Sticking-Free Reduction of Titanomagnetite Ironsand in a Fluidized Bed Reactor

  • Sigit W. Prabowo
  • Raymond J. Longbottom
  • Brian J. Monaghan
  • Diego del Puerto
  • Martin J. Ryan
  • Chris W. BumbyEmail author


Fluidized bed reduction of iron ore fines is typically inhibited by the onset of “sticking” at temperatures above 973 K, which leads to particle agglomeration and defluidization of the bed. Here, we report the sticking-free fluidized bed reduction of titanomagnetite (TTM) ironsand at 1223 K in Ar-H2 gas mixtures. We show that sticking is prevented by the formation of a protective titanium-rich oxide shell around each particle during the initial reduction stage. This protective shell prevents iron-iron contact between particles throughout the reduction process, enabling metallization degrees of 93 pct to be attained without sticking occurring. Phase evolution during the reaction has also been analyzed using q-X-ray diffraction and scanning electron microscope/energy dispersed spectroscopy. We find that the reduction proceeds through four separate stages. During the initial stage, approximately half of the initial TTM phase is converted to wüstite, forming a network of sub-micron wüstite channels which interlace the TTM matrix. During this stage, Ti and Al are enriched within the TTM matrix, due to the low solubility of both species in wüstite. This enrichment stabilizes the remaining TTM, meaning that wüstite is then preferentially reduced to metallic iron in stage 2 of the reduction. In stage 3, the remaining Ti-enriched TTM is reduced directly to metallic iron and ilmenite. The final stage of reduction involves the conversion of ilmenite into rutile and pseudobrookite. Our findings clarify the important role played by titanium species during the reduction of TTM and suggest that New Zealand ironsand can offer significant advantages over conventional hematite ores when used as a feedstock for fluidized bed direct-reduced iron processes.


Titanomagnetite (TTM) ironsand has been studied as a potential alternate low-cost iron ore due to the ease with which it can be extracted and magnetically concentrated.[1, 2, 3, 4, 5] Substantial deposits of TTM ironsand are found throughout the west coast of the North Island of New Zealand (NZ), which are utilized as the primary source of iron for domestic steel production.[6,7] At present, this TTM ironsand ore is commercially processed in a coal-fired rotary kiln process.[8] However, this approach, similar to other carbothermic process, exhibits low energy efficiency and produces high emissions of gases such as CO2 and SO2.

The increased focus for reducing CO2 emission in global steel production has attracted renewed interests in the gas-based reduction utilizing hydrogen gas to treat TTM ironsand.[5,9, 10, 11] Among the process technologies used for gas-based reduction, fluidized bed processing offers advantages as it enables fast processing times and direct use of the raw ironsand fines. This eliminates the need for additional pelletizing and sintering plant required when processing fines using other traditional ironmaking routes.[12, 13, 14] Moreover, NZ TTM ironsand ore exhibits a naturally occurring particle size distribution in the range 90 to 250 µm which are approximately spherical in shape, making it well suited to fluidized bed processing.

A significant technical obstacle to the high throughput fluidized bed reduction of iron ores is the sticking of particles at high temperatures (≳ 973 K).[12,15,16] “Sticking” refers to the agglomeration of particles in the whole bed during the reduction process.[15,17,18] Once agglomerated, fluidization is lost and the process is arrested, limiting further reduction of the iron ore fines.[19, 20, 21] This phenomenon is often associated with the growth of fibrous whiskers on the surface of freshly reduced iron particles which can lead to particles becoming entangled together upon contact.[15,21,22] Direct iron-iron contact at particle surfaces can also lead to the sticking of iron particles even in the absence of fibrous structures.[16,23, 24, 25]

Several methods have been proposed to prevent the sticking of iron fines during fluidized bed reduction. In particular, the coating of ore particles with either inert oxides or carbon has been investigated, due to the simple practical implementation of this method.[14,26] Coating hematite fines with inert oxides such as MgO and CaO has been reported to effectively prolong the fluidization at temperatures up to 1073 K, but also leads to a substantially reduced reaction rate.[18,27,28] However, at temperatures above 1073 K, these methods were found to be ineffective, with iron-iron contact again becoming the predominant cause for sticking.[18,28] The coating of hematite fines with carbon was reported to effectively resist sticking at temperatures up to 1173 K.[14,29,30] However, this method requires a long multi-step process to pre-deposit carbon on the hematite particles.[12,14] Doping metal oxides (including Al and Mg) into the crystal lattice of synthetic hematite has also been reported to suppress the growth of iron whiskers during reduction in CO at 1073 K, although this also decreased the reduction rate.[31] While these methods have been experimentally demonstrated for hematite ores, none have been investigated specifically with TTM ironsand.

The literature dealing with sticking phenomena during fluidized bed reduction of TTM ironsand is limited. The few studies available indicate that there are important differences in sticking tendency for different sources of ironsand,[23,32, 33, 34] and different reducing gases. In a CO gas reduction study of NZ TTM ironsand,[32] it was reported that sticking started to occur at temperatures ~ 1073 K. In contrast, no sticking was reported in a CO gas fluidized bed reduction investigation of South African ironsand at temperatures up to 1223 K.[33,34] In H2 gas reduction studies of NZ TTM ironsand,[22,32] no sticking was reported at temperatures up to 1173 K, although some sticking did occur at 1273 K.[32] However, the reason why the TTM particles did not stick in these cases was not identified and hence requires further investigation.

It is generally accepted that the reduction of TTM is slower than hematite ores because the titanium in TTM stabilizes the spinel structure.[5,10,11,35] Several authors have reported that H2 gas reduction of the TTM occurs in a sequence of several steps.[5,10,36] The initial step involves the reduction of TTM to wüstite, which is then followed by the reduction of wüstite to metallic iron.[5,36] However, Park and Ostrovski[10] reported only a very small wüstite X-ray diffraction (XRD) peak during the initial reduction stage. The final step of reduction involves the formation of Ti-bearing oxides, which are ilmenite (FeTiO3), rutile (TiO2), and pseudobrookite (Fe2TiO5). It should be noted that in all the published literature, it was pellet or packed bed reduction that was studied and not reduction of ironsand in a fluidized bed.

