(Received on December 20, 2011; accepted on February 10, 2012)
The sampling of raw materials in the rotary kiln, firing experiment by the experimental kiln, and waterquenching experiment have been performed, and the clarification of behavior of the reduction and thegrowth of metal has been attempted by SEM-EDS. In the amorphous serpentine region, NiO has very highreducibility than FeO. The increase in heating temperature above 1 273 K allows the fine reduced metal tobe confined in the silicate, which causes the extraction by the bromine methyl alcoholic solution to be dif-ficult. This allows the degree of reduction to enter apparently the depression region. However, theincrease in heating temperature above 1 525 K allows melt to occur, which causes the fine metal tocoalesce with each other. Therefore, the degree of reduction leaves the depression region, andapproaches the equilibrium values. Low-MgO and high-FeO·NiO silicate is enriched by the fractional crys-tallization, and thereafter reduction reaction is enhanced. From the result of SEM observation that metalsoccur from the lower temperature in the low-MgO and high-FeO ore containing much point defect, NiOand FeO in the crystal lattice are reduced via cationic and electronic defect species, the oxygen occurreddiffuses toward the crystal surface via vacancies. However, the lack of experimental data of defect chem-istry in Ni-ore requires the further investigation. From the fact that the temperature of the melt occurrencecoincides with the temperature of metal beginning to grow largely in SEM observation, it is confirmed thatthe fine metals coalesce with each other via melt to grow.
KEY WORDS: saprolite Ni-ore; reduction; fractional crystallization; fine metal particles; growth of metal;point defects; bromine methyl alcoholic method.
1. Introduction
The pyrometallurgical smelting of saprolite Ni-ore(henceforth, Ni-ore), which is weathered ultramafic rockand mainly composed of serpentine, has been generallyperformed by the Elkem process, in which the melting-reduction is conducted in the electric furnace (around1 600°C), enabling the high Ni-recovery of around 96 mass%to be acquired.
On the other hand, in “Nippon Yakin Oheyama Process,”which is the improved Krupp-Renn process, Ni-ore is smeltedwith the semi-fused state at a maximum temperature ofaround 1 400°C. Subsequently, clinker discharged from thekiln is water-quenched to obtain easy crushing. Further,clinker is crushed, thereafter divided into the metal and slagby jigs and magnetic separaters: metal is granular ferro-nickel contained Fe and Ni around 75–80 and 25–20 mass%respectively; particle size ≧ 0.1 mm φ; average particle sizearound 1mm φ. In this case, maintenance of constructionmaterials standard and reductions of the power cost requirethe restriction of degree of crushing. This causes fine ferro-nickel (–a few μm) contained in clinker to be taken in slagand not be recovered as metal, which results in lower Ni
recovery than the Elkem process. Hence, the improvement ofNi-recovery requires the growth of fine-grain metal duringsmelting process in the rotary kiln. In particular, recently,there is the tendency for Ni-grade to lower due to conserva-tion of resources. As a result, some Ni-grade in availableores lower than 2 mass%, which further causes low Ni-recovery.
Many studies have been performed concerning the reduc-tion of Ni-ores and the growth of grain size of the reducedmetal. For example, Shirane1–3) has reported that the reduc-tion of Ni in the (Mg–Ni) silicate depends on the content ofMgO. Further, Johannsen,4) who has developed the Krupp-Renn process, has deduced for the growth of metal as fol-lows: 1) metals contained in heated raw materials in themetal growing zone in the rotary kiln are re-oxidized by theair contained in a gas phase, causing metal to melt due tothe heat of oxidation of Fe, which allows metals to coalescewith each other and grow larger; 2) Fe plays a role of Ni col-lector, which allows metals to grow largely. Furthermore,Matsumori5) has proposed that the fine metal wrapped withsulfur can coalesce with each other due to the reduction ofmelting temperature of the surface layer, metal growinglargely. However, Ni-ore is the mixture of several minerals,
furthermore chemical reactions in the rotary kiln are verycomplicated, which allows those technical subjects to bestill left unsolved. Whereas recently, Kobayashi,6–8) Tsuji9)
have performed the mineralogical examination concerningthe softening behavior related to the reduction of ores andgrowth of metal, thus having found out that point defectsformed during the weathering, CaO, Al2O3, and (MgO/SiO2)in ore have a large significant effect on softening. However,the behavior of the reduction and the growth of metal havenot been still made fully clear. Therefore, clarification hasbeen attempted by the reduction experiments and SEM-EDS, focusing attention on the fine metal particles of sev-eral tens μm or less.
