Kinetics and Mechanism of the Simultaneous Carbothermic Reduction of FeO and MnO from High-Carbon Ferromanganese Slag

Three different carbonaceous materials consisting of pure graphite, a single coke produced from a single coal, and eucalyptus charcoal were used as the carbon material substrates. Powders of these materials were prepared and then pressed to make pellet disks with flat surfaces. To prepare the carbon powder substrates, they were crushed in a jaw crusher to less than 10 mm, ground in a tungsten carbide disc mill to a fine powder, and sieved to a size fraction of 44 to 105 μm. The charcoal powder was heat treated in CO gas at 1600 °C for 10 minutes to remove its large volatile materials content. These powders were then dried at 100 °C for 24 hours and were subsequently mixed with 3 wt pct stearic acid (binder) to achieve some green strength in the pressed substrates. The powder-binder mix was pressed into a small graphite crucible (10 mm outer diameter, 8 mm inner diameter, 3 mm height, and 1 mm deep) using 63.7 kg/cm² pressure.

A part of the study was dedicated to characterize the carbon materials. The proximate analysis of the carbon materials was determined. The specific surface area of the carbon powders was measured by Brunauer, Emmett, and Teller method using helium-nitrogen gas as the surface adsorbate agent. The structural parameters such as carbon layer spacing (d ₀₀₂) and the average crystallite size (L _{C (002)}) were determined using X-ray diffraction and applying the developed procedure by Iwashita et al.[18] Moreover, the CO₂ reactivity for carbonaceous materials was measured by thermogravimetery at 1000 °C in a 100 pct CO₂ atmosphere on the particles that ranged from 1.68 to 3.36 mm in size. The mass loss changes during the experiment were subsequently converted to the initial reaction fraction X of the sample mass. The changes in the initial reaction fraction with time (dX/dt), which were fairly linear, were calculated and used to characterize the CO₂ reactivity. The carbon material characterization has been described in detail,[19] and the results are summarized in Table I.

High purity (+99.99 pct) fine powders of CaO, MgO, SiO₂, and Al₂O₃ were mixed in a ratio of CaO/MgO = 2, CaO/Al₂O₃ = 1.34, and Al₂O₃/SiO₂ = 0.4. The mixture was then melted in a graphite crucible in air and higher than 1600 °C using an induction furnace. The obtained slag was crushed in a tungsten carbide disc mill to a fine powder and was then mixed with appropriate amount of high purity MnO (+99.95 pct) fine powder by hand in plastic containers. The obtained mixture was melted in graphite crucible in air at 1550 °C. The produced slag was again crushed in the tungsten carbide disc mill to a fine powder and was then mixed with pure 99.5 pct FeO powder (CAS: 1345-25-1 from Alfa Aesar GmbH & Co KG, Karlsruhe, Germany). The mixture was melted in platinum crucible in air at 1550 °C. The slag composition was measured by electron probe microanalyzer (EPMA) and the composition in wt pct was 38.7 pct MnO-12.1 pct CaO-24.1 pct SiO₂-9.3 pct  Al₂O₃-5.4 pct MgO-10.4 pct FeO. It is worth noting that a portion of Fe²⁺ and Mn²⁺ are oxidized to Fe³⁺ and Mn³⁺ during the melting in air. These oxides will be reduced rapidly again to Fe²⁺ and Mn²⁺ in the slag reduction experiments through reduction by carbon. In the present study the measured elemental Fe and Mn in the slags by EPMA were converted to their oxides in the form of FeO and MnO.