In this study, the fluidized bed reduction of TTM ironsand in hydrogen has been investigated at 1223 K, a temperature well above the onset of sticking for conventional hematite ores. Reduction experiments were performed using a range of different hydrogen-argon gas mixtures, with the primary purpose of investigating the sticking tendency and phase evolution of the TTM ironsand under these conditions. In situ sampling enabled XRD and scanning electron microscope (SEM) analysis of samples obtained at different reaction times in each experimental run.


Characterization of Raw Ironsand Concentrate

The ironsands sample studied in this work was TTM ironsand concentrate from Waikato North Head (WNH) Mine. The particle size of the samples used in this study was screened to be in the range of 106 to 125 µm, which corresponds to the average particle size of the as-received raw concentrate (see also data provided in Electronic Supplementary Material). X-ray fluorescence (XRF) samples of this size fraction were prepared following a standard procedure, in which the sample was oxidized in air at 1273 K for 1 hour prior to analysis. The sample was then mixed with a lithium tetraborate melt as a standard calibration matrix. The XRF result for the composition of the 106 to 125 µm size fraction is presented in Table I.
Table I

Chemical Composition of the WNH Ironsands Concentrate, 106 to 125 μm Size Fraction, as Measure by XRF

Compositions (Equivalent Mass Pct)



















TTM is a solid solution of magnetite (Fe3O4) and ulvöspinel (Fe2TiO4). Magnetite and ulvöspinel exhibit identical crystal structure, with only a small change in lattice constant (1.6 pct).[37,38] As such, it is not possible to resolve the peak-shift between XRD patterns from TTM containing small differences in Ti composition, using a standard laboratory XRD instrument. Figure 1 shows the XRD patterns obtained from the raw ironsand concentrate used in this work. Previous reports have found that NZ TTM ironsand has the stoichiometric composition, Fe3−xTixO4, where x = 0.27 ± 0.02.[1,5,10,11] This is consistent with the ratio calculated from the XRF data in Table I. XRF data also indicate that there are significant levels of Al and Mg present within ironsand. This is confirmed by spot energy dispersed spectroscopy (EDS) measurements of TTM grains (see Section III–D).
Fig. 1

XRD patterns of the raw ironsand concentrate, 106 to 125 μm size fraction

The particles of NZ TTM ironsand fall into two broad classes of naturally occurring morphology. The great majority of particles are uniform homogeneous grains of TTM, but a small minority (< 10 pct of particles) are non-uniform particles containing lamellar structures of titanohematite (TTH), (Fe2−yTiyO3) exsolved in a TTM matrix.[39, 40, 41] This is discussed further in detail later.

Experimental Set-up

A schematic design of the fluidized bed reactor and sampling system used in this work is presented in Figure 2. A quartz tube with an inner diameter of 30 mm is used as the reactor tube. A quartz frit with a porosity of 40 to 90 µm is used as the gas distributor. The reactor tube is located in a vertical tube furnace with a distance of 25 cm between the distributor and the top of the hot zone. The top part of the reactor is wrapped with an alumina fiber, and the top flange is wrapped with a band heater to prevent water condensation at the reactor outlet. Within the reactor tube, one end of the sampling tube is located 20 mm above the quartz frit. The other end is connected to the cyclone-type sampler. Thermocouples are located at the top and bottom of the reactor with the tip of each thermocouple close to the quartz distributor. Pressure transducers are connected at the inlet and outlet of the gas line to record the pressure drop across the bed during fluidization.
Fig. 2

Schematic diagram of (a) the fluidized bed reactor, and (b) its sampling system. Fluidized bed (FB) comprises 100 g of ironsand sample

The sampling system consists of a Venturi nozzle, a cyclone sampler, and a sample container as presented in Figure 2(b). This system utilizes the Venturi effect to sample material from the reactor bed during an experimental run. Solenoid valves are installed on the sampling system line and automatically controlled for rapid actuation of the open–close cycle. During sample collection, a high flow rate of nitrogen gas (≥ 75 L/min) is introduced to the inlet of the Venturi nozzle, and then the valves are activated by the software to an “open” state. A small mass (≲ 5 g) can extracted directly from the bed by activating the “open” state of the valves for one second.

A 100 g sample of ironsand was charged into the reactor for each reduction experiment. The sample was purged by Ar gas with a flow rate of 3 L/min while heating up the bed to the experimental temperature. After the bed temperature reached 1223 K, the fluidizing gas was then switched to the reducing gas of varying compositions. The total gas flow rate (H2-Ar) for the reduction was set to 5 L/min which corresponded to the bubbling fluidization regime[42] for the ironsand in the reactor. At predetermined sampling times, samples were collected by activating the pre-set open–close cycle. After the sample was extracted into the sample container, the sample container was then quench cooled in a water pot. The sample container is independent and isolated from the reactor, so multiple samples can be exchanged without stopping reactor operation. When exchanging the sample container, a low flow rate of Ar was introduced to purge the container opening and prevent re-oxidation of the sample.

Sample Characterization

The crystal phases within the reduced ironsand were characterized using XRD (Bruker D8 Advance) with Co Kα radiation, at 0.05 deg step size and 2 s/step collection time. Quantitative XRD analysis (q-XRD) was used to determine the crystal phase composition of each sample, and this was performed using Topas 4.2 (Bruker) software. The metallization degree (pct met) was calculated from the q-XRD data using Eq. [1].
$$ {\text{pct}}\;{\text{met}} = {\text{Fe}}^{\text{o}} /{\text{Fe}}_{\text{Tot}} \times 100\,{\text{pct}}. $$

The total iron (FeTot) is determined as the total iron within all iron-containing phases, while the metallic iron (Feo) is determined as the independent metallic iron phase detected by XRD. The FeTot and Feo contents were calculated by the q-XRD analysis. The accuracy of the q-XRD technique was confirmed by comparison with standard titration tests (ISO 16878:2016)[43] on a series of partially reduced magnetite and ironsand samples (see data provided in Electronic Supplementary Material). This comparison showed close agreement between the q-XRD and titration methods (to within 3 pct), across the full measurement range considered in this work.

Microstructure characterization of the samples was performed using a field-emission SEM (FEG-SEM, FEI, Nova 450) at 20 kV. Elemental spot analysis and mapping were analyzed using an integrated EDS detector at 15 kV.