2. Experimental Procedure
2.1. Sampling of Raw Materials in the Rotary Kiln outof Operation
First, raw materials of around 20–30 kg were sampledevery 5–10 m in the rotary kiln, shown in Fig. 1(a), cooledin order to get an outline of how Ni ore is reduced: aboutthree days is needed for cooling after being out of operation.The samples were crashed, and separated into slag and met-al, chemical analysis being performed. Here, Table 1 showsthe chemical analysis values of charged raw materials.
2.2. Firing Experiment by the Experimental KilnIt is impossible to take the sample of raw materials from
the rotary kiln in operation. Then, Ore-NC shown in Table1 were blended with anthracite (C, hardcoal) and limestone(80 kg/t of ore), and thereafter 700 kg-dry briquette of pil-low type 30×25×15 mm were produced: the additions ofanthracite is 4 A0 about 180 kg/t of ore for the consumptionof anthracite by CO2 contained in the LPG combustion gas-es; A0 is the quantity of anthracite for the reduction of Niand Fe in Ni-ore. Kiln simulation experiments have been
performed using the batch type experimental kiln shown inFig. 1(b): air ratio 1.0; LPG 30 m3/h; kiln rotation 20 rph.Air for combustion is automatically controlled, and damperopening is manually operated for the oxygen concentrationof 0 (vol%) in the exhaust-gas, then raw materials beingheated until the occurrence of softening and melting. Theresulting time for experiment was taken around 7 h. The 2–3 kg of raw materials samples in the experimental kiln weretaken every 0.5–1.0 h from the sampling hole of 200 mm indiameter, directly being quenched into the water. That wasdried, crashed and separated slag and metal, the chemicalanalysis being performed.
2.3. Water Quenching Experiment10)
Supplement for firing experiment by the experimentalkiln requires the understanding of the exact relation betweentemperature and degree of reduction. Then, ore-anthracite(1.2 A0), limestone composite pellets were prepared usingOre-A shown in Table 1, thus being heated in the verticalwater quenching furnace (NIKKATO CORPORATION)under the argon gas flow of 200 cc/min. Water quenchedpellets were dried, embedded in resin and polished, thenSEM observation and EDS analysis have been carried outwith SHIMAZU SSX-550. The remaining samples werecrushed in –44 μm to be analyzed by the bromine methyl
Table 1. Chemical analysis values of ores and raw materials(mass%).
Ore C SiO2 Fe Al2O3 Ni CaO MgO
Charge* 8.98 40.43 10.78 1.82 2.21 3.60 20.06
NC – 44.36 9.05 0.41 2.45 0.08 26.68
A – 40.58 14.67 1.57 2.07 0.14 20.98
B – 39.53 9.84 0.49 2.29 0.12 30.08
*in actual rotary kiln
(a) Simplified schematic illustration of rotary kiln (72 mL×3.6 mφ)
3.1. Sampling of Raw Materials in the Actual KilnFigure 2(a) shows the relation between degree of Ni and
Fe reduction, C content in the slag and distance from the dis-charge end. The degree of Ni and Fe reduction begins toincrease from around 40 m (raw materials temperature ofaround 1 000°C), but from around 25 m the degree of Nireduction has relatively gentle gradients and that of Fe doesnot has large change. That is, the degree of reduction entersthe plateau region. However, from around 15 m the degreeof both Ni and Fe reduction increases with steeper gradientsin spite of very low C content in the slag, and simultaneous-ly +0.105 mm metal (mass%) also increase. This indicatesthat reduction reaction does not occur in that region, but thefine metals grow to the size enough to be extracted by thebromine methyl alcohol solution, thus allowing the apparentdegree of reduction to approach the equilibrium values.
