Mechanism and Kinetics of the Reduction of Hematite to Magnetite with CO–CO 2 in a Micro-Fluidized Bed

: The mechanism and kinetics of the reduction of hematite to magnetite in a laboratory-scale, micro-ﬂuidized bed reactor were isothermally investigated. The procedure consisted of the isothermal heating in a ﬂow of a 20%CO–80%CO 2 mixture at temperatures from 500 ◦ C to 600 ◦ C. It was found that the Avrami–Erofe’ev model of nucleation and 1D growth ( n = 1.58) successfully described the phase transition of hematite to magnetite, and the value of activation energy ∆ E a of the reaction was estimated to be 48.70 kJ/mol. The microstructure properties for specimens with different conversion degrees were analyzed using the Brunauer, Emmett and Teller (BET) method and scanning electron microscopy (SEM). The results showed that the magnetite nuclei were needle-like, and the hematite specimens became thoroughly porous after complete reduction to magnetite. The physical and chemical processes of the reaction were also discussed.


Introduction
The preparation of magnetite (Fe 3 O 4 ) has become of long-standing interest because of the diversified applications of magnetite in the industry as a pigment, in magnetic tapes, or as raw material for iron-making [1][2][3]. Therefore, it is of great interest to study innovative synthetic methods to reduce the production costs of magnetite and improve its production efficiency. Several chemical methods have been proposed to synthesize magnetite [4][5][6][7][8]. Fluidized reduction roasting has been proven to bean industrially economical and efficient method for the chemical synthesis of magnetite, with hematite as the raw material [9,10]. The reaction in fluidized bed reactors offers the advantages of a uniform temperature in the reactor, and of fast heat and mass transfer [11].
Several researchers have examined the reduction of synthetic Fe 2 O 3 reagents or of hematite to magnetite by gaseous reductants, and the data reported can be summarized in Table 1. According to the reported results, the ∆E a (activation energy) for the reduction of synthetic Fe 2 O 3 reagents or for hematite to magnetite varies from 33.28 kJ/mol to 139.2 kJ/mol, and there are four mechanism models for this reaction, including the random nucleation model, the phase boundary model, the non-topochemical model, and the unimolecular model. The difference between these mechanism models may be caused by the different experimental conditions, which include thereactor type, the reduction temperature, the gaseous atmosphere, and the particle size. Especially, the solid material used in the previous experiments consistedmostly of synthetic Fe 2 O 3 reagents with the same chemical composition as natural hematite (Fe 2 O 3 ). However, the physicochemical properties, such as the specific surface area, density, hardness, and granularity of the synthetic Fe 2 O 3 reagents and of natural hematite are different. Therefore, the kinetics parameters and mechanisms for the reduction of synthetic Fe 2 O 3 to magnetite cannot fully explain the process and mechanism for the reduction of hematite to magnetite.  [25] In this work, in order to avoid the overreduction of the newly generated magnetite to wüstite (FeO), a mixture of 20% CO-80% CO 2 was used as a reducing reagent, which was thermodynamically in equilibrium with the magnetite. The kinetics of the reduction of hematite to magnetite in a fluidized bed reactor at temperatures between 500 • C and 600 • C (the temperature range is consistent with the industrial production temperature) was studied for the first time. Meanwhile, the microstructure changes after reduction roasting were analyzed by the Brunauer, Emmett and Teller (BET) method and scanning electron microscopy (SEM), and the physical and chemical processes of the reduction of hematite to magnetite with CO-CO 2 in a micro-fluidized bed were also discussed.

Materials
The material used in this study was a powdered natural hematite ore (containing 93.5 % Fe 2 O 3 and 6.5% SiO 2 , w/w). The X-ray diffraction pattern of the hematite powder is presented in Figure 1, indicating that the main phase structures of the sample werehematite (Fe 2 O 3 ) and quartz (SiO 2 ). The grain size of the hematite ore was 74 µm-100 µm. The SEM image of the sample is shown in Figure 2, indicating that the surface of the sample wasrough. The BET surface area and porosity volume of the hematite ore were 0.669 m 2 /g and 0.003 m 3 /g, respectively. A gas mixture of 20% CO-80% CO 2 with a purity of 99.9%was used as a reducing agent, and nitrogen (N 2 ) with a purity of 99.9% was used as a protective gas.