Results and Discussion

Reproducibility of the Fluidized Bed Reactor System

Initial tests were performed to confirm the reproducibility of the experimental fluidized bed reactor and sampling system. Five experimental runs were performed under identical conditions. In each case, a fresh 100 g sample of ironsand powder was reduced in a gas composition of 80 vol pct H2 and 20 vol pct Ar at 1223 K. Figure 3 shows plots of pct met as a function of time from each of the five reduction runs. It is clear that these experiments show excellent reproducibility, and in each case the reaction reached a pct met of ~ 93 pct after 40 minutes.
Fig. 3

Pct met curves for the five reduction runs of TTM ironsand in 80 vol pct H2-20 vol pct Ar gas mixture at 1223 K

Importantly, despite the relatively high temperature employed (1223 K), no sticking was observed throughout these reduction runs. This is in direct contrast to previous reports on the fluidized bed hydrogen reduction of hematite ores, where sticking starts to occur at ∼ 923 K.[17,29] In addition, XRD measurements of post experimental reduced ironsand samples indicated that the samples appeared to be stable after weeks of exposure to air. In Section III–D, both these properties are discussed and are likely a result of the formation of a protective Ti-enriched oxide shell around the particles.

Effect of Reducing Gas Composition on Metallization Rate

The effect of hydrogen gas concentration on the fluidized bed reduction of TTM ironsand was studied using a range of H2-Ar gas mixtures at 1223 K. The hydrogen gas composition was varied between 50 and 100 vol pct. Figure 4(a) shows the resulting metallization degree as a function of time for each gas composition measured, indicating that the reduction rate increased with increasing hydrogen gas content. A maximum pct met ~ 93 pct was achieved after 30 minutes in 100 pct H2 gas. At lower H2 concentrations, the maximum pct met was achieved after 60 minutes of reduction (the end-point of each reduction run).
Fig. 4

Metallization degree of the TTM ironsand reduced by H2-Ar gas mixture at 1223 K using different hydrogen gas concentrations. (a) Pct met as a function of time. (b) Pct met as a function of total delivered hydrogen volume

What can be clearly seen from Figure 4(a) is that after 60 minutes of reduction, the metallization curves meet at the same point for all reducing gas compositions. Figure 4(b) reveals that in each case the reduction degree follows a similar relationship with the total volume of hydrogen delivered to the reactor. This is as would be expected for a simple first-order reaction.

Phase Evolution During Reduction by H2-Ar Gas Mixture

Figure 5 shows a series of XRD patterns obtained from samples extracted at various times during reduction in 80 vol pct H2-20 vol pct Ar at 1223 K. After 5 minutes of reduction (corresponding to a pct met of ~ 13 pct), the initial TTH in the ironsand has been completely reduced to TTM, while the metallic iron and wüstite peaks are already prominent. After 30 minutes, the wüstite peak has completely disappeared, and minority Ti-containing phases such as ilmenite and rutile start to appear. The formation of ilmenite and rutile in the final stage of the reaction is consistent with the previous reports.[10] After 40 minutes, the reduction has reached its maximum pct met, and pseudobrookite peaks start to appear. Further reduction to 60 minutes does not completely reduce TTM to iron as indicated by the remaining small TTM peaks in the final product. However, rutile is no longer observed after 60 minutes, indicating that reactions involving minority Ti-containing phases continue beyond the apparent end-point of the metallization process.
Fig. 5

XRD patterns of reduced ironsands showing the progression of reduction reaction by 80:20 vol pct H2:Ar gas mixture at 1223 K. (a) Full XRD pattern. (b) Magnified pattern with the y-axis limited to the maximum intensity of 1500 cps for each pattern

q-XRD analysis has been used to obtain the wt pct of each crystalline phase from each of the XRD patterns shown in Figure 5. Figure 6 shows the time evolution of each crystalline phase for each of the H2 gas concentrations studied. It is clear that the hydrogen gas concentration significantly affects the reduction rate, particularly in the later stages of the reaction. However, for all of the gas compositions studied, the phase evolution during reduction follows the same sequence of stages. The duration of each reduction stage for each hydrogen gas composition is indicated by the numbered spans at the top of each plot in Figure 6. In the first stage, the small amount of TTH present in the raw ironsand (~ 8 wt pct) is fully reduced to TTM. This occurs rapidly, within 5 minutes of starting the reaction. During this same stage, approximately 50 pct of the TTM phase is also reduced to wüstite.
Fig. 6

The evolution of the amount of crystalline phases as a function of time during the reduction of ironsand at 1223 K in H2-Ar gas mixtures. Volume ratios (H2:Ar) of the gas mixtures used are (a) 50:50, (b) 60:40, (c), 70:30, (d) 80:20, and (e), 100 pct H2. Numbered spans at the top of each plot indicate the duration of each reduction stage, as described in the text

In the second reduction stage, wüstite is reduced to metallic iron, such that the wüstite composition decreases to zero from a maximum composition of ~ 40 wt pct in each case. During this stage, the reduction rate of the TTM phase slows significantly, indicating that hydrogen is being consumed primarily by the wüstite reduction. The observation of wüstite during this stage is consistent with previous work,[5,36] which also reported that wüstite occurred as a transitional phase during early reduction of TTM. However, a different result was reported in a previous study of packed beds of NZ ironsand reduced by hydrogen gas at 1173 K.[10] In that study, the authors did not observe significant levels of wüstite and concluded that the reduction of TTM to wüstite is much slower than the subsequent reduction of wüstite to metallic iron. By contrast, the results in Figure 6 show that the initial reduction of TTM to wüstite (in stage 1) occurred rapidly and faster than the subsequent reduction of wüstite to metallic iron in stage 2.