Here, the return to the kiln from the dressing processprovides the existence of metallic Ni and Fe in around 50–70 m in Figs. 2(a) and 2(b). The transportation of fine metalfrom raw materials to ferro-nickel products causes M–Fe inthe slag to begin to decrease rapidly from around 15 m, still
more, re-oxidation during the cooling of kiln leads to theconsiderably high content of Fe3+ in Fig. 2(b).
3.2. Experimental KilnFigures 3(a) and 3(b) show the presence of the depression
region in the degree of Ni and Fe reduction between 4 h(raw materials temperature of around 1 120°C) and 5 h(around 1 200°C) in spite of the considerably high residualC content of 8–10 mass% in the slag. Furthermore, thedegree of reduction leaves the depression region, and againbegins to increase from 5 h to reach the peak at 5.5 h. How-ever, the low C content due to the consumption of anthracitecauses re-oxidation of metallic Ni and Fe in the process ofsampling. Table 2 shows the chemical analysis values.
3.3. Water Quenching ExperimentFigures 4(a) and 4(b) show the degree of reduction and
the reduction behavior of Ni and Fe in heated ore at the var-ious temperatures, and Table 3 shows the chemical analysisvalues. The completion of reduction of Fe2O3 to FeO ataround 900°C is followed by the beginning of reduction ofFeO to Fe from around 1 000°C, thus the degree of Fereduction quickly increasing. On the other hand, the reduc-tion of NiO to Ni quickly begins from around 900°C as wellas Fe, the degree of Ni reduction reaching as much as
Fig. 2. Chemical analysis of the raw materials in the rotary kiln.Fig. 3. Chemical analysis of the quenched sample in the experi-
around 70 mass%. This, hence, shows that Ni has higherreducibility than Fe.
The increase in heating temperature above around1 000°C allows the degree of reduction of the waterquenched raw materials to decrease apparently and to enterthe depression region as well as firing experiment by theexperimental kiln. However, re-oxidation of Ni and Fe does
not occur beyond 1 300°C. Simultaneously, the degree of Niand Fe reduction increase in spite of the very low residualC content in the slag. This seems to be almost the same phe-nomena as that of raw materials in the actual rotary kiln.
4. Discussion
4.1. Calculated Degree of Ni and Fe Reduction in Equi-librium in Ni-ore
Then, the degree of Ni and Fe reduction in equilibrium iscalculated using the thermodynamic software packageMELTS11) in fixed oxygen partial pressure (Po2=10–10–10–13)at 1 100–1 400°C to examine whether or not the measure-ment values are reasonable. As for the activity of each oxidein this software, pure solid is taken as the standard state.Regular solution is approximated for liquid phase. Here,Mg, Ni and Fe share the cation component in silicate, the
Table 2. Chemical analysis values of the raw materials in the experimental kiln (mass%).
Slag (mass%) Metal (mass%) Weight ratio Reduction degree
Time (h) C T–Fe M–Fe Fe2+ Fe3+ T–Ni M–Ni C Fe Ni slag metal Fe Ni
Ni content of 2.0–2.4 mass% not largely varying. Hence,with maintenance of the value of components other than Feand MgO having a great influence on reduction, the molarratio Fe/(Mg+Fe) in the low-MgO and high-FeO Ore-Ashown in Table 1 is changed into the range of 0.1–0.4, thusthe degree of Ni and Fe reduction being calculated. Where,a small amount of Fe in goethite (FeO·OH) in Ni-ores couldallow the assumption that those are incorporated in the sil-icate. Figures 5(a) and 5(b) show that the decrease in valueof Ni/(Mg+Ni+Fe) in the cation component due to theincrease in Fe content leads to the tendency for the degreeof Ni reduction to decrease. The degree of Ni and Fe reduc-tion are around 96, 68 mass% at 1 400°C, respectively, at thevalue of Fe/(Mg+Fe)=0.3 near Ore-A.
Both of degrees of Ni and Fe reduction in equilibriumhave larger values than that determined using the brominemethyl alcohol method, in particular, Ni having serious dif-ferences.