The sessile drop wettability technique was used to study the interaction between the synthetic slag and the carbon substrates. A schematic diagram of the experimental set up is shown in Figure 1. The carbon substrate was located in the graphite sample holder, and a 40 ± 1-mg slag particle was put on the carbon surface. The furnace chamber was evacuated initially, and then the furnace was heated up slowly in pure argon (99.9999 pct Ar) with 0.5 Nl/min flow rate. The furnace was heated to 950 °C in approximately 10 minutes, followed by a rapid heating rate of 120 °C/min to 1400 °C, and held for 4 minutes at 1400 °C. The furnace was then heated up from 1400 °C to 1600 °C in 1 minute, and then it was held for reaction times of 2, 5, and 8 minutes with rapid cooling. The first 4-minute reduction at 1400 °C was required to maintain a relatively slow FeO reduction stage so that the pressed powder substrate was not destroyed. The temperature was controlled using a Keller PZ40 two-color pyrometer (Keller GMBH, Ibbenburen, Germany) focused on the edge of the graphite sample holder. A video camera (XCD-SX910CR; Sony, Tokyo, Japan) was used to record images from the sample at 960 × 1280 pixels.

The experiments were carried out through the stop of reduction within the above-mentioned reduction times to obtain the kinetic data points. After the experiments were completed, the samples were quenched, mounted in epoxy, and prepared for EPMA supported by wavelength dispersive spectroscopy (WDS). The slag and metal chemical compositions were measured in ten different points of each phase, and the averages were considered as the total slag and metal compositions. It is worth noting that the proper reproducibility of the experimental technique has been already observed by the authors for MnO reduction from a silicate slag by graphite substrates at different temperatures.[15,19]

The ternary phase diagram for (MnO + FeO)-SiO₂-(CaO + MgO + Al₂O₃) slag system (MnO/FeO = 4, CaO/MgO = 2, CaO/Al₂O₃ = 1.34) at different temperatures between 1180 °C and 1680 °C was calculated using FactSage thermodynamics software (FactSage 5.5; C.R.C.T.-Ecole Polytechnique de Montreal, Quebec, Canada) as illustrated in Figure 2(a). Assuming a negligible SiO₂ reduction, the (MnO + FeO) content of slag during reduction changes along a straight line and forms a completely liquid slag above 1400 °C. The reduction of FeO mainly takes place before the MnO reduction because of the greater MnO stability. Thus, the slag chemical composition changes through the reduction from point A to point B with mostly FeO reduction. When the FeO content is relatively low, MnO is reduced from the MnO-SiO₂-(CaO + MgO + Al₂O₃) slag system and from point B in Figure 2(b). The MnO reduction from this point takes place again from a completely liquid slag above 1400 °C. Regarding the reduction of FeO and MnO from a liquid phase, their activities are changed with the changes of FeO and MnO concentrations during reduction. These changes were also calculated by the FactSage (FACT oxide database) for 1600 °C as illustrated in Figure 3.

Although the wettability technique was applied in the current study, the wettability parameters such as the contact angle, the rate of contact angle changes, and the contact area between the slag and carbon substrate were not used to discuss the reduction kinetics. Because considerable gas bubble formation and coexistence of the gas phase in the slag droplet make it difficult to measure the contact angle and the slag drop volume properly. Hence, the changes in the chemical compositions of the slag and the produced metal during reduction were considered to study the reduction kinetics.

The slag chemical analysis of the reduced slag indicated that the chemical composition around the slag phase is more less the same and that it is not dependent on the distance from the slag/metal and slag/carbon interfaces. It is worth noting that the standard deviation was small, which indicates that the EPMA is reliable. For example, the standard deviation for MnO concentration in 10 analyzed points was 0.4 wt pct, whereas the average MnO concentration was 38.7 wt pct. Figure 4 shows the changes in the FeO and MnO contents of slag through reduction with carbon reductants. Although the FeO activity in the slag is always less than the MnO activity (Figure 3), FeO reduction takes place with a much faster reduction rate because of the lesser thermodynamic stability than MnO. Figure 4 indicates also that the FeO reduction rate is dependent on the type of carbon material and that the greatest reduction rate is obtained with the coke and the least reduction rate is obtained with the graphite. The reduction of FeO takes place mainly at the first few minutes, and it is approximately completed for coke in 4 minutes, for charcoal in more than 4 minutes, and for graphite in around 7 minutes. Figure 4 shows that the MnO content of slag increases initially because of the FeO reduction, and then it starts to decrease so that a hump for MnO concentration appears. When the FeO initial fast reduction stage is completed, the MnO concentration is approximately on the top of this hump. Obviously, the hump position for the reduction with coke is before the first sampling, whereas for graphite it is before the second sampling, and for charcoal it is before or after the first sampling. In this article, the rapid part of the FeO reduction is called the main FeO reduction stage, and the MnO reduction after reaching the hump of MnO is called the main MnO reduction stage.