The third stage occurs once all the wüstite has been consumed. At this point, the reduction rate of the remaining TTM accelerates, and metallic iron is formed directly i.e., not through wüstite as an intermediate. At the same time, ilmenite is also observed within the sample for the first time. The direct conversion of TTM to iron observed in stage 3 represents a significant change in the reaction pathway from the stepwise reduction via wüstite that was observed in stages 1 and 2. This behavior differs significantly from that reported for conventional magnetite ore,[44,45] which is always reduced via wüstite at these temperatures. This change in favored pathway implies that the thermodynamic potential of the TTM phase has itself changed over the course of the reaction. This is thought to be due to gradual Ti-enrichment of the TTM phase during the first two reaction stages, as described by reactions [2] and [3] below.
$$ {\text{Stage}}\;1 :\;(x + \delta ) \cdot {\text{Fe}}_{(3 - x)} {\text{Ti}}_{x} {\text{O}}_{4} + \delta {\text{H}}_{2} \to 3\delta \cdot {\text{FeO}} + x \cdot {\text{Fe}}_{3 - (x + \delta )} {\text{Ti}}_{(x + \delta )} {\text{O}}_{4} + \delta \cdot {\text{H}}_{2} {\text{O}}, $$
$$ {\text{Stage}}\;2 :\;{\text{FeO}} + {\text{H}}_{2} \to {\text{Fe}} + {\text{H}}_{2} {\text{O}} . $$

In stage 1, the TTM phase becomes increasingly enriched with Ti, which appears to stabilize it limiting further reduction. Ultimately, the Ti-enriched TTM reaches a point at which it is less reducible than the co-existing wüstite. This represents the transition to stage 2 and the reaction pathway switches to prefer reduction of the wüstite phase, while the reduction rate of the remaining TTM slows significantly. From Figure 6 it can be seen that this transition occurs at the point when the original TTM has dropped to ~ 50 pct of its initial value. If it is assumed that Eq. [2] fully describes the fate of Ti during stage 1, then this implies that the transition to stage 2 occurs once the TTM composition reaches approximately Fe(3−Δ)TiΔO4, with Δ ≈ 0.5.

Once the wüstite has been fully consumed, the reduction enters the third stage of the reaction, whereby further reduction can proceed according to reaction [4].
$$ {\text{Stage}}\;3 :\;{\text{Fe}}_{(3 - \Delta )} {\text{Ti}}_{\Delta } {\text{O}}_{4} + (4 - 3\Delta ){\text{H}}_{2} \to \Delta \cdot {\text{FeTiO}}_{3} + (3 - 2\Delta ) \cdot {\text{Fe}} + (4 - 3\Delta ) \cdot {\text{H}}_{2} {\text{O}}. $$
A final fourth stage in the reduction sequence is observed in Figure 6 once the conversion to metallic iron nears completion. During this fourth stage, reactions between the residual Ti-bearing minority phases continue, with ilmenite being converted first into rutile and then pseudobrookite. Figure 7 shows magnified plots of the phase evolution of these three phases throughout the reduction process. Ilmenite is the first Ti-containing phase observed, as it is formed via reaction [4] in stage 3. The ilmenite is then itself reduced to metallic iron and rutile. Prolonged exposure to high temperature reducing conditions, then results in the rutile combining with residual ilmenite to form pseudobrookite (which reaches a maximum 6 wt pct of the final product after 60 minutes reduction with 100 pct H2).
Fig. 7

The proportions of the minority Ti-containing phases [ilmenite (a), rutile (b), and pseudobrookite (c)] as a function of time during the reduction of ironsand by various H2-Ar gas mixtures at 1223 K

Microstructural Evolution of Ironsand Particles During Fluidized Bed Reduction

The fact that the ironsand was not observed to stick during the fluidized bed reduction is a surprising and positive result and differs from previous reports of fluidized bed reduction of iron ore fines.[17,23,29] In order to investigate the underlying cause of this behavior, SEM microstructural analysis was performed on the samples reduced by 100 vol pct hydrogen. Similar micrographs and microstructural features have also been obtained for samples reduced at lower H2 concentrations but are not reproduced here due to space constraints.

Figure 8 shows back-scatter electron (BSE) images of both the unreduced ironsand and the final product after reduction for 60 minutes in 100 pct H2. It is clear that the reduced particles retain a similar size distribution to the unreduced material and are individually separated, showing no evidence of agglomeration (see Figures 8(a) and (b)). There are also no fibrous or whiskers of iron visible on the surface of the reduced particles as shown in Figures 8(b) and (d) (these are common features of reduced iron ore particles that do exhibit sticking).
Fig. 8

SEM micrographs for unreduced ironsand particles (a, c) and ironsand particles reduced in 100 vol pct H2 at 1223 K for 60 min (b, d). Red markers indicate the location of magnified image (Color figure online)

There are significant differences between the microstructural evolution of the two different types of particles found in the original unreduced ironsand, namely the uniform homogeneous TTM particles and the non-uniform particles containing TTH lamellae. It is useful to discuss each type of particle separately. Starting with the uniform homogenous TTM particles, as these form by far the majority of the sample (≳ 90 pct) and hence determine to a large degree the bulk reduction behavior measured in the XRD.

Microstructural evolution of uniform TTM particles

Figure 9 shows a typical cross-section of the BSE images formed by reduction of homogenous TTM particles. Images are shown for particles extracted at each sampling time during reduction in 100 pct H2 gas.
Fig. 9

The microstructural evolution of the uniform grain during the reduction by 100 vol pct hydrogen at 1223 K: (a) raw ironsand, (b) 5 min, (c) 10 min, (d) 20 min, (e) 40 min, and (f) 60 min

Several features are immediately apparent. The fully reduced samples (Figures 9(e) and (f)) show a filigree-like structure of metallic iron (bright white) throughout the inner region of the particle, which is surrounded by a gray shell. This protective outer oxide shell is also apparent in all other images (except for the original unreduced ironsand) and is the key reason that these particles do not exhibit sticking in the fluidized bed. Importantly, formation of the shell layer does not prevent further reduction from occurring, as the reduction continues to progress steadily until a metallization of > 90 pct is achieved after 30 minutes (Figure 4). This implies that gas can readily diffuse through the shell layer.

In the sample taken after 10 minutes (Figure 9(c)), a shrinking-core type boundary is observed for the conversion to metallic iron. However, the latter stages of metallization (Figure 9(d)) do not show a core-like structure, and there is no evidence of a topochemical boundary between the wüstite and TTM phases.