4.2. Differences Between Equilibrium Values and Mea-sured Values of the Degree of Ni and Fe Reduction
4.2.1. Quantitative Determination of Metallic Ni and Feby Bromine Methyl Alcohol Method
Then, bromine methyl alcohol method is examined
whether or not above-mentioned differences are attributed toa chemical analysis method. To be sure, that has high reac-tivity with metal separated from the slag according to Eq.(1), but low reactivity with the metal confined in the silicate.
Ni + Br2 + 2CH3OH=Ni(OCH3)2 + 2HBr......... (1)
Then, the observation by an optical microscope of residualprepared by filtration the bromine methyl alcoholic solutionreveals that around –2 μm metals remain without beingextracted as shown in Fig. 6.
As expected, there is a high possibility that the fine metal-lic Ni particles confined to the silicate leads to the lowdegree of Ni reduction than true values. Similarly,Okajima12) reported that the metallic Ni particle of around0.6 μm confined to (Mg–Ni) olivine heated at 1 000°C isobserved by EPMA, which probably causes the degree of Nireduction to decrease.
4.2.2. Difference in the Experimental KilnIt seems that fine metallic Ni and Fe being confined in the
silicate leads to the difficulty in the extraction by the bro-mine methyl alcoholic solution, allowing the occurrence ofthe depression region of the degree of reduction. Thereduced fine metals are scattered in the silicate under con-ditions high temperatures (i.e., 1 100–1 200°C). Waterquenching in that state allows fine metal to be enveloped inthe silicate, which metallic Ni and Fe not to be accuratelyanalyzed. However, the occurrence of melt from 5 h causesthe fine metal to grow to the size enough to be extracted bythe bromine methyl alcoholic solution, thus the degree ofreduction approaching the equilibrium values. At the end ofexperiment, the decrease in residual C in the slag due to theconsumption of anthracite causes re-oxidation, thus thedegree of reduction decreasing.
4.2.3. Difference in the Water Quenching ExperimentIt is considered that amorphous serpentine, which occurs
by the dehydration from 600°C and is not still recrystallized,have high reducibility.1,2) Ni particularly has higher reduc-ibility than Fe as shown in the free energy change of Eqs.(2)–(5).13)
NiO+CO=Ni+CO2 ........................... (2)
Fig. 5. Relation between the degree of Ni and Fe reduction andmolar ratio of Fe/(Mg+Fe), temperature.
Fig. 6. Observation by an optical microscope of residual preparedby the filtration bromine methyl alcoholic solution.
However, the degree of Ni and Fe reduction enters thedepression region from around 1 000, 1 100°C, respectively.This seems to be not attributed to the re-oxidation but to thedifficulty in the extraction by the bromine methyl alcoholicsolution due to the fine metallic Ni and Fe being confinedin the silicate, because C remains as much as around2 mass% in the slag.
Here, reason for Ni entrance depression region from thelower temperature than Fe is considered as follows: 1) fromthe fact that the degree of Ni, Fe reduction at 900°C are10.0, 1.9 mass%, respectively, as shown in Fig. 4(a), thepreferential reduction of Ni allows the high-Ni and high-Femetal to be formed at initial stage of reduction; 2) the con-tent of Ni, Fe in Ni-ore is 2.07, 14.67 mass%, respectivelycauses the high-Ni metal of small grain size and high-Femetal of large grain size to be formed in spite of high Ni andlow Fe reducibility.
This concept can also demonstrate that the differencesbetween equilibrium values and measured values are con-siderably large and almost none for Ni and Fe, respectively:compared the equilibrium values Ni, Fe for 96.4, 34.6 mass%respectively at Fe/(Mg+Fe)=0.3 as shown in Figs. 5(a) and5(b) with the measured values Ni, Fe for 63.6, 33.7 mass%respectively as shown in Fig. 4(a) at 1 100°C.
4.2.4. Difference in the Actual Rotary KilnThere is not depression region, which is present in the
experimental kiln and water quench experiment, but plateauregion in the degree of reduction of raw materials sampledin the actual rotary kiln as shown in Fig. 2(a). This mayseem to be explained by the fact that fine-Ni and Fe particlesconfined in the silicate are released outside by the fractionalcrystallization occurring with the crystallization of silicateduring the cooling of the rotary kiln, Ni and Fe being ableto be precisely analyzed.
