3.1. Removal of General Elements and the Selective Reduction of Oxides in Direct Reduction Process
The reduction of oxidized pellets by hydrogen can effectively decrease sulfur and carbon in iron ore.
Figure 2 shows the changes in carbon and sulfur content in each step, in which “ppmw” means parts per million by weight. As can be seen from
Figure 2, the sulfur content reduces during the roasting of pellets.
The sulfur in iron ore usually exists in the form of sulfide, which can be removed by the oxidation reaction during roasting, as shown in Equations (1) and (2). The desulfurization rate of pellets during roasting can reach 96–98%, while the remaining sulfur is further reduced in the refining step.
The decrease of carbon content occurs mainly in the reduction step of pellets, which is the basic law of iron oxide reduction, as shown in Equation (3). In addition, the residual FeO in DRI can continue to react with carbon in the melting separation step.
The efficient and simultaneous removal of carbon and sulfur is difficult to achieve by coal-based direct reduction or gas-based reduction using a CO–H2 mixture. When the direct reduction iron contains a large amount of carbon, it is laborious to remove it completely in the subsequent smelting process. Although the reaction between carbon and FeO in the melting process is considered, it is almost impossible to reduce carbon to a very low level in the end. For coal-based direct reduction, the main source of sulfur is solid reducing agent, which is easy to enter the iron in reduction process, and it brings additional desulfurization burdens for subsequent refining.
Because of the reduction characteristics of hydrogen, the reduction process is selective, that is, the valuable element Fe can be extracted from the iron ore, while most of the valueless gangue minerals cannot be reduced into metal. The selective reduction of oxides by H
2 can be explained by
Figure 3, which shows the standard Gibbs free energy change for the reactions of hydrogen with various oxides that may exist in ores [
12]. At the temperature where direct reduction occurs, i.e., 1073–1273 K, the general oxides such as manganese oxide, silicon oxide and alumina cannot be reduced into iron, while only phosphorus oxide, molybdenum oxide, nickel oxide and copper oxide can be completely reduced. The removal of phosphorus can be accomplished in subsequent steps, while the reduced copper, nickel and molybdenum are difficult to eliminate. Fortunately, the content of these oxides is relatively low in common iron ores. The blast furnace ironmaking uses coke as a reducing agent, and the temperature of blast furnace ironmaking is about 1773 K. This leads to the partial reduction of manganese oxide, silicon oxide, etc. into the molten iron, which imposes a burden on the refining process. This is why the hot metal of a blast furnace is not as pure as DRI.
3.2. Dephosphorization in the Melting Separation Process
According to the analysis in
Section 3.1, all the phosphorus in the ore is reduced after direct reduction. If the DRI is simply melted and separated at 1823 K without slagging for dephosphorization, the phosphorus content in the obtained pure iron is at 166 ppm, and if the reduced phosphorus is removed by slagging refining during the melting process, the phosphorus content drops to 18 ppm, as shown in
Figure 4.
The dephosphorization reaction in the refining process is carried out between slag and metal. Previous studies by other researchers and us [
13,
14,
15,
16,
17,
18,
19] have clearly investigated the mechanism of dephosphorization between slag and metal. Firstly, phosphorus in iron is oxidized to (P
2O
5) with the conditions of an oxidizing atmosphere or an oxidizing slag (FeO), as shown in Equation (4). Secondly, free CaO in high basicity slag reacts with (P
2O
5) to form (3CaO·P
2O
5), as shown in Equation (5). The formation of (3CaO·P
2O
5) decreases the activity of (P
2O
5), and this reaction is so strong that Equation (4) is promoted. Accordingly, the comprehensive dephosphorization reaction equation can be considered to be carried out by Equation (6).
The phosphorus distribution ratio between slag and metal can be expressed as
. Basicity and FeO play an important role in dephosphorization. According to previous studies, increasing the basicity and FeO content of slag can improve the phosphorus distribution ratio, as shown in
Figure 5 [
19].
It is qualitatively concluded that the favorable conditions for dephosphorization are low temperature, high basicity of slag and oxidizing condition. Due to the spongy porous structure, DRI is easily oxidized during transportation and storage. Alternatively, in the actual reduction process, a certain amount of FeO can be preserved in the DRI by controlling the reduction conditions. The presence of FeO in DRI creates an innate advantage for dephosphorization, which means that no additional oxygen is needed to obtain FeO during refining. At this time, we only have to get a high basicity slag to effectively dephosphorize.
In order to quantitatively study the effect of slag composition on dephosphorization in the melting separation/dephosphorization process and further determine the suitable slag composition for dephosphorization, the thermodynamic calculation and experiments were carried out based on the designed 15 groups of slag with different CaO and FeO contents. The thermodynamic model for predicting the phosphorus distribution ratio in the melting separation process was developed based on the ion-molecule coexistence theory (IMCT) of slag [
18,
19]. The detailed calculation is not described repeatedly here. The composition of experimental slag used for dephosphorization is shown in
Table 3.