The reduction rate of MnO from a slag with a similar composition as the current synthetic slag and without FeO content by the same reductants was studied previously,[19] and based on their importance in the current discussions, the results are represented in Figure 5. Comparing Figures 4 and 5 reveals that the rate of MnO reduction from slag that contains FeO (from the hump point) by each carbon material is significantly higher than that from slag devoid of FeO. During the reduction of the latter slag, no metal phase was observed at the slag/carbon interface because of the rapid evaporation of the produced manganese.[15,19] When metal is not present, the charcoal has the greatest reactivity with the slag and graphite has the lowest reactivity (Figure 5). Whereas, when metal is present, coke is the most reactive carbon material (Figure 4). It may be concluded that the kinetics of MnO reduction is affected by the type of carbon material, and it can be extensively changed through a metal phase in the system.

Microscopic studies of the cross sections of the samples revealed that a metal phase is always present at the slag/carbon interface, as typically shown in Figure 6. The measured concentrations of Fe, Mn, C, and Si in the metal produced by different carbon materials are listed in Table II and are shown in Figure 7 by solid lines. The initial Fe content was assumed 100 pct, because pure iron is expected to be first nucleated. As observed, the metal phase after a 4-minute reduction at 1400 °C includes a Fe-Mn-C alloy that contains a low Si content. The Mn/Fe ratio in the metal produced by coke is always the highest and by graphite is the lowest. The measured carbon concentrations in the produced metal phase are between 6.9 and 8.5 pct, except for the metals produced by graphite in 4 minutes and by coke in 13 minutes.

Regarding the importance of carbon in the metal phase, the carbon solubility in metal produced by the carbonaceous materials was calculated by the formula suggested by Fenstad.[20] In this case, the measured Fe and Mn contents of the produced metal at various reduction times and the corresponding sample temperatures were considered. The calculated carbon solubilities in each alloy are listed and compared with the measured carbon concentrations in Table II. Moreover, these calculated carbon solubilities are illustrated on the carbon graph in Figure 7 by the dashed lines. Obviously, the measured carbon concentrations are higher than saturation, except for the metal produced by graphite in 4 minutes. This finding is expected with regard to the sample preparation for microprobe, where a thin layer of carbon was coated on the samples, which influences the resulting carbon concentrations higher than the actual carbon. It is worth noting that the reported measured carbon concentrations by EPMA in this article are around 10 random analyzed points of the metal drops, on average. For instance, the average carbon concentration in the metal produced through slag reduction by coke after 4 minutes is 6.7 wt pct with a standard deviation of 0.4 wt pct. The carbon graph in Figure 7 shows that the concentration of carbon is not significantly changed after exceeding around 6 wt pct, which indicates that the Fe-Mn alloys are approximately saturated above this level. Therefore, the metal is approximately carbon saturated in all coke and charcoal samples, except for the metal produced by graphite within a reduction time of 4 minutes. It is worth mentioning that the lower carbon concentration in the produced metal by graphite than by charcoal and coke in the first 4 minutes was confirmed by the microstructure studies. In particular, relatively lesser amounts of carbides were observed in the metal produced by graphite than that produced by charcoal, whereas they have similar concentrations of Mn. Although the measured carbon concentrations by WDS are not absolute and they always show higher concentrations than reality (because of the carbon coating), they can be used to indicate how much the metal produced by graphite in 4 minutes is under saturation. Considering the reduction by coke as an outlier, the measured carbon concentrations in the metal phase are on average 1.2 ± 0.5 wt pct higher than the calculated concentrations for the carbon saturation in the alloys (Table II). This may indicate that the real carbon concentration in the metal produced by graphite within 4 minutes can be in the range 3.45 to 3.95 wt pct, which is less than the expected saturation of 5 wt pct (Table II).

