the reduction of 381

molten silicate, Fe-Ni-Si alloy formed as the major alloy phase at 1200 ºC with minor amount of 382

Fe0.64Ni0.36. The alloy produced at 1400 ºC is a relatively homogeneous phase of Ni-Si-Fe, showing 383

enhanced merging of the alloy particles. 384

385

4. Conclusions 386

Reduction of the leach residue by both carbon monoxide and graphite was conducted using TGA and 387

DTA. The reduced products were analyzed by SEM/EDS, EPMA and XRD. Results show that the 388

reduction by CO initiated at around 400 ºC during the non-isothermal reduction tests. The reduction of the 389

hematite phase took place in two stages with magnetite as the intermediate products. Nickel ferrite was 390

11

reduced mostly in between 520 and 800 ºC resulting in the formation of an alloy phase Fe0.64Ni0.36. 391

Significant sintering of the alloys was observed above 800 ºC, and the extent of the sintering was greater 392

at higher temperatures. Below 1200 ºC, the isothermal reduction was firstly controlled by the gaseous 393

diffusion in the pores of the sample bed. More alloy products formed with the progress of reduction, 394

resulting in the gradual shift of the rate-controlling mechanism to the diffusion through the alloy products 395

in individual particles. This change of rate-controlling mechanism led to the increase of the apparent of 396

activation energies from 6.7 kJ/mol (α=0.2) to 41.1 kJ/mol (α=0.9). The siliceous gangue melted at 1200 397

ºC, which caused the partial dissolution of hematite into the silicate melt and removal of pores from the 398

sample bed. This resulted in slow reduction kinetics. 399

Substantial reduction of the leach residue by graphite powder took place only above 1000 ºC. The 400

formation of CO or CO2 as the main gaseous product is largely dependent on temperature, and is 401

suggested to be dominantly influenced by the Boudouard reaction. The Fe and Si components of the 402

silicates could be reduced by graphite above their melting temperature (1200 ºC) forming an Fe-Si alloy. 403

The alloys formed from the reduction of hematite, nickel ferrite or silicates tend to be more homogeneous 404

at higher temperatures resulting from the enhanced merging of the alloy particles. 405

406

Acknowledgements 407

The authors wish to thank Xstrata Nickel and Vale Canada Ltd. for sponsoring the project. The funding 408

for this research was provided by the Natural Sciences and Engineering Research Council (NSERC) of 409

Canada. 410

411

References 412

[1] D. Yu, T.A. Utigard, M. Barati, Fluidized bed selective oxidation-sulfation roasting of nickel sulfide 413 concentrate: part I. oxidation roasting, Metall. Mater. Trans. B, accepted for publication. 414 [2] D. Yu, T.A. Utigard, M. Barati, Fluidized bed selective oxidation-sulfation roasting of nickel sulfide 415 concentrate: part II. sulfation roasting, Metall. Mater. Trans. B, accepted for publication. 416 [3] D. Yu, M. Zhu, T.A. Utigard, M. Barati, TGA kinetic study on the hydrogen reduction of an iron 417 nickel oxide, Miner. Eng. (2013) http://dx.doi.org/10.1016/j.mineng.2013.03.018. 418 [4] W.K. Jozwiak, E. Kaczmarek, T.P. Maniecki, W. Ignaczak, W. Maniukiewicz, Reduction behavior of 419 iron oxides in hydrogen and carbon monoxide atmospheres, Appl. Catal., A 326 (2007) 17-27. 420 [5] A.A. El-Geassy, M.I. Nasr, M.M. Hessien, Effect of reducing gas on the volume change during 421 reduction of iron oxide compacts, ISIJ Int. 36 (1996) 640-649. 422 [6] J.-m. Pang, P.-m. Guo, P. Zhao, C.-z. Cao, D.-w. Zhang, Influence of size of hematite powder on its 423 reduction kinetics by H2 at low temperature, J. Iron Steel Res. Int. 16 (2009) 07-11. 424 [7] A.A. El-Geassy, M.I. Nasr, Influence of original structure on the kinetics and mechanisms of carbon 425 monoxide reduction of hematite compacts, ISIJ Int. 30 (1990) 417-425. 426 [8] H.W. Kang, W.S. Chung, T. Murayama, Effect of iron ore size on kinetics of gaseous reduction rate, 427 ISIJ Int. 38 (1998) 109-115. 428 [9] H.W. Kang, W.S. Chung, T. Murayama, Effect of iron ore shape on gaseous reduction rate, ISIJ Int. 429 38 (1998) 1194-1200. 430 [10] K. Higuchi, R.H. Heerema, Influence of sintering conditions on the reduction behaviour of pure 431 hematite compacts, Miner. Eng. 16 (2003) 463-477. 432