Higher magnification images of the same particles are shown in Figure 10. Figure 10(a) shows a region of the particle in which no metallic iron has yet been formed after 5 minutes of reduction. Here, it can be seen that the inner region of the particle has been partially converted to form light-gray channels (white arrow) which penetrate throughout the darker gray TTM matrix (black arrow). The increased contrast indicates higher Fe/O ratios in the light-gray areas, indicating that they comprise the wüstite phase observed in XRD (see Figure 6). This phase assignment is consistent with the EDS maps shown in Figure 11. These EDS maps also show that the iron-rich wüstite regions do not contain significant levels of Al or Ti, and that the Al and Ti remain within the darker gray TTM regions. This is consistent with reaction [2].
Fig. 10

High magnification back-scatter SEM images of cross-sectioned uniform particles after reduction by 100 vol pct hydrogen at 1223 K. Images shown for reduction after: (a) 5 min, (b) 10 min, (c) 20 min, and (d) 60 min. The magnification is 8000 times

Fig. 11

Elemental EDS map of a TTM ironsand uniform grain reduced by 100 vol pct H2 at 1223 K for 5 min. (a) is the BSE image, and (b) to (f) refer to the element noted in each image

The interlaced wüstite-TTM microstructure in Figures 10(a) and 11 also provides evidence of the mechanism by which Ti and Al migration affects the reduction process. The low solubility of Al and Ti in wüstite,[11,35,46] means that these species must be exsolved from all wüstite as it is formed (as described in reaction [2]). This is because Ti4+ and Al3+ ions can substitute onto octahedral Fe3+ sites in the TTM crystal structure,[47, 48, 49] but are not readily incorporated onto the solely Fe2+ sites within the wüstite lattice. By contrast, Mg2+ can substitute onto Fe2+ sites in both TTM and wüstite crystal lattices implying that Mg2+ is not exsolved as wüstite is formed. This is confirmed in Figure 11(e) where it can be seen that the distribution of Mg within the particle remains homogenous. Locally, this leads to the formation of the wüstite channels surrounded by Ti-enriched TTM as shown in Figure 11. This process continues until all of the surrounding Ti-enriched TTM becomes more stable to reduction than the wüstite channels. At this point, wüstite is then preferentially reduced to metallic iron, marking the transition point between stages 1 and 2 of the reduction process.

Similar bright-channel microstructures have also been observed during the early reduction of the TTM ironsand in a fixed bed reactor by References 3 and 36, although the phase was neither identified nor explained. It should be noted that this microstructure is quite different from that reported for conventional (i.e., titanium-free) magnetite and hematite ores. In those ores, the conversion of TTM to wüstite typically proceeds topochemically such that a wüstite ring develops from the particle surface, and this spatially separates the interface between any subsequently formed metallic iron and unreacted magnetite.[44,45,50]

Figure 10(b) shows the phase evolution during stage 2 of the reaction when wüstite is being converted to metallic iron. The inner region of the particle shows a similar interlaced structure of wüstite and TTM as observed in Figure 10(a). However, outside of this region bright white channels of iron are observed. These channels follow the same pattern as the wüstite through the surrounding TTM matrix. This is consistent with the conversion of wüstite to iron, while the surrounding Ti-stabilized TTM remains unreduced, as described in reaction [3]. Figure 10(c) shows the situation after 20 minutes of reduction, during stage 3 of the reduction. At this point, all of the wüstite has been reduced and metallization proceeds via the direct conversion of the remaining Ti-enriched TTM regions which are interspersed between the channels of metallic iron (black arrow). As a result, the fine channels of iron become thicker, and some voids (white arrow) appear due to the volume change as oxygen is removed from the TTM lattice. These voids will slightly reduce the apparent density of the particles within the fluidized bed, which could affect the fluidization behavior. However, the gas flow rate has been chosen to lie well inside the bubbling bed regime, [42] to ensure that such changes in particle density will not affect the operating mode of the bed.

Figure 10(d) shows the end-point of the reduction after 60 minutes, at which point 93 pct metallization has been achieved. The interior region is now fully occupied by porous metallic iron and small regions of gray slag oxides. Interestingly, Figure 10(d) is the first image in which metal precipitation is observed in the outer shell of the particle. This suggests that the small iron precipitations (white arrow) observed on the surface of the final particles (Figure 10(d)) are formed at a very late stage in the process and are a result of the final reduction of the Ti-rich phases in stage 4 of the reaction.

Development of the protective outer oxide shell

The outer oxide shell of each particle forms within the first 5 minutes of the reduction and is maintained throughout the reduction process (Figures 9(b) through (f)). Although only a few microns thick, this shell provides a protective barrier which prevents metallic iron-to-iron contact between particles during fluidization. It is likely that the formation of this shell around the uniform grains is the key reason that the TTM ironsand investigated in this study did not stick during fluidized bed reduction at 1223 K. An important point to note is that formation of the protective shell is a direct result of reduction by hydrogen and is not simply caused by heating under inert conditions. Samples taken following fluidization of ironsand in Ar gas at 1223 K for 30 minutes do not form any observable outer shell.

The approximate elemental composition of the shell at each stage of the reduction was determined by EDS point analysis for each of the numbered spots shown in Figure 10. This EDS data is shown in Table II.
Table II

Elemental Spot EDS Analysis of Points Shown in Figs.  9 and 10


Point (At. Pct)










































































In the early stage of reduction (Figure 10(a)), the shell is typically rich in titanium and aluminum (points 1, 3, 6, and 7) compared to the inner particle (points 2, 4, and 5). The initial development of the shell likely occurs via a process similar to that which causes the segregation of wüstite and Ti-enriched TTM within the interior of the particle. During the initial reduction stages of the particle, oxygen is removed from the outer layer of TTM. The remaining species combine to form a Ti (and Al) enriched TTM, but no wüstite is found at the outer surface of the particle. Instead the excess iron species seem to migrate inwards, and only form wüstite within the interior of the particle. We speculate that this might be due to the higher relative concentration of [H2]/[H2O] at the particle surface, compared to its interior. This appears to prevent wüstite from forming at the outer surface, such that the surface oxide is quickly depleted in Fe and correspondingly enriched in Ti and Al. As the reaction proceeds, further wüstite is formed within the particle and additional Al and Ti ions migrate to the outer shell. As a result, the outer shell becomes thicker as the reduction proceeds (Figures 10(a) through (c)), and the Fe/Ti ratio in the shell becomes progressively smaller over time (see points 1, 3, 6, and 7 in Table II).

