Introduction

Steel slag is classified as basic oxygen furnace (BOF) slag, electric arc furnace (EAF) slag, and ladle metallurgy slag. The EU produced around 20 million tons steel slag annually in the recent years, of which 46 wt pct is BOF slag.[1] Recycling and reutilization of such large amounts of steel slag are of great importance, both for the sustainability of the metallurgical industry and for the environment. After modification of its chemistry and solidification mineralogy by means of hot stage engineering, steel slag can be recycled for internal use in the steelmaking process, and utilized for road construction, for cementitious substitutes, and as agricultural fertilizer.[2,3,4,5,6,7,8]

Using the slag for construction or agricultural purposes, however, does not consider the large amount of Fe present in the slag. Typically, BOF slags contain from 14 to 29 wt pct of Fe, present in iron oxides and iron-containing minerals.[9] “Zero waste” of BOF slag can be achieved by both metal recovery and slag utilization. However, the difficulty of controlling the P distribution between slag and the recovered metal prevents the slag from achieving a full reutilization. Because of P, the recovered Fe has a limited application. Also, P stabilizes β dicalcium silicate (β-C2S). The complete removal of P from BOF slag leads to disintegration of the slag due to the transformation of β-C2S to γ dicalcium silicate (γ-C2S) during cooling.[10] Thus, it is favorable to concentrate P in the oxides while limiting its concentration in metallic Fe. In previous studies, the carbothermic reduction of slag has been described to recover Cr and/or Fe from stainless steel slag, hot metal dephosphorization slag, and other steelmaking slags,[11,12,13,14] without focusing on improving the purity and controlling the morphology of the metal. Furthermore, the utilization of the remaining slag by modifying the chemical composition did not receive as much attention as the metal. Ye et al. studied the reduction of steel slags to recover both metals (Fe, Mn, V, and Cr) and oxide materials. The oxides could be reused as cementitious materials and/or desulfurization fluxes in secondary metallurgy, but no chemical modification of the oxides was considered.[15] To utilize the steel slag as aggregates in road construction, Yang et al. studied the chemical modification of the slag after reduction, focusing on avoiding the disintegration of slag caused by the transformation of β-C2S to γ-C2S.[16] Kim et al. proposed a two-stage reduction of EAF slag followed by water quenching, with the aim to reuse the amorphous slag in the cement preparation.[17] However, the concurrent implementation of reduction to achieve high value-added metal products, and chemical improvement of the remaining oxides for application in cements and geopolymers, has not been described.

In this study, the carbothermic reductions of Fe oxides and P-containing compound and/or oxides of BOF slags with different amounts of Al2O3 and SiO2 additions were investigated. The effect of Al2O3 on the size distribution of the reduced Fe was discussed. Based on experimental observations and thermodynamic calculations, a potential method to recover low-P-containing Fe particles with a reasonably homogeneous size distribution is suggested. Concurrently, the phase modification of residual BOF slag with the aim to produce slag product with enhanced cementitious property was studied. A potential method to achieve the “zero-waste” concept of the BOF slag, contributing to the sustainability of the steelmaking industry, is proposed as well.

Experimental Methods and Materials

Slag Preparation

Industrial BOF slag samples were collected at different positions in a slag yard to obtain a representative chemical composition of the slag. The samples were ground, followed by milling to less than 200 µm. Thereafter, the powders were thoroughly mixed for homogenization, and then applied in this study as the master slag material. Table I shows the chemical composition of this BOF slag, as determined by X-ray fluorescence spectroscopy (Panalytical PW2400). Fe2+ and Fe3+ were measured using chemical titration by potassium dichromate. 10.2 wt pct Fe2+ and 10.4 wt pct Fe3+ were determined for the starting master slag.

Table I Chemical Composition of the Master Slag (Wt Pct)

Due to the presence of different iron oxides (FeO, Fe3O4, and Fe2O3) in the BOF slag, a preliminary experiment was performed. An appropriate range of C addition for reducing the oxides was determined to be from 5 to 8 wt pct. Thus, in the present study, to study the effect of carbon on the reduction behaviors of iron- and phosphorus-containing compounds, respectively, 5, 6, 7, and 8 wt pct C (Superior Graphite, 6.5 µm) were added to the master slag in the form of powder. To reveal the effects of SiO2 and Al2O3 additions on phase modification of the slag and metal formation in the reduction experiments, different amounts of SiO2 (Sibelco, 1 to 40 µm) and Al2O3 (Sasol, 25 µm) were mixed with the master slag, while keeping the C addition amount fixed at 7 wt pct. Table II shows the various additions of SiO2, Al2O3, and C to the master slags. Each mixture was wet-mixed using ethanol in a multidirectional mixer (Turbula type) for 24 hours. The mixture was then dried by a rotating evaporator at 338 K (65 °C) and further dried at 353 K (80 °C) for 24 hours.

Table II Different Amounts of Additions for Preparing the Experimental Slags

Experimental Procedure and Characterization

Each slag mixture (10 g) was loaded in a high-purity magnesia crucible (21 mm ID, 50 mm H), which was suspended by Mo hooks in a vertical tube furnace (100-250/18, HTRV, GERO) under an Ar flow rate of 0.4 L/min. The samples were introduced at room temperature and held in the furnace at 1873 K (1600 °C) for 1 hour, followed by water quenching.

The microstructure of the slags was quantitatively analyzed using field-emission electron probe micro analysis (FE-EPMA, JXA-8530F, JEOL Ltd.) at fixed accelerating voltage (15 kV) and beam current (15 nA). For the wavelength-dispersive spectroscopy (WDS) analysis of the light element C, the analyzing crystal adopted is a layered diffracting element 1 at K-α line. Phase identification was achieved via X-ray diffraction (XRD, 3003-TT, Seifert, Ahrensburg), with 2θ in the range from 10 to 80 deg using Cu Kα radiation at 40 kV and 40 mA.

Results and Discussion

Effects of C and Al2O3 on Metal Recovery

Figure 1 shows the Fe and P elemental distributions after reduction by different amounts of C additions. With the additions of 5 and 6 wt pct C, most of the Fe is concentrated in the spherical metallic phase. Some Fe is found in other phases, mainly in magnesia wustite (RO). This phase is composed of monoxides, such as FeO, MnO, and MgO. With the increasing C addition, Fe eventually concentrates entirely in the metallic phase, as the higher C additions create stronger reduction conditions. In addition, the increasing C additions also favor the transfer of P from slag to metal. As indicated by the color level, with 5 and 6 wt pct C additions, very low concentrations of P can be seen in the metallic phase. At 8 wt pct C addition, the concentration of P in the metallic phase is much higher. Furthermore, free lime and periclase precipitate during the reduction due to the high basicity of the slags. It is known that the hydration of free lime and periclase induces about 10 vol pct expansion,[18] which should be avoided for construction applications.

Fig. 1
figure 1

Fe and P distributions at various C additions

In order to acquire accurate data on the reduction behavior of BOF slags, more than 10 metallic grains were analyzed using WDS. The dissolved