12

[11] C.E. Seaton, J.S. Foster, J. Velasco, Structural changes occurring during reduction of hematite and 433 magnetite pellets containing coal char, Trans. Iron Steel Inst. Jpn. 23 (1983) 497-503. 434 [12] B. Weiss, J. Sturn, S. Vogslam, S. Strobl, H. Mali, F. Winter, J. Schenk, Structural and 435 morphological changes during reduction of hematite to magnetite and wustite in hydrogen rich reduction 436 gases under fluidised bed conditions, Ironmaking Steelmaking 38 (2011) 65-73. 437 [13] A.V. Bradshaw, A.G. Matyas, Structural changes and kinetics in the gaseous reduction of hematite, 438 Metall. Trans. B 7B (1976) 81-87. 439 [14] K.S. Abdel-Halim, M. Bahgat, O.A. Fouad, Thermal synthesis of nanocrystalline fcc Fe-Ni alloy by 440 gaseous reduction of coprecipitated NiFe2O4 from secondary resources, Mater. Sci. Technol. 22 (2006) 441 1396-1400. 442 [15] M.O. Boudouard, Recherches sur les equilibres chimiques, Ann. Chim. Phys. 24 (1901) 1-85. 443 [16] A. Roine, HSC Chemistry, Outokumpu Research Oy, Pori, Finland, 2007. 444 [17] R. Haque, H.S. Ray, Role of ore/carbon contact and direct reduction in the reduction of iron oxide by 445 carbon, Metall. Mater. Trans. B 26B (1995) 400-401. 446 [18] M.S. Bafghi, M. Fukuda, Y. Ito, S. Yamada, M. Sano, Effect of CO gas formation on reduction rate 447 of iron oxide in molten slag by graphite, ISIJ Int. 33 (1993) 1125-1130. 448 [19] J. Moon, V. Sahajwalla, Investigation into the role of the Boudouard reaction in self-reducing iron 449 oxide and carbon briquettes, Metall. Mater. Trans. B 37B (2006) 215-221. 450 [20] K. Mondal, H. Lorethova, E. Hippo, T. Wiltowski, S.B. Lalvani, Reduction of iron oxide in carbon 451 monoxide atmosphere - reaction controlled kinetics, Fuel Process. Technol. 86 (2004) 33-47. 452 [21] S.K. Sharma, F.J. Vastola, J. P.L. Walker, Reduction of nickel oxide by carbon: I. Interaction 453 between nickel oxide and pyrolytic graphite, Carbon 34 (1996) 1407-1412. 454 [22] S.K. Sharma, F.J. Vastola, J. P.L. Walker, Reduction of nickel oxide by carbon: II. Interaction 455 between nickel oxide and natural graphite, Carbon 35 (1997) 529-533. 456 [23] S.K. Sharma, F.J. Vastola, J. P.L. Walker, Reduction of nickel oxide by carbon: III. Kinetic studies 457 of the interaction between nickel oxide and natural graphite, Carbon 35 (1997) 535-541. 458 [24] D. Yu, T.A. Utigard, TG/DTA study on the oxidation of nickel concentrate, Thermochim. Acta 533 459 (2012) 56-65. 460 [25] G.C. Ulmer, W.C. Elliott, T. Buntin, J. Edwin S. Erickson, J.J. Friel, Role of selected cations and gas 461 speciation on the reduction of fayalite at 1300ºC, J. Am. Ceram. Soc. 75 (1992) 1476-1483. 462 [26] S. Vyazovkin, A.K. Burnham, J.M. Criado, L.A. Perez-Maqueda, C. Popescu, N. Sbirrazzuoli, 463 ICTAC Kinetics Committee recommendations for performing kinetic computations on thermal analysis 464 data, Thermochim. Acta 520 (2011) 1-19. 465 [27] S. Vyazovkin, Kinetic concepts of thermally stimulated reactions in solids: A view from a historical 466 perspective, Int. Rev. Phys. Chem. 19 (2000) 45-60. 467 [28] A.K. Galwey, What is meant by the term ‘variable activation energy’ when applied in the kinetic 468 analyses of solid state decompositions (crystolysis reactions)?, Thermochim. Acta 397 (2003) 249-268. 469 [29] A. Khawam, D.R. Flanagan, Solid-State Kinetic Models:  Basics and Mathematical Fundamentals, J. 470 Phys. Chem. B 110 (2006) 17315-17328. 471 [30] C.J.B. Crowe, T.A. Utigard, Kinetics of carbon monoxide reduction of oxidized calcines, Can. 472 Metall. Q. 42 (2003) 447-454. 473