Figure 12 plots the change in the Fe, Ti, Al, and Mg content in the protective shell as the reduction process proceeds. The value shown for each element in Figure 12 is taken from an average of the EDS spot analysis from eight locations in each sample (from four particles), with the error bar denoting the standard error. The M value plotted on the y-axis is defined in Eq. [5],
$$ M = M_{\text{i}} /({\text{pct}}\;{\text{Fe}} + {\text{pct}}\;{\text{Ti}} + {\text{pct}}\;{\text{Al}} + {\text{pct}}\;{\text{Mg}}), $$
where Mi denotes the measured at. pct of the element of interest (i.e., Fe, Ti, Al, or Mg). All values are in at. pct. By plotting this M value, the relative ratio of the key metallic species within the shell oxide over time can be followed. From this figure it can be seen that there is a marked decrease of iron content in the shell as the reduction progresses, while the Ti content increases substantially. The relative concentrations of Al and Mg also increase over time. We find that the average Fe/Ti ratio in the shell in after 60 minutes is 0.97 ± 0.20, which is comparable to the Fe/Ti ratio expected for Ilmenite (= 1.00). As discussed previously, ilmenite and small amount of TTM phases are reduced to metallic iron, rutile, and pseudobrookite in the final stage of reduction. Figure 10(d) shows that small metallic iron precipitates have started to form in the shell after 60 minutes. This implies that in the final stage of the reaction the oxide shell is also being reduced.
Fig. 12

Plot showing relative change in ratios of key metallic elements within the shell, as a function of time during the reduction by 100 vol pct H2 at 1223 K. The value M plotted on the y-axis is defined in Eq. [5]

Figure 13 shows an EDS map of a particle reduced for 60 minutes, again emphasizing that the iron content of the shell matrix is greatly depleted compared to the inner region of the particle. The shell is rich in Ti, Al, and Mg as indicated by point 7 in Table II. Besides in the shell, small (sub-micron) exsolved regions of (Ti, Al, Mg) oxides are also found interspersed within the metallic iron.
Fig. 13

Elemental EDS map of the TTM ironsand uniform grain reduced by 100 vol pct H2 at 1223 K for 60 min. (a) is the BSE image, (b) to (f) refer to the element noted in each image

Microstructural evolution of non-uniform lamella TTH/TTM particles

A small proportion (< 10 pct) of the initial ironsand particles contain lamellae of TTH within the TTM matrix. Although these particles do not dominate the observed bulk reduction behavior, they are the main source of TTH in the initial unreduced sample. It is noteworthy that the reduction process in these lamellar particles initially proceeds quite differently from that of the uniform TTM particles. Figures 14 and 15 show SEM images of non-uniform lamellar particles at each stage of the reduction process, at low and high magnifications, respectively. It is apparent that reduction initiates along each lamella, and metallization then subsequently encroaches into each of the surrounded TTM regions.
Fig. 14

The microstructural evolution of non-uniform lamella particles during the reduction by 100 vol pct hydrogen at 1223 K: (a) original ironsand prior to reduction process, (b) after 5 min of reduction, (c) after 10 min of reduction, (d) after 20 min of reduction, (e) after 40 min of reduction, and (f) after 60 min of reduction

Fig. 15

High magnification image (16,000 times) of non-uniform lamella particles during the reduction by 100 vol pct hydrogen at 1223 K: (a) after 5 min of reduction, (b) after 10 min of reduction, (c) after 20 min of reduction, and (d) after 60 min of reduction

Figures 14(b) and 15(a) show that the reduction of the TTH lamellae occurs in the first 5 minutes, which is consistent with the disappearance of TTH from the XRD traces within this timescale (Figure 6(e)). At this point, voids (white arrow) have already formed within the lamella, and these may enable rapid gas diffusion along these channels during the latter stages of the reaction. As a result, the metallization front within each enclosed region of TTM appears to start at the surrounding lamella boundary and progress inwards from there (as shown for example in Figures 15(b) and (c)).

Table III shows data from EDS point analysis of selected points in Figure 15. Point 8 (Figure 15(a)) shows that after 5 minutes of reduction the lamellar region exhibits a Fe/Ti ratio similar to that observed in the outer shell of the uniform particles at the same time. This implies that a similar migration of excess Fe has occurred out of the lamellar regions. Evidence for this are the bright white iron layers which lie on either side of the lamella.
Table III

Elemental Spot EDS Analysis of Points Shown in Fig. 15


Point (At. Pct)













































In the top right region of Figure 15(a), a region of intermixed wüstite and TTM can be observed. Similarly to the uniform grains after the same reduction time, bright gray channels of wüstite are found to interlace throughout the TTM matrix. As a result, after 5 minutes each lamella-particle comprises multiple regions of wüstite-TTM which are each bounded by thin lamina of porous Ti-stabilized oxide. During the subsequent stages of reduction, the Ti-enriched lamellae then play a similar role to that played by the oxide protective shell in the uniform particles, and reduction proceeds through the same basic stages. Figure 15(b) shows that metallization during stage 2 follows the wüstite channels. Figure 15(c) then shows the transition to stage 3 of the reduction process in which direct metallization of the residual Ti-enriched TTM occurs. Figure 15(d) shows the final state after 60 minutes of reduction, with regions of porous metallic iron (white arrows) surrounded by the oxide lamella.


The fluidized bed reduction of TTM ironsand by H2-Ar gas mixture at 1223 K has been demonstrated and investigated. Metallization of about 93 pct was achieved in 30 minutes when the ironsand was fluidized in 100 pct H2 gas. Importantly, there were no occurrences of sticking between particles within the fluidized bed at any point during the experimental series. Sticking appears to be inhibited by the formation of a protective outer oxide shell on the majority of particles, which prevents iron-iron contact at the surface of each particle. This outer shell forms in the initial stage of the reduction as iron species migrate inwards, leaving an (Ti, Al)-rich oxide layer around the exterior of the particle.