4.3. Observation of the Occurrence Process of ReducedMetal by SEM
4.3.1. Occurrence of Low-MgO and High-SiO2·NiO·FeOSilicate by the Fractional Crystallization
Then, the observation of the occurrence process ofreduced metal is performed by SEM. Fig. 7(a) shows thatlow-MgO and high-SiO2·NiO·FeO silicate is released fromthe right side high-MgO silicate with the recrystallization ofserpentine. Then, the fine metal particles occur at the point1, and the sharp peak of FeO/SiO2 appears in the map lineanalysis as shown in Fig. 7(b). Here, the metal (Ni, Fe) anddivalent ions (Ni2+, Fe2+) cannot be distinguished by theSEM quantitative analysis, which permits total of both to berepresented by NiO and FeO in this paper.
4.3.2. Relation Between Occurrence of Fine Metal andChemical Composition of Ni-ore
Figures 8(a) and 8(b) show the conditions of occurrenceof fine metal in the low-MgO and high-FeO Ore-A at1 100°C. The high-NiO·FeO·SiO2 silicate is released from
(a) SEM-image
(b) Distribution of elements
Fig. 7. Inter-diffusion of FeO, SiO2 and MgO in low-MgO andhigh-FeO silicate in Ore-A at 1 000°C (1.2A0).
(a) SEM image; Occurrence of fine metals at 1 100°C
(b) Values of SEM quantitative analysis
Fig. 8. Occurrence of the fine metals at 1 100°C in Ore-A (1.2A0).
points 5, 6, 7, 8 to points 1, 2, 3, 4, then NiO and FeO beingreduced to the fine metal particles containing the low con-tent of Ni (around 7 mass%). On the other hand, in the high-MgO and low-FeO Ore-B, Figs. 9(a) and 9(b) shows thathigh-NiO·FeO·SiO2 silicate is released from points 3 topoints 1, 2, the fine metal particles containing the low con-tent of Ni (around 11 mass%) being unable to occur until1 250°C. However, reducibility is very low, and fine metalparticles occur over the bulk, large metal occurring only atthe surface of crystal and surrounding pores.
4.3.3. Procedure of the Occurrence of the Fine MetalThereby, it could be inferred from the above mentioned
fact that fine metal particles are not formed by the agglom-eration of reduced metal, but occur in this order as shownin Fig. 10: 1) the separation of low-MgO and high-NiO·FeOsilicate released from serpentine with the recrystallization tothe quasi-(Mg–Fe) olivine or pyroxene and high-NiO·FeOsilicate (impurity) by fractional crystallization; 2) theagglomeration of high-NiO·FeO silicate; 3) the reduction ofhigh-NiO·FeO silicate to the fine metal particles containingthe low content of Ni, and 4) the agglomeration of the finemetal particles via melt occurred by the migration of high-SiO2 silicate enveloping fine metal particles to the surround-ing mineral. Here, a small amount of goethite contained inNi-ore transforms to hematite by dehydration, being easilyreduced to metal, which allows the behavior of reduction notto be discussed.
On the other hand, in the reduction of iron ores,14) CO gascomes into direct contact with oxygen on the surface ofhematite. Subsequently, oxygen is released by the formation
of CO2 to the gas phase, allowing the ratio of Fe/O at thesurface of ore to increase, which enables Fe to move towardthe bulk by the gradient of Fe/O via vacancies to cause met-al to grow largely. Thus, the difference in the mechanism ofmetal formation between iron ores and Ni-ores is attributedto the content of Fe and bonding state with oxygen.
4.4. Mechanism of the Reduction of (Mg–Ni–Fe) Sili-cate in the Ni-ores
4.4.1. Previous Works Concerning Reduction ReactionProcess
Shirane2) has performed a reduction experiment of a partof lump of Ni-ore (i.e., green garnierite; Ni=11.0, Fe=1.3,MgO=12.3, SiO2=61.0 mass%) using hydrogen. Olivineaccompanied by the recrystallization of amorphous serpen-tine at 800°C following dehydration from 600°C is reduced.In this case, it is reported that the increase in the mole frac-tion of Mg2SiO4 leads to the decrease in activity of Ni2SiO4,which causes the reducibility of Ni to lower.