In this experiment, the FeO in slag does not need to be additionally added, while the different FeO content contained in DRI can be controlled by changing the reducing conditions in the direct reduction step. Al2O3 is added in the slag to reduce the melting point of slag for better meltability, while MgO is added to avoid the crucible erosion. The SiO2 content in the DRI is relatively constant, about 9%, so other components are added proportionally according to the slag composition, which results in different total slag mass wslag for different experimental slags. After the experiment, the phosphorus content in different pure iron samples was determined.
The effect of basicity (%CaO)/(%SiO
2) and FeO content on dephosphorization is shown in
Figure 6. It can be seen from the figure that increasing the basicity can greatly promote dephosphorization, but when the basicity is more than 3, the increase of basicity has a limited effect on promoting dephosphorization. Increasing the FeO content also has a significant effect on promoting dephosphorization, but the ideal dephosphorization effect can be achieved when the FeO content reaches 5%. Excessively high FeO content is not necessary, especially when the basicity is high. Furthermore, due to the SiO
2 content in DRI being high and relatively constant, to obtain a high basicity slag, it requires much CaO to be added, which results in a quantity of slag (see
Table 3) and a lot of energy consumption in the production. The FeO content also determines the yield of iron after the dephosphorization step, and too high of an FeO content leads to an increase in iron loss.
Consequently, based on the above results and analysis, controlling the basicity at 4 and the FeO content at 5% (Slag 9) can effectively dephosphorize as well as make the yield of iron and the energy consumption within an acceptable range. In actual production, ores with low SiO2 should be used for production convenience and lower energy consumption. On the other hand, the necessity for the dephosphorization operation can be considered depending on the phosphorus content in ore and the purity of pure iron required.
3.3. Deoxidation by Secondary Refining
In metallurgy, the commonly used methods of deoxidation are precipitation deoxidation, diffusion deoxidation and vacuum deoxidation. For conventional technically pure iron, deoxidation is carried out by adding aluminum as a deoxidizer, and then the generated inclusions are removed by slagging. This results in a generally higher aluminum content in technically pure iron (see
Table 1), which reduces the purity of the iron. The gas impurities such as oxygen and nitrogen in electrolytic iron are sometimes not used as an evaluation criterion. If necessary, they are usually removed by adding a small amount of carbon under the ultra-high vacuum [
3]. Deoxidation for pure iron is a difficult problem because elements commonly used for deoxidation, such as carbon, silicon and aluminum, are themselves impurities for pure iron. Owing to the presence of chemical equilibrium, it is quite challenging to control the oxygen and deoxidizing elements simultaneously to a very low level.
In this experiment, the primary pure iron after melting separation/dephosphorization had an oxygen content of 300 mass ppm, and after the step of refining/deoxidizing, the amount of oxygen decreased to 10 mass ppm. Oxygen in the primary pure iron is removed by the reaction between slag and molten iron, which is to say the inclusions in iron are eliminated by the transportation into slag.
As the main component of the gangue in ores, SiO
2 remains in the DRI and coexists with the reduced metallic iron particles, as shown in
Figure 7a,b.
These SiO
2 particles gather and float up to become slag during the melting process, while the iron that has a greater density becomes molten iron and separates from the slag. Research suggests that the floating velocity of particles in molten iron is proportional to the square of the radius of particles, as shown by the Stokes Equation and related correction formula [
20]. For small liquid deoxidized products, taking into account the particle viscosity
, the rise velocity should be given by Equation (7).
In this study, the viscosity of SiO
2 particles is much higher than that of molten iron, so the viscosity of molten iron in the formula
can be ignored, and Equation (7) is simplified as Equation (8).
In the above formulas, is the rise velocity of SiO2 inclusions in molten iron; is the inclusion’s radius; and are the density of the metal and of the inclusion, respectively, with values of and ; is the gravity constant, with a value of ; and is the viscosity of molten iron, with a value of 0.006 at 1823 K.
According to the experimental conditions, if there is about 100 g of molten iron in the crucible, the depth of the molten iron is about 1.8 cm. From this, the time taken for SiO
2 particles of different sizes in molten iron to float up can be calculated, as shown in
Figure 8.
It can be seen from
Figure 8 that under current experimental conditions, the duration of the melting separation process is about 20–30 min, during which not all of the SiO
2 particles can float up into slag. Only SiO
2 particles larger than 5 μm can rise into slag, and there are still many small-sized SiO
2 particles left in the iron, as shown in
Figure 7c,d. These SiO
2 particles may prove to be the main source of oxygen in the primary pure iron. At this time, the task is to make suitable slag to promote the residual SiO
2 inclusions to float and be absorbed by the slag, so as to achieve the purpose of deoxidation.
After a large number of experiments, the appropriate slag for deoxidation was determined. The composition and characteristics are shown in
Table 4. The completely melting temperature and viscosity of slag were calculated using FactSage
TM 7.3 software. In fact, the residual SiO
2 inclusions are equivalent to the products of the reaction of silicon and oxygen, and the driving force for their removal is highly dependent on the properties of the slag. Valdez et al. [
21] and Park et al. [
22] suggested that the dissolution of the inclusion into the slag is controlled by the slag phase mass transfer and that the total dissolution time of the inclusion into the slag (
) is given by Equation (9). This result qualitatively implies that the driving force of the dissolution of the inclusion and the viscosity of the slag directly affect the removal rate of inclusion by the slag.
where
is the particle density,
R is radius of inclusion,
C is the driving force for the dissolution (concentration difference),
k is the Boltzmann constant,
T is the temperature,
is the ionic diameter, and
is the viscosity of the slag.