The difference in the carbon content of the metals produced in the first sampling indicates that the rate of iron production with coke and charcoal is small compared with the rate of carbon transfer/dissolution into the metal, so that the metal phase is always saturated by carbon. In contrast, the rate of metal production with graphite is relatively high compared with the rate of carbon dissolution/transfer in metal, and therefore, the carbon concentration of metal is below the saturation. Considering the carbon content differences and the rates of FeO reduction (Figure 4) with two carbon materials, we can say that the FeO reduction in the system is dependent on the carbon content of the metal phase, and FeO is mainly reduced by the dissolved carbon in the liquid metal. In this case, charcoal is a better source of carbon than graphite, because carbon dissolution/transfer in iron is not a rate-limiting step, although it can be rate limiting for the graphite.

The masses of the produced Fe and Mn through slag reduction were calculated using mass balance. In this case, the initial slag chemical composition, the initial slag mass, and the measured chemical compositions of the reduced slags (Table III) were used. Considering CaO and Al₂O₃ as the stable oxides in the slag phase, the mass of the reduced slags was calculated. The mass of the remained FeO and MnO was calculated with regard to the determined mass of the reduced slag and the measured slag composition. Consequently, the produced Fe and Mn for different reduction times were calculated. The obtained results (Figure 8) indicate that the Fe production is rapid at the beginning for all samples and that it levels off after 4 minutes at 1400 °C for coke and charcoal samples and later for the graphite sample (Figure 8(a)). In contrast, the manganese production starts slowly and simultaneously with the Fe production (solid curves in Figure 8(b)), and then it increases rapidly after the main FeO reduction stage. Figure 8(b) indicates that the amounts of the produced manganese after 4 minutes at 1400 °C for different reductants are close and in the range of 0.75 to 1.17 mg. However, the manganese concentration of the produced metal by different substrates is different, as illustrated in Figure 7. This indicates that the manganese evaporation occurs parallel to the manganese production (dotted curves in Figure 8(b)) and that it affects the total transferred manganese to the metal. Moreover, the Mn concentration in the metal levels off after awhile, whereas MnO is still reduced. This may indicate that the rate of Mn production and evaporation become similar after some reduction. Therefore, the manganese concentration in the metal does not vary, which is observed in particular for graphite and coke samples after 7 minutes in Figure 7.

Considering the Mn content of the metal phase in Figure 7 and the manganese production rate (Figure 8(b)), it is found that the concentration of Mn in the metal is mostly higher for the carbon samples with higher MnO reduction rates. The maximum Mn evaporation rate is about 0.5 mg/min for the coke sample. This evaporation rate is much lower than the Mn evaporation rate from the pure Mn at 1600 °C, which is about 5 mg/min as measured previously in Ar atmosphere by sessile drop method.[19] It is worth noting that when a similar FeO-free slag is reduced, no metallic Mn have been found at the slag/carbon interface, which indicates a faster Mn evaporation in Fe-free system.[19,21] This finding indicates that Fe plays an important role to stabilize the liquid state of the produced Mn with maintaining low Mn activity in the metal phase. This may show how Fe benefits the industrial process by decreasing the amount of Mn, which is evaporated in the coke bed area and condensed in the upper zones of the furnace.

Regarding the above discussions, a possible explanation of the reduction mechanism pertaining to a slag containing MnO and FeO as well as SiO₂ is given in the next section in an effort to combine general knowledge about the possible reactions in the system and the experimental observations.