Further, Matsumori15–17) has proposed the reductionmechanism of Ni-ore as follows: 1) hematite formed by thedehydration of goethite is reduced to FeO; 2) (Mg–Fe–Ni)olivine is formed by the reaction between FeO and amor-phous serpentine, being directly reduced to metal by C aftersoftening and melting. However, such behavior cannot beentirely found in this SEM observation.
4.4.2. Relation Between Point Defects and DiffusionThen, in order to discuss the process of the occurrence of
fine metal particles, at first, relation between diffusion andpoint defects giving a substantial effect on reduction hasbeen checked in the literature. Many experiments on therelation between point defects and diffusion have been per-formed. For example, Dimanov18) has carried out Fe–Mgand Mn–Mg inter-diffusion experiments between ferro-
(a) SEM image; Occurrence of fine metals at 1 250°C
(b) Values of SEM quantitative analysis
Fig. 9. Occurrence of the fine metals at 1 250°C in Ore-B (1.2A0).Fig. 10. Schematic representation of the mechanism of the forma-
johannsenite (Ca, Fe, Mn)O·SiO2 and diopside (Ca,Mg)O·SiO2. In that experiment, it is found out that the acti-vation energy for diffusion is largely reduced with theincrease in point defects. Furthermore, Azough19) have per-formed the iron diffusion experiments using natural and syn-thetic diopside, thus having revealed that diffusion takesplace more rapidly in natural diopside with much pointdefect than synthetic diopside. From the above mentioned,it is confirmed that point defects enhance the diffusion.
4.4.3. Concentration of Point Defect in the Heated Ni-oreThen, what extent point defects occurred during the
weathering remains in the silicate during the reduction reac-tion of NiO and FeO is calculated by Eq. (7)10) from the val-ues of SEM-EDS quantitative analysis measured at 900–1 250°C. In this case, silicates seem to exist in the form ofas follows: 1) olivine; 2) serpentine-anhydride; 3) pyroxene;and 4) excessive SiO2 silicate, but those cannot be clearlydistinguished.
Then, the ratio of the lack of the mole number in the cat-ion component is denoted as the concentration of the pointdefects by the using olivine as an indicator for the degree ofpoint defects.
Concentration of point defects (atom%)=((MgO+CaO+MnO+FeO+NiO)/(SiO2+2×Al2O3)–2)/2×100............... (7)
Where, elements refer to the mole number of those. The dis-tribution of point defects at 1 000°C shown in Fig. 11(a)indicates that point defects still remain after the recrystalli-zation, Ore-A having some higher content of point defectsthan Ore-B.
Then, the average of concentration of point defects at var-ious temperatures in Fig. 11(b) shows that increase in tem-perature provides the increase in the concentration of pointdefects by the fractional crystallization in Ore-A, silicatetransforming to the quasi-pyroxene or excessive SiO2 sili-cate. On the other hand, high content of MgO in Ore-Bcauses SiO2, and so on (i.e., extra composition for recrystal-lization) to be released, allowing the concentration of pointdefects to decrease, which causes silicate to approach quasi-olivine. The facts above-mentioned indicate that some highreducibility develops in Ore-A, conversely, some low reduc-ibility develops in Ore-B. This can be deduced from theresults of SEM observation in section 4.3.2.
4.4.4. Removal of the Oxygen by the Diffusion Path viaPoint Defects and Oxygen Vacancies
Here, even in the reduction of the rock-salt type oxide(Mg, Ni, Fe)O, which have a few point defects and strongerbonding with oxygen than Ni-ore sharing oxygen in theSiO4 tetrahedron, metals can be occurred in the bulk. Thisis interpreted by Schmalzried20,21) as follows: reductionreaction occurs by the cationic and electronic defect species(i.e., pairs of negatively charged cation vacancy (V//) andholes (h·)) corresponding to the oxygen partial pressure inthe gas phase. NiO is reduced by Eqs. (8) and (9).