The result of the automated SEM/EDS inclusion analysis (ASPEX) is shown in
Table 5, which shows that these inclusions are almost entirely SiO
2, so a high basicity slag is necessary to provide sufficient chemical driving force. Previous studies by Ren et al. [
23] and us have demonstrated the effectiveness of high basicity slag in removing SiO
2 inclusions. According to the research on the ion-molecule coexistence theory (IMCT) for slag [
24,
25], the free CaO in slag has a strong binding ability toward SiO
2, as shown in Equations (10)–(12). Consequently, an increase in the basicity of slag causes an increase in the chemical driving force for the adsorption of SiO
2 inclusions by the slag.
The excessive CaO affects the melting point and fluidity of slag, so it is necessary to add Al2O3 to lower the melting point of slag, and CaF2 has a great effect on reducing the viscosity of slag, which can greatly promote the adsorption of inclusions by slag, as Valdez studied. Actually, the effect of the properties such as viscosity and melting point on the adsorption capacity of slag can be considered as a physical driving force. Under the combined action of chemical driving force and physical driving force, the SiO2 inclusions in the primary pure iron are effectively removed.
3.4. Feasibility of Industrialization and Simple Estimation of Cost
The process of producing high-purity iron is shown in
Figure 9. Iron ore fines are transformed into oxidized pellets by a disc pelletizer with pelletizing and roasting functions. The direct reduction step is carried out in a shaft furnace using hydrogen as a reducing agent to produce sponge iron. Although most of the current direct reduction processes use a CO–H
2 mixture from natural gas pyrolysis as reducing agent, hydrogen metallurgy without carbon emissions is gradually developing. Some enterprises, such as the HYBRIT project in Sweden, have carried out attempts to use pure hydrogen in metallurgy [
26]. Hydrogen metallurgy is the future development direction, with the goal of reducing CO
2 emission and developing cleaner industry. The melting process of DRI can be carried out using a mid-frequency induction furnace or an electric arc furnace, and the secondary refining can be carried out in an electric arc furnace or an LF furnace. These devices are conventional metallurgical equipment and have very mature experience in use.
The cost of high-purity iron produced by this process is shown in
Table 6. It is a rough estimate, based on a project design we have done previously. In fact, the data in
Table 6 are a magnified result, while the actual cost should be lower. The iron ore grade is calculated as 60% TFe. If one tonne of iron is produced, 1.667 tonnes of ore is needed. The market price of this kind of iron ore is
$100 per tonne, so the ore cost is about
$166.7 per tonne of iron. The price of hydrogen obtained by different processes varies greatly. Considering that more and more attention has been paid to the future use of hydrogen metallurgy, the cost of hydrogen used in industry will be lower and lower. According to the current market, the price of high-purity hydrogen with 99.999% purity is 0.7 USD/Nm
3. It takes about 200 Nm
3 hydrogen to produce one tonne of iron, so the total cost of hydrogen is
$140 per tonne of iron. Owing to a large amount of slag with high basicity needed in the dephosphorization and melting process, about 0.6 tonnes of fluxes, such as lime and dolomite, are needed for smelting one tonne of pure iron. The price of flux is about
$110 per tonne. Furthermore, 1.5 tonnes of water (recyclable), 20 Nm
3 of natural gas (for a small amount of energy supply) and 1200 kwh of electricity are required to produce each tonne of iron. The price of energy for industrial use in China is about
$0.7 per tonne of water,
$0.6 per Nm
3 of natural gas and
$0.1 per kwh of electricity. Based on a plant with an annual output of 50,000 tonnes of pure iron, if there are 100 workers, and the annual salary of each worker is
$20,000, then the labor cost is
$40 per tonne of iron. Additionally, the annual cost of maintenance, overhaul and loss is about
$1.5 million US dollars, so the average cost of maintenance is about
$30 per tonne of iron.
It can be seen that the cost is at the same level as that of ordinary steel, which makes it possible to produce high-purity iron on a large scale. According to the investigation, the market price of pure iron with 99.9% purity is about $1000 US dollars/tonne, and the price of pure iron with 99.95% purity has climbed to $4000 US dollars/tonne, while the price of ultra-high-purity iron with 99.99% or higher purity reaches $7000–200,000 US dollars/tonne. Combined with the cost estimation, it can be seen that there will be huge economic benefits.
At present, the production of DRI in the world has grown rapidly, reaching 100.5 million tonnes per year in 2018. Almost all of this DRI is used as a substitute for scrap steel in the production of steel products, which is a waste of the purity of DRI. The high-purity iron produced by this approach can be used as the high-quality raw material for smelting various steels and iron-containing alloys. It only requires the addition of corresponding alloying elements to meet the requirements of products without additional purification and upgrading. Therefore, this high-purity iron and its production method will have broad application prospects for future steel manufacturing processes.