474

475

Figures 476

13

477

Fig. 1. Stability diagram for Fe-O system under CO-CO2 atmospheres (solid lines) and CO partial 478

pressure established by the Boudouard reaction (dashed line). 479

480

Fig. 2. BSE image of the leach residue [3]. 481

14

482

Fig. 3. Non-isothermal reduction (15 ºC/min) of the leach residue in CO. 483

484

Fig. 4. XRD patterns for the leach residue (a), and samples collected after non-isothermal reduction in CO 485

to 520 ºC (b) and 800 ºC (c). 486

15

487

Fig. 5. TGA isothermal reduction of the leach residue with CO. 488

489

Fig. 6. Variation of apparent activation energies as a function of extent of reduction (α) for the isothermal 490

reduction tests. lnt v.s.1/T ×104 (isoconversional method) is also plotted as an inset. 491

16

492

Fig. 7. Effect of the variation of the flowrates of CO on the isothermal reduction of the leach residue. 493

494

Fig. 8. Effect of the variation of sample sizes on the isothermal reduction of the leach residue. 495

17

496

Fig. 9. Arrhenius plot for the isothermal reduction tests between 500 and 1100 ºC. The relationship 497

between the reduction rate and the 2D diffusion model (f(α)=[-ln(1-α)]-1) is also plotted as an inset. 498

499

Fig. 10. Isothermal reduction of the porous hematite particles with CO at various temperatures. 500

18

501

19

Fig. 11. Microstructure of particles reduced by CO under isothermal conditions (Px: Pyroxene, silicate 502

containing Fe, Al, Na; Ol: Olivine, silicate containing Fe, Mg). 503

504

Fig. 12. Laminar structure of the silicates formed upon cooling the reduced residue from 1300 ºC with a 505

gas bubble formed in the melt (left) and interior of the bubble (right). 506

507

Fig. 13. Non-isothermal reduction (15 ºC/min) of the leach residue with graphite in TGA and DTA. 508

509

Fig. 14. TGA isothermal reduction of the leach residue with graphite. 510

20

511

Fig. 15. EPMA analysis on the alloy particles formed from the isothermal reduction tests by graphite. 512

513

Fig. 16. (a) Partial conversion of the nickel sulfide to alloy at 900 ºC; (b) Alloy surrounded by 514

monosulfide solid solution (Mss, (Ni,Fe)S) reduced at 1000 ºC (Px: Pyroxene, silicate containing Fe, Al). 515

516

Fig. 17. The composition of silicate phases in isothermal reduction by graphite. 517

21

518

Fig. 18. (a) Formation of SixFe alloy particles from the silicate melt at 1200 ºC (Px: Pyroxene, silicate 519

containing Al, Mg, Fe); (b) Reduction of Fe and Si from a silicate particle forming Ni-Si-Fe alloy at 1200 520

ºC (Px: Pyroxene, silicate containing Al, Mg, Na); (c) NixSiyFe alloy formed at 1300 ºC; (d) An NixSiyFe 521

alloy particle formed from the reduction at 1400 ºC. 522

523

Fig. 19. XRD analysis on the product from the reduction at 1000 ºC, 1200 ºC and 1400 ºC. 524

525

Table 526

Table 1. Chemical and mineralogical compositions of the leach residue. 527

Chemical composition

(wt%)

Fe Ni S Cu Co Mg Al Si O

49.3 10.5 1.1 1.0 0.06 1.1 0.9 5.1 30.94

Mineralogical

composition (wt%)

Fe2O3 NiFe2O4 NiS CuO CoO MgO Al2O3 SiO2

47.4 33.7 3.1 1.2 0.08 0.8 1.8 10.8

528