Experimental Apparatus and Procedures
All the experiments of this study were conducted in a cylindrical quartz fluidization furnace tube with an inner diameter of 20 mm, and the schematic diagram of the experimental apparatus is shown in Figure 3. A perforated quartz plate bed containing holes of 15 µm in diameter wasfitted in the middle of the fluidization furnace tube. . Schematic diagram of the experimental apparatus: 1, gas cylinder; 2, gas valve; 3, mass flow meter; 4, mixing cylinder; 5, gas drying; 6, heating furnace; 7, fluidized bed (porous quartz plate); 8, thermocouple; 9, temperature controller.

Experimental Apparatus and Procedures
All the experiments of this study were conducted in a cylindrical quartz fluidization furnace tube with an inner diameter of 20 mm, and the schematic diagram of the experimental apparatus is shown in Figure 3. A perforated quartz plate bed containing holes of 15 µm in diameter wasfitted in the middle of the fluidization furnace tube.

Experimental Apparatus and Procedures
All the experiments of this study were conducted in a cylindrical quartz fluidization furnace tube with an inner diameter of 20 mm, and the schematic diagram of the experimental apparatus is shown in Figure 3. A perforated quartz plate bed containing holes of 15 µm in diameter wasfitted in the middle of the fluidization furnace tube.

Experimental Apparatus and Procedures
All the experiments of this study were conducted in a cylindrical quartz fluidization furnace tube with an inner diameter of 20 mm, and the schematic diagram of the experimental apparatus is shown in Figure 3. A perforated quartz plate bed containing holes of 15 µm in diameter wasfitted in the middle of the fluidization furnace tube.  The powdered hematite ore was initially preheated (10 • C/min) under flowing nitrogen (1000 mL/min) to the predetermined temperature (selected from the 500 • C-600 • C range). At the defined temperature, the nitrogen flow was switched to the flow of the reducing gas mixture (1000 mL/min, gas mixture composition: 20% CO + 80% CO 2 ). After roasting for a given time, the reducing gas mixture was switched back to pure nitrogen, and the reactor was cooled down to ambient temperature. Before entering the micro-fluidized bed reactor, gases were dried by anhydrous calcium chloride. All tests were carried out with 5 g of fresh hematite powder.

Methods
The surface area, porosity volume, and pore size measurements were carried out using the BET (liquid N 2 ) method based on the adsorption of N 2 at −197 • C, withan ASAP 2020-Physisorption Analyzer instrument (Micromeritics Instrument Corp, Atlanta, GA, USA), and the microstructures of the roasted specimens were analyzed by SEM (S-3400N; Hitachi, Ltd., Tokyo, Japan). The kinetics of the hematite-magnetite reduction process was determined by monitoring the change of ferrous oxide content (w FeO ) in the specimens during their transformation under selected isothermal conditions. At the end of the reduction test, the roasted specimens were ground to pass through a 74 µm sieve. The content of FeO (w FeO ) in the roasted specimens was determined by chemical analysis.
We considered the following chemical reaction for hematite to magnetite: It can be noted that, based on the above reaction's stoichiometry, the w FeO of the roasted specimens corresponds to the conversion degree α (of Fe 2 O 3 to Fe 3 O 4 ). Based on the law of mass conservation, the numerical relationship between w FeO and α is shown in Equation (2): The reaction rate, r, can be calculated by the following Equation (3):

Determination of the Kinetic Model
The α = f (t) and r = f (t) experimental profiles, calculated with Equations (2) and (3), and corresponding to different process temperatures, are presented in Figures 4 and 5, respectively.
The powdered hematite ore was initially preheated (10 °C/min) under flowing nitrogen (1000 mL/min) to the predetermined temperature (selected from the 500 °C-600 °C range). At the defined temperature, the nitrogen flow was switched to the flow of the reducing gas mixture (1000 mL/min, gas mixture composition: 20% CO + 80% CO2). After roasting for a given time, the reducing gas mixture was switched back to pure nitrogen, and the reactor was cooled down to ambient temperature. Before entering the micro-fluidized bed reactor, gases were dried by anhydrous calcium chloride. All tests were carried out with 5 g of fresh hematite powder.