Here, (V//+2h·) must be eliminated to the outer surface
according to Schmalzried, however, in this case assumed tobe regenerated in the bulk. Ni-ore have much point defectoccurred during weathering. Furthermore, the reduction ofFe3+ in serpentine to Fe2+ by Eq. (10) enables 1 mole of oxy-gen per 2 moles of Fe to be removed from serpentine, thusoxygen vacancies probably occurring. Then, it is deducedthat those defects have a great effect on the reduction reac-tion in the silicate.
Fe2O3+CO=2FeO+CO2 ..................... (10)
For example, it seems that atomic oxygen accompanied bythe reduction in the silicate is surrounded by SiO2, MgO,and so on, which allows the removal of oxygen to be verydifficult. Thus, the diffusion path in which point defects andoxygen vacancies probably play an important role is essen-tial to the oxygen removal. Therefore, reduction model tak-ing that into the consideration can be inferred. However, thelack of experimental data concerning defect chemistry inNi-ore cause model to be still only a hypothesis.
4.5. Mechanism of the Growth of Metal in the Ni-ores4.5.1. Growth of Fine Metal via Diffusion and Melt
Figure 8(a) shows that the fine metal particles of around2.0 μm in diameter occur at 1 100°C in the point 4 in theOre-A, and that metal occurs from the agglomerated low-MgO and high-NiO·FeO silicate due to the fractional crys-tallization. It seems to considered that the content of Ni2+
Fig. 11. Concentration of the point defects calculated from the val-ues of SEM-EDS in case of regarding all minerals as oliv-ine. (a) at 1 000°C (b) average.
and Fe2+ ion at the around point 4 lower with reduction,allowing the concentration gradient between point 4 and 8in the surrounding mineral phase to be formed, which causethe ion to be provided by the diffusion due to via pointdefects. This seems to enhance the growth of fine metal par-ticles.
Furthermore, the high-SiO2 silicate enveloping fine metalparticles migrates to the surrounding mineral phase with areduction, allowing the occurrence of melt surrounding met-al, which metals to coalesce with each other to grow morelargely.
On the other hand, Fig. 9(a) shows the fine metal particlesoccurred in Ore-B. It seems that the mechanism of growthof fine metal is almost similar to Ore-A. However, the con-tent of high MgO and low FeO in ore allows Ore-B to havea small amount of point defect and the occurring tempera-ture of melt to shift to higher temperatures. This leads to nooccurrence of fine metal of max around 1.6 μm at the sur-face of silicate and surrounding pores until 1 250°C as men-tioned at section 4.3.2.
4.5.2. Relation Between Heating Temperature, Grain Sizeand Ni Content of Metal
Figure 12(a) shows relation between heating tempera-ture, the grain size and the average grain size of metal in
Ore-A: where average values are calculated by excludingthe maximum and minimum values. The grain size increasesfrom around 1 250°C. This indicates that the fine metalgrows largely via melt because of melt occurring fromaround 1 250°C.10) Further, Fig. 12(b) shows relationbetween heating temperature, the Ni content, the averagevalues of that and the standard deviation values of that inOre-A. The average values almost reach the saturation ataround 1 250°C. On the other hand, the standard deviationvalues are low in the range 1 050–1 150°C. This seems to beattributed to the difficulty in the detecting the metal of highNi content due to the low Ni content in Ni-ore. However, theincrease in heating temperature to around 1 200–1 250°Cprovides the larger size particles of both the high Ni andhigh Fe metal due to the diffusion, allowing both metals toco-exist, which causes the standard deviation values to varylargely. Whereas, it seems that the increase in heating tem-perature above 1 250°C leads to the occurrence of melt,allowing the high Ni and Fe metal to coalesce with each oth-er, which causes the standard deviation values to be greatlyreduced.
5. Conclusions
The sampling of raw materials in the rotary kiln out ofoperation, firing experiment by the experimental kiln, andwater quenching experiment have been performed, and theclarification of behavior of the reduction and growth of met-al have been attempted by SEM-EDS analysis. The result isobtained as follows.
(1) In the amorphous serpentine region which existsfrom the beginning of dehydration at around 600°C to therecrystallization to the olivine, Ni has very high reducibilityto be reduced as much as around 70 mass% at 1 000°C.However, Fe has some low reducibility to be reduced onlyup to around 35 mass%. On the other hand, hematite fromdehydrated goethite has very high reducibility.