The reduction process is observed to proceed in four distinct stages. In the first stage, TTH lamellae are reduced, while at the same time approximately half of the TTM in the original sample is reduced to wüstite. In the second stage, wüstite is reduced to iron metal while the remaining TTM is not reduced any further. A significant decrease in the reduction rate of TTM is observed during this stage, which may be due to titanium enrichment caused by the migration of species exsolved from the wüstite regions. In the third stage, the remaining Ti-enriched TTM is reduced to iron and ilmenite. In the fourth stage, ilmenite is partially converted to rutile and then to pseudobrookite.

The titanium content of the ironsand plays a highly significant role in the progress of this reduction, as it is primarily responsible for the formation of the protective outer oxide shell. In addition, the low solubility of Ti in wüstite also prevents the total conversion of TTM to wüstite in the second reduction stage. Instead channels of wüstite are formed which interlace the Ti-enriched TTM matrix. These channels are then preferentially reduced to iron metal, prior to reduction of the surrounding Ti-enriched TTM.

A minority of particles in the original TTM ironsand contained multiple TTH lamellae running through the TTM matrix. In these non-uniform particles, the initial reduction proceeded along each lamella to form porous Ti-rich oxide similar to that observed in the shell formed around the uniform TTM particles. Subsequent reduction then proceeded through the same stages as observed in the majority uniform particles.

In summary, our findings show that titanium bearing ironsand can be reduced at 1223 K without incurring the usual intractable problems associated with iron particle sticking in the fluidized bed. This opens the way to develop novel high throughput fluidized bed processes for the production of direct-reduced iron (DRI) from TTM ironsand. A further important feature of the final product from this process is that the thin oxide shells seem to render each particle air-stable without further treatment. This property means that this product might ultimately be shipped as a free-flowing powder, without the need for hot-briquetting or other measures typically used to mitigate the highly oxidizable nature of other types of DRI products.



This research was supported by funding received from the Endeavour Fund of the New Zealand Ministry of Business Innovation and Employment (Grant No. RTVU1404).

Supplementary material

11663_2019_1625_MOESM1_ESM.doc (206 kb)
Supplementary material 1 (DOC 205 kb)