Methods
The surface area, porosity volume, and pore size measurements were carried out using the BET (liquid N2) method based on the adsorption of N2 at −197 °C,withan ASAP 2020-Physisorption Analyzer instrument (Micromeritics Instrument Corp, Atlanta, GA, USA), and the microstructures of the roasted specimens were analyzed by SEM (S-3400N; Hitachi, Ltd., Tokyo, Japan). The kinetics of the hematite-magnetite reduction process was determined by monitoring the change of ferrous oxide content (wFeO) in the specimens during theirtransformation under selected isothermal conditions. At the end of the reduction test, the roasted specimens were groundto pass through a 74 µm sieve . The content of FeO (wFeO) in the roasted specimens was determined by chemical analysis.
We considered the following chemical reaction for hematite to magnetite: It can be noted that, based on the above reaction's stoichiometry, the wFeO of the roasted specimens corresponds to the conversion degree α (of Fe2O3 to Fe3O4). Based on the law of mass conservation, the numerical relationship between wFeO and α is shown in Equation (2): The reaction rate, r, can be calculated by the following Equation (3):

Determination of the Kinetic Model
The α = f(t) and r = f(t) experimental profiles, calculated with Equations (2) and (3), and corresponding to different process temperatures, are presented in Figures 4 and 5, respectively.  All α = f (t) conversion curves at different temperatures were S-shaped ( Figure 4) and couldbe divided into three different periods, namely, induction, acceleration, and deceleration period ( Figure 5). As shown in Figure 4, the incubation period shortened and the reaction rate increased with the increase of the roasting temperature.
Minerals 2017, 7, 209 5 of 12 All α = f(t) conversion curves at different temperatures were S-shaped ( Figure 4) and couldbe divided into three different periods, namely,induction, acceleration, and deceleration period ( Figure 5). As shown in Figure 4, the incubation period shortened and the reaction rate increased with the increase of the roasting temperature. The S-shaped α = f(t) curves are commonly analyzed by the Hancock and Sharp's method [26,27]: where α is the conversion degree of Fe2O3 to Fe3O4 at a certain time t, β is the isothermal crystallization rate constant, and n is the Avrami exponent related to the growth and morphology of the crystals. The calculation results obtained by applying Equation (5) tothe test data are presented in Figure 6, and their linear fitting analysis resultsare listed in Table 2. The S-shaped α = f (t) curves are commonly analyzed by the Hancock and Sharp's method [26,27]: where α is the conversion degree of Fe 2 O 3 to Fe 3 O 4 at a certain time t, β is the isothermal crystallization rate constant, and n is the Avrami exponent related to the growth and morphology of the crystals. The calculation results obtained by applying Equation (5) tothe test data are presented in Figure 6, and their linear fitting analysis resultsare listed in Table 2. All α = f(t) conversion curves at different temperatures were S-shaped ( Figure 4) and couldbe divided into three different periods, namely,induction, acceleration, and deceleration period ( Figure 5). As shown in Figure 4, the incubation period shortened and the reaction rate increased with the increase of the roasting temperature. The S-shaped α = f(t) curves are commonly analyzed by the Hancock and Sharp's method [26,27]: ln(−ln(1 − α)) = lnβ + nlnt (5) where α is the conversion degree of Fe2O3 to Fe3O4 at a certain time t, β is the isothermal crystallization rate constant, and n is the Avrami exponent related to the growth and morphology of the crystals. The calculation results obtained by applying Equation (5) tothe test data are presented in Figure 6, and their linear fitting analysis resultsare listed in Table 2.  An average value of n = 1.58 was obtained, suggesting that the universal Avrami-Erofe'ev kinetic model describing nucleation and growth processes, shown in Equation (6), mightbe applicable for the modeling of the Fe 2 O 3 to Fe 3 O 4 reductionreaction: In this equation, the parameter k isthe kinetic (reaction rate) constant. Meanwhile, the values of n = 1−2 represent the 1D growth of nuclei [28]. Thus, this equationindicates that this phase transition of hematite to magnetite proceeded through a 1D growth process, suggesting that the newly generated magnetite nuclei wereneedle-like, which wasalso proven by Et-Tabirou, Swann, and Bahgat [29][30][31]. Additionally, the kinetic constant (k) values for different roasting temperatures could be obtained from the experimental α = f (t) data through the Avrami-Erofe'ev equation (n = 1.58). The calculation results obtained by applying Equation (6) to the test data are presented in Figure 7, and their linear fitting analysis results are listed in Table 3.  