It was observed in the sessile drop experiments that the slag droplet moves and dances during reduction. This phenomenon is caused by the CO gas bubble formation and growth in the slag and its bursting on the slag drop surface. The cross section of the reduced slag by the graphite substrate in Figure 10 shows that a gas bubble has been trapped during cooling, while it was passing through the slag. The formation and growth of the gas bubbles in the slag can indicate that the reduction reactions take place extensively by the dissolved carbon in the metal through reactions [3] and [6] at the slag/metal interface. If the reactions do not take place at this interface and they occur at slag/solid carbon interface, then no bubbling is expected, because in such situations, the released gas through the chemical reactions would prefer to leave the interface through the powder substrate and does not penetrate to the slag. It is worth noting that many small metal prills, even of submicron size, were observed in the solidified slag drop, and they may react with the slag as well. These metal prills have been transferred into the slag through the gas bubbles and the fluid flow in the slag droplet. Regarding the small size of these metal prills, it may be reasonable to assume it is a suspended metal phase in the slag. The role of iron in the mechanism of MnO reduction may also be confirmed with regard to the composition of the metal prills, which are distributed in the whole slag. It was observed that these metal prills are Fe-Mn-C alloys with much less carbon than the large metal particles at the slag/carbon interface. This finding may show that the transferred metal prills to the slag react with slag and decarburized.

Considering the obtained results and the discussions, the carbothermic reduction of MnO from FeO-containing slags can be explained. To simplify the system, the drawn schematic for the important reactions and the slag-reduction mechanism in Figure 11 can be considered. As mentioned previously, the reduction of FeO in the main FeO reduction stage can occur by both the solid and dissolved carbon in the produced metal. Regarding the correlation between the FeO reduction kinetics and the dissolved carbon concentration in the melt (Figures 4 and 7), and also the extensive gas bubble release from the slag surface, it may be reasonable to consider the reduction of FeO by the dissolved carbon as the dominant reaction. And as discussed previously, the kinetics of this reaction are controlled by the solid carbon transfer into the produced metal. In this stage, the reduction of MnO by Fe takes also place because of the rapid formation of iron containing high activity. In this stage, the rate of MnO reduction is dependent on the rate of the FeO consumption in the system. This means that the reduction of both FeO and MnO in the main FeO reduction stage is controlled by the carbon dissolution/transfer into the produced metal phase.

Two possible mechanisms can be studied for the MnO reduction in the main MnO reduction stage. The first mechanism is the MnO reduction by the solid carbon according to reaction [5] and without Fe contribution. If such a reaction takes place in the system, then the order of sequence of the MnO reduction rates by carbonaceous materials in Figure 4 has to be similar to the MnO reduction rates from FeO-free slag (Figure 5). But, as we can observe, the order of the MnO reduction rates from theses two type of slags is different. Moreover, reaction [5] leads to the formation of CO gas at the slag/carbon interface with no bubble bursting from the slag. However, gas bubble bursting is always observed from the slag drop top, which means that reaction [5] cannot be dominant in the system. As described in section E, this reaction involves four phases, and it is limited to a few points in the system. In contrast, if this reaction proceeds at slag/carbon contact area without Fe contribution, it leads to the formation of a metal rich of Mn with high manganese activity. However, some metal in the system that contains Mn with low activity (Table II) already exists, which maintains better thermodynamic conditions for the reaction. Thus, MnO reduction does not take place in the system at the slag/carbon interface without the contribution of the metal droplets that contain iron.

A possible mechanism for manganese oxide reduction in the main MnO reduction stage can be MnO reduction by Fe through reaction [9] and then the reduction of the produced FeO by the solid carbon. This can be confirmed with obtaining similar MnO and FeO reduction rate orders with regard to the carbonaceous materials in both reduction stages (Figure 4). It was also observed that in the main MnO reduction stage, the metal phase is almost saturated of carbon. Therefore, if the FeO reduction by the dissolved carbon in the metal is dominant, then the corresponding MnO reduction rates for different substrates must not be significantly different, which is not observed in the experimental results (Figure 4). Hence, the FeO reduction in the main MnO reduction stage may occur mainly by the solid carbon through reaction [2]. This reaction involves four phases (slag, metal, carbon, and gas), and it seems to be impossible in the system. However, this reaction is the first reaction in the system after interacting the slag with carbon, and it may be reasonable to consider it as a proceeding reaction with an unknown mechanism in the main MnO reduction stage. This result can indicate that the kinetics of FeO reduction, and consequently the MnO reduction, in the second stage are dependent on the reactivity of the solid carbon material. Subsequent work would be beneficial to increase our knowledge about the mechanism details and to clarify the remaining uncertanties.