  1. 1.
    1 E. Park and O. Ostrovski: ISIJ Int., 2004, vol. 44, pp. 999–1005.CrossRefGoogle Scholar
  2. 2.
    2 T. Hu, X. Lv, C. Bai, Z. Lun, and G. Qiu: Metall. Mater. Trans. B, 2013, vol. 44, pp. 252–60.CrossRefGoogle Scholar
  3. 3.
    Z. Wang, J. Zhang, Z. Liu, K. Jiao, and X. Xing: J. Miner. Met. Mater. Soc., 2018, Scholar
  4. 4.
    4 X.F. She, H.Y. Sun, X.J. Dong, Q.G. Xue, and J.S. Wang: J. Min. Metall. Sect. B Metall., 2013, vol. 49, pp. 263–70.CrossRefGoogle Scholar
  5. 5.
    5 H. Sun, J. Wang, Y. Han, X. She, and Q. Xue: Int. J. Miner. Process., 2013, vol. 125, pp. 122–28.CrossRefGoogle Scholar
  6. 6.
    T. Christie and B. Brathwaite: Mineral Commodity Report 15Iron, vol. 22, 1997. Accessed 12 November 2018.
  7. 7.
    7 R.L. Brathwaite, M.F. Gazley, and A.B. Christie: J. Geochemical Explor., 2017, vol. 178, pp. 23–34.CrossRefGoogle Scholar
  8. 8.
    8 N. Evans: Steel TImes Int., 1986, vol. 10, pp. 38–40.Google Scholar
  9. 9.
    9 S. Hayashi: Ironmak. Steelmak., 2015, vol. 42, pp. 233–40.CrossRefGoogle Scholar
  10. 10.
    10 E. Park and O. Ostrovski: ISIJ Int., 2003, vol. 43, pp. 1316–25.CrossRefGoogle Scholar
  11. 11.
    R.J. Longbottom, B. Ingham, M.H. Reid, A.J. Studer, C.W. Bumby, and B.J. Monaghan: Miner. Process. Extr. Metall., 2018, Scholar
  12. 12.
    12 T. Zhang, C. Lei, and Q. Zhu: Powder Technol., 2014, vol. 254, pp. 1–11.CrossRefGoogle Scholar
  13. 13.
    13 J.L. Schenk: Particuology, 2011, vol. 9, pp. 14–23.CrossRefGoogle Scholar
  14. 14.
    14 C. Lei, T. Zhang, J. Zhang, C. Fan, Q. Zhu, and H. Li: ISIJ Int., 2014, vol. 54, pp. 589–95.CrossRefGoogle Scholar
  15. 15.
    15 J.F. Gransden and J.S. Sheasby: Can. Metall. Q., 1974, vol. 13, p. 1974.CrossRefGoogle Scholar
  16. 16.
    16 Q. Zhu, R. Wu, and H. Li: Particuology, 2013, vol. 11, pp. 294–300.CrossRefGoogle Scholar
  17. 17.
    17 M. Komatina and H. Gudenau: Metalurgija, 2004, vol. 10, pp. 310–28.Google Scholar
  18. 18.
    18 J. Shao, Z. Guo, and H. Tang: Steel Res. Int., 2013, vol. 84, pp. 111–8.CrossRefGoogle Scholar
  19. 19.
    19 H.W. Gudenau, J. Fang, T. Hirata, and U. Gebel: Steel Res. Int., 1989, vol. 60, pp. 138–44.CrossRefGoogle Scholar
  20. 20.
    20 Y. Zhong, Z. Wang, Z. Guo, and Q. Tang: Powder Technol., 2013, vol. 249, pp. 175–80.CrossRefGoogle Scholar
  21. 21.
    21 X. Gong, B. Zhang, Z. Wang, and Z. Guo: Metall. Mater. Trans. B, 2014, vol. 45, pp. 2050–6.CrossRefGoogle Scholar
  22. 22.
    22 S. Hayashi, S. Sayama, and Y. Iguchi: ISIJ Int., 1990, vol. 30, pp. 722–30.CrossRefGoogle Scholar
  23. 23.
    23 S. Hayashi and Y. Iguchi: ISIJ Int., 1992, vol. 32, pp. 962–71.CrossRefGoogle Scholar
  24. 24.
    24 B. Zhang, X. Gong, Z. Wang, and Z. Guo: ISIJ Int., 2011, vol. 51, pp. 1403–9.CrossRefGoogle Scholar
  25. 25.
    25 J. Shao, Z. Guo, and H. Tang: ISIJ Int., 2011, vol. 51, pp. 1290–5.CrossRefGoogle Scholar
  26. 26.
    26 L. Guo, J. Tang, H. Tang, and Z. Guo: Mater. Today Proc., 2015, vol. 2, pp. S332–41.CrossRefGoogle Scholar
  27. 27.
    27 S. Hayashi, S. Sawai, and Y. Iguchi: ISIJ Int., 1993, vol. 33, pp. 1078–87.CrossRefGoogle Scholar
  28. 28.
    28 Y. Zhong, Z. Wang, Z. Guo, and Q. Tang: Powder Technol., 2013, vol. 241, pp. 142–8.CrossRefGoogle Scholar
  29. 29.
    29 D. Neuschütz: Steel Res. Int., 1991, vol. 62, pp. 333–7.CrossRefGoogle Scholar
  30. 30.
    30 K. Miyagwa, T. Kamijo, and M. Deguchi: J. Iron Steel Inst. Japan, 1992, vol. 782, pp. 1258–65.CrossRefGoogle Scholar
  31. 31.
    31 X. Gong, Z. Zhao, Z. Wang, B. Zhang, and L. Guo: Metall. Mater. Trans. B, 2016, vol. 47, pp. 1137–46.CrossRefGoogle Scholar
  32. 32.
    32 W. Shannon, W. Kitt, and T. Marshall: NEW Zeal. J. Sci., 1960, vol. 3, pp. 74–90.Google Scholar
  33. 33.
    33 H. Sun, A.A. Adetoro, Z. Wang, F. Pan, and L. Li: ISIJ Int., 2016, vol. 56, pp. 936–43.CrossRefGoogle Scholar
  34. 34.
    H. Sun, A.A. Adetoro, F. Pan, Z. Wang, and Q. Zhu: Metall. Mater. Trans. B Process Metall. Mater. Process. Sci., 2017, vol. 48, pp. 1898–1907.CrossRefGoogle Scholar
  35. 35.
    35 R.J. Longbottom, O. Ostrovski, and E. Park: ISIJ Int., 2006, vol. 46, pp. 641–6.CrossRefGoogle Scholar
  36. 36.
    36 Z. Wang, J. Zhang, J. Ma, and K. Jiao: ISIJ Int., 2017, vol. 57, pp. 443–52.CrossRefGoogle Scholar
  37. 37.
    37 F. Bosi, U. Hålenius, and H. Skogby: Am. Mineral., 2009, vol. 94, pp. 181–9.CrossRefGoogle Scholar
  38. 38.
    38 H. Tanaka and M. Kono: J. Geomagn. Geoelectr., 1987, vol. 39, pp. 463–75.CrossRefGoogle Scholar
  39. 39.
    E. Park and O. Ostrovski: 2004, vol. 44, pp. 74–81.CrossRefGoogle Scholar
  40. 40.
    Z. Wang, D. Pinson, S. Chew, H. Rogers, B.J. Monaghan, M.I. Pownceby, N.A.S. Webster, and G. Zhang: Metall. Mater. Trans. B Process Metall. Mater. Process. Sci., 2016, vol. 47, pp. 330–43.CrossRefGoogle Scholar
  41. 41.
    41 J.B. Wright: NEW Zeal. J. Geol. Geophys., 1963, vol. 7, pp. 424–44.CrossRefGoogle Scholar
  42. 42.
    Wen C. Yang: in Handbook of Fluidization and Fluid-particle System, Marcel Dekker, Inc., 2003, p. 53.CrossRefGoogle Scholar
  43. 43.
    International Standard ISO 16878:2016: Iron Ores—Determination of Metallic Iron Content—Iron(III) Chloride Titrimetric Method, 2016.Google Scholar
  44. 44.
    44 B. Weiss, J. Sturn, S. Voglsam, S. Strobl, H. Mali, F. Winter, and J. Schenk: Ironmak. Steelmak., 2011, vol. 38, pp. 65–73.CrossRefGoogle Scholar
  45. 45.
    45 A. Pichler, H. Mali, F. Plaul, J. Schenk, M. Skorianz, and B. Weiss: Steel Res. Int., 2016, vol. 87, pp. 642–52.CrossRefGoogle Scholar
  46. 46.
    46 G.D. McAdam: Ironmak. Steelmak., 1974, vol. 1, pp. 138–50.Google Scholar
  47. 47.
    J.C.W. Richards, J.B.O. Donovan, Z. Hauptman, W.O. Reilly, and K.M. Creer: 1973, vol. 7, pp. 437–44.CrossRefGoogle Scholar
  48. 48.
    48 C.I. Pearce, O. Qafoku, J. Liu, E. Arenholz, S.M. Heald, R.K. Kukkadapu, C.A. Gorski, C.M.B. Henderson, and K.M. Rosso: J. Colloid Interface Sci., 2012, vol. 387, pp. 24–38.CrossRefGoogle Scholar
  49. 49.
    B.A. Wechsler, D.H. Lindsley, and Charles T. Prewitt: Am. Mineral., 1984, vol. 69, pp. 754–70.Google Scholar
  50. 50.
    50 B. Weiss, J. Sturn, S. Voglsam, S. Strobl, H. Mali, F. Winter, and J. Schenk: Steel Res. Int., 2010, vol. 81, pp. 93–9.CrossRefGoogle Scholar

Copyright information

© The Minerals, Metals & Materials Society and ASM International 2019

Authors and Affiliations

  • Sigit W. Prabowo
    • 1
    • 2
  • Raymond J. Longbottom
    • 2
  • Brian J. Monaghan
    • 2
  • Diego del Puerto
    • 3
  • Martin J. Ryan
    • 3
  • Chris W. Bumby
    • 1
    Email author
  1. 1.Faculty of Engineering, Robinson Research InstituteVictoria University of WellingtonWellingtonNew Zealand
  2. 2.Pyrometallurgy Group, School of Mechanical, Materials, Mechatronic and Biomedical EngineeringUniversity of WollongongWollongongAustralia
  3. 3.Callaghan InnovationWellingtonNew Zealand

Personalised recommendations