An average value of n = 1.58 was obtained, suggesting that the universal Avrami-Erofe'ev kinetic model describing nucleation and growth processes, shown in Equation (6), mightbe applicable for the modeling of the Fe2O3 to Fe3O4 reductionreaction: In this equation, the parameter k isthe kinetic (reaction rate) constant. Meanwhile,the values of n = 1−2 represent the 1D growth of nuclei [28]. Thus, this equationindicates that this phase transition of hematite to magnetite proceeded through a 1D growth process, suggesting that the newly generated magnetite nuclei wereneedle-like, which wasalso proven by Et-Tabirou, Swann, and Bahgat [29][30][31]. Additionally, the kinetic constant (k) values for different roasting temperatures could be obtained from the experimental α = f(t) data through the Avrami-Erofe'ev equation (n = 1.58). The calculation results obtained by applying Equation (6) to the test data are presented in Figure 7, and their linear fitting analysis results are listed in Table 3.  As shown in Table 3, the kinetic constant k increasedwith the increase of the roasting temperature, indicating that the required roasting time for the total conversion of Fe2O3 to Fe3O4was gradually shortened with the increase of the roasting temperature.  As shown in Table 3, the kinetic constant k increasedwith the increase of the roasting temperature, indicating that the required roasting time for the total conversion of Fe 2 O 3 to Fe 3 O 4 was gradually shortened with the increase of the roasting temperature. Figure 8 exhibits the graphical comparison between the experimental and the modeled values of α = f (t) using theAvrami-Erofe'ev topochemical kinetic model (n = 1.58). It is clearly visible that the experimental data do not exactly coincide with the theoretical predicted values. This is because this kinetic model only applies to the surface reaction of the hematite particles, without considering the influence of mass or heat transfer resistances during the reaction process [15]. However, the mass and heating transfer resistances occur, induced by the gradually thicker magnetite layer during the fluidized reduction process of hematite, and lead to a deviationbetween the experimental data and the theoretical predicted values.
Minerals 2017, 7, 209 7 of 12 Figure 8 exhibits the graphical comparison between the experimental and the modeled values of α = f(t) using theAvrami-Erofe'ev topochemical kinetic model (n = 1.58). It is clearly visible that the experimental data do not exactly coincide with the theoretical predicted values. This is because this kinetic model only applies to the surface reaction of the hematite particles, without considering the influence of mass or heat transfer resistances during the reaction process [15]. However, the mass and heating transfer resistances occur, induced by the gradually thicker magnetite layer during the fluidized reduction process of hematite, and lead to a deviationbetween the experimental data and the theoretical predicted values. The data from Table 3 were then usedto calculate the pre-exponential factor (A) and the activation energy (ΔEa) values from the Arrhenius equations, Equations (7) and (8): where k is the kinetic constant, A the pre-exponential factor, R the gas constant (8.314 J/mol·K), and ΔEa the activation energy. The linear fitting of the function lnk vs. 1/T for the selected process The data from Table 3 were then usedto calculate the pre-exponential factor (A) and the activation energy (∆E a ) values from the Arrhenius equations, Equations (7) and (8): where k is the kinetic constant, A the pre-exponential factor, R the gas constant (8.314 J/mol·K), and ∆E a the activation energy. The linear fitting of the function lnk vs. 1/T for the selected process temperatures is plotted in Figure 9. From the regression calculations, the following were obtained: A = 5.58 s −1 , ∆E a = 48.70 kJ/mol. Thus, k = 5.58exp(−48700/RT). temperatures is plotted in Figure 9. From the regression calculations, the following were obtained: A = 5.58 s −1 , ΔEa = 48.70 kJ/mol. Thus, k = 5.58exp(−48,700/RT).