(2) The increase in heating temperature above 1 000°Callows the fine reduced metallic Ni and Fe to be confined inthe silicate accompanied by recrystallization, which causesthe extraction by the bromine methyl alcoholic solution tobe difficult. This allows the degree of reduction to enter thedepression region and decrease apparently. However, theincrease in heating temperature above 1 250°C allows meltto occur, which causes the fine metal to coalesce with eachother. Therefore, the metallic Ni and Fe can be preciselyanalyzed by the bromine methyl alcoholic solution, allowingthe apparent degree of reduction to approach the equilibriumvalues (i.e., Ni, Fe around 96, 60 mass%, respectively at1 300°C).
(3) The recrystallization of serpentine causes low-MgOand high-FeO·NiO silicate to be released outside, which isenriched by the fractional crystallization. Thereafter, thecontent of point defects is increased, and reducibilitybecomes high, which allows reduction reaction to beenhanced.
(4) From the micro standpoint of the reduction mecha-nism of Ni-ore, it seems that Ni2+, Fe2+ in the crystal latticeare reduced by the cationic and electronic defect speciesproposed by Schmalzried, the occurred oxygen diffusestoward the crystal surface with the low concentration of
Fig. 12. Relation obtained by SEM observation between the diam-eter of metal, the content of Ni in the metal and tempera-ture.
oxygen to be released outside. For this diffusion process, itis deduced that the point defects formed during the weath-ering play an important role from the result of SEM obser-vation that metals occur from the lower temperature in thelow-MgO and high-FeO ore than the high-MgO and low-FeO ore. However, the lack of experimental data concerningdefect chemistry in Ni-ore cause model to be still only ahypothesis.
(5) The fine metals agglomerated by the diffusioncoalesce with each other via melt to grow largely. This isconfirmed by the fact that the temperature of the melt occur-rence coincides with the temperature of metal beginning togrow largely in SEM observation.
REFERENCES
1) Y. Shirane: J. MMIJ, 85 (1969), 1001.2) Y. Shirane: J. MMIJ, 87 (1971), 419.
3) Y. Shirane: J. MMIJ, 89 (1973), 811.4) F. Johannsen: Tetsu-to-Hagané, 25 (1939), 976.5) T. Watanabe, S. Ono, H. Arai and T. Matsumori: Int. J. Miner. Process.,
19 (1987), 173.6) Y. Kobayashi, H. Todoroki and H. Tsuji: Proc. MMIJ Fall Meeting,
MMIJ, Tokyo, (2007), 159, CD.7) Y. Kobayashi, H. Todoroki and H. Tsuji: The 4th Int. Cong. on the
Science and Technology of Steelmaking, ISIJ, Tokyo, (2008), 518.8) Y. Kobayashi, H. Todoroki and H. Tsuji: ISIJ Int., 51 (2011), 35.9) H. Tsuji, H. Todoroki and Y. Kobayashi: Proc. MMIJ Fall Meeting,
MMIJ, Tokyo, (2007), 157, CD.10) H. Tsuji: ISIJ Int., 52 (2012), 333.11) M. S. Ghiorso: Geochimica et Cosmochimica Acta, 58 (1994), 5489.12) Y. Okajima: J. MMIJ, 103 (1987), 41.13) Y. Waseda: Thermodynamics, AGNE Gijutsu Center, Tokyo, (1998).14) R. Nicolle and A. Rist: Metall. Mater. Trans. B, 10 (1979), 429.15) T. Matsumori: J. MMIJ, 115 (1999), 603.16) T. Matsumori: J. MMIJ, 116 (2000), 141.17) T. Matsumori: J. MMIJ, 116 (2000), 597.18) A. Dimanov and M. Wiedenbeck: Eur. J. Mineral., 18 (2006), 705.19) F. Azough and R. Freer: Phys. Chem. Miner., 27 (2000), 732.20) H. Schmalzried and D. L. Ricoult: J. Mater. Sci., 22 (1987), 2257.21) H. Schmalzried: Ber. Bunsenges. Phys. Chem., 88 (1984), 1186.