Microstructure Changes after Reduction Roasting
In order to investigate the microstructure changes during the reduction process, measurements of the surface area, porosity volume, and pores size were carried out for specimens with different conversion degrees, using the BET (liquid N2) method. The resultsareshown in Table 4. It is obvious that the surface area, porosity volume, and pore size of the specimens greatly increased with the increasing of the conversion degree of hematite to magnetite.  Figure 10 presents the microstructures of roasted samples at 600 °C atdifferent times. As shown in Figure 10a, the conversion of hematite to magnetite initially occurred at the outer surface of the hematite particles because of the presence of an external magnetite layer in the original hematite particle and of the fact that the fresh magnetite nuclei were needle-like (Figure 10d). With time, the needle-like magnetite nuclei increased, and the fresh magnetite particles were randomly filled with micropores and cracks (Figure 10b). Eventually, the original hematite particles were completely transformed into magnetite particles with a porous structure (Figure 10c). As mentioned above, the reduction process of hematite to magnetite may be theoretically divided into three periods, namely,an induction period (0 < α < 0.15), an acceleratory period (0.15 < α < 0.50), and a deceleratory period (0.5 < α < 1) [15,28]. In the initial induction period, the fresh microcracks were formed as a result of normal swelling due to the crystal structure difference between the starting hematite (hexagonal closed-packed) and the newly formed magnetite (cubic spinal). During the second acceleratory period, the microcracks increased and enlarged with more magnetite nuclei

Microstructure Changes after Reduction Roasting
In order to investigate the microstructure changes during the reduction process, measurements of the surface area, porosity volume, and pores size were carried out for specimens with different conversion degrees, using the BET (liquid N 2 ) method. The resultsareshown in Table 4. It is obvious that the surface area, porosity volume, and pore size of the specimens greatly increased with the increasing of the conversion degree of hematite to magnetite.  Figure 10 presents the microstructures of roasted samples at 600 • C atdifferent times. As shown in Figure 10a, the conversion of hematite to magnetite initially occurred at the outer surface of the hematite particles because of the presence of an external magnetite layer in the original hematite particle and of the fact that the fresh magnetite nuclei were needle-like (Figure 10d). With time, the needle-like magnetite nuclei increased, and the fresh magnetite particles were randomly filled with micropores and cracks (Figure 10b). Eventually, the original hematite particles were completely transformed into magnetite particles with a porous structure ( Figure 10c). As mentioned above, the reduction process of hematite to magnetite may be theoretically divided into three periods, namely, an induction period (0 < α < 0.15), an acceleratory period (0.15 < α < 0.50), and a deceleratory period (0.5 < α < 1) [15,28]. In the initial induction period, the fresh microcracks were formed as a result of normal swelling due to the crystal structure difference between the starting hematite (hexagonal closed-packed) and the newly formed magnetite (cubic spinal). During the second acceleratory period, the microcracks increased and enlarged with more magnetite nuclei formation, and theirgrowth facilitated the gas diffusion to the unreacted inner hematite layers. Therefore, the reduction reaction progressed. As the conversion degree increased, the reduction entered the final deceleratory period, which was characterized by the termination of growth upon the impingement of different growth regions or their reaching the grain boundaries. As a result, the hematite specimens became wholly porous after the complete reduction to magnetite.
Minerals 2017, 7, 209 9 of 12 formation, and theirgrowth facilitated the gas diffusion to the unreacted inner hematite layers. Therefore, the reduction reaction progressed. As the conversion degree increased, the reduction entered the final deceleratory period, which was characterized by the termination of growth upon the impingement of different growth regions or their reaching the grain boundaries. As a result, the hematite specimens became wholly porous after the complete reduction to magnetite.

Physical and Chemical Processes
Based on the above analysis, the physical and chemical processes duringthe reduction of hematite to magnetite in a micro-fluidized bed reactor areshownschematically in Figure 11.

Physical and Chemical Processes
Based on the above analysis, the physical and chemical processes duringthe reduction of hematite to magnetite in a micro-fluidized bed reactor are shown schematically in Figure 11. formation, and theirgrowth facilitated the gas diffusion to the unreacted inner hematite layers. Therefore, the reduction reaction progressed. As the conversion degree increased, the reduction entered the final deceleratory period, which was characterized by the termination of growth upon the impingement of different growth regions or their reaching the grain boundaries. As a result, the hematite specimens became wholly porous after the complete reduction to magnetite.

Physical and Chemical Processes
Based on the above analysis, the physical and chemical processes duringthe reduction of hematite to magnetite in a micro-fluidized bed reactor areshownschematically in Figure 11. . Physical and chemical processes around and inside the hematite ore grain in the micro-fluidized bed reactor: steps 1 and 5 represent thediffusion through the gas boundary layer situated around the hematite ore grain; steps 2 and 4 represent the gas diffusion through the pores of the hematite ore grain; step 3 represents the phase-interface reaction at the crystal surface, and the diffusion of the iron ions through the dense magnetite layers. Figure 11. Physical and chemical processes around and inside the hematite ore grain in the micro-fluidized bed reactor: steps 1 and 5 represent thediffusion through the gas boundary layer situated around the hematite ore grain; steps 2 and 4 represent the gas diffusion through the pores of the hematite ore grain; step 3 represents the phase-interface reaction at the crystal surface, and the diffusion of the iron ions through the dense magnetite layers.
The hematite ore particles were fluidized in the reactor with the reducinggas mixture (1000 mL/min). The reducing gas CO diffused through the boundary layer, reaching the external surface of the hematite particles and continuing further into the micropores to the internal surface of the hematite grains. The phase-interface reaction described by Equation (1) occurred at this surface, where the reducing gas picked up oxygen and freed two electrons [9,11,29]: The electrons remained in the lattice where theyreduced Fe 3+ to Fe 2+ , as shown in Equation (10).
The generated ferrous ions and the electrons diffused through the magnetite product layer into the hematite core, where the hematite lattice deformed, and additional magnetite was formed through lattice reconstruction [9,11,29]: Oxygen would diffuse from the hematite core to the external border of magnetite, but the mobility of the oxygen anions in the magnetite was negligible in comparison with the ferrous cations [11]. The gaseous reaction product CO 2 migrated back to the gas flow where it could be removed from the fluidized bed reactor [11]. As the above process proceeded, the reduction reaction progressively extended to the inner layer. Finally, the hematite particles were completely reduced to magnetite [9]. The actual reaction is illustrated in all its phasesin Figure 12, which is modified from [11]. The hematite ore particles were fluidized in the reactor with the reducinggas mixture (1000 mL/min). The reducing gas CO diffused through the boundary layer,reaching the external surface of the hematite particles and continuing further into the micropores to the internal surface of the hematite grains. The phase-interface reaction described by Equation (1) occurred at this surface, where the reducing gas picked up oxygen and freed two electrons [9,11,29]: The electrons remained in the lattice where theyreduced Fe 3+ to Fe 2+ , as shown in Equation (10).
The generated ferrous ions and the electrons diffused through the magnetite product layer into the hematite core, where the hematite lattice deformed, and additional magnetite was formed through lattice reconstruction [9,11,29]: Oxygen would diffuse from the hematite core to the external border of magnetite, but the mobility of the oxygen anions in the magnetite was negligible in comparison with the ferrous cations [11]. The gaseous reaction product CO2 migrated back to the gas flow where it could be removed from the fluidized bed reactor [11]. As the above process proceeded, the reduction reaction progressively extended to the inner layer. Finally, the hematite particleswerecompletely reducedto magnetite [9]. The actual reaction is illustrated in all its phasesin Figure 12, which is modified from [11].

Conclusions
In this study, the kinetics of the reduction of hematite to magnetite with 20%CO-80% CO2 in a micro-fluidized bed reactor was investigated for the first time. The reduction mechanism was also discussed. The following are the major findings obtained in this study: (1) The Avrami-Erofe'ev model of nucleation and 1D growth (n = 1.58) was successfully applied to describe the phase transition of hematite to magnetite, and the value of activation energy ΔEa of the reaction was estimated to be 48.70 kJ/mol. (2) The newly formed magnetite nuclei were needle-like in shape. Microcracks were formed and increased asthe magnetite nuclei formed, and thus the original hematite sample became porous after complete reduction to magnetite.

Conclusions
In this study, the kinetics of the reduction of hematite to magnetite with 20% CO-80% CO 2 in a micro-fluidized bed reactor was investigated for the first time. The reduction mechanism was also discussed. The following are the major findings obtained in this study: (1) The Avrami-Erofe'ev model of nucleation and 1D growth (n = 1.58) was successfully applied to describe the phase transition of hematite to magnetite, and the value of activation energy ∆E a of the reaction was estimated to be 48.70 kJ/mol. (2) The newly formed magnetite nuclei were needle-like in shape. Microcracks were formed and increased asthe magnetite nuclei formed, and thus the original hematite sample became porous after complete reduction to magnetite. (3) Micropores or cracks induced by normal swelling were conducive to the reduction of hematite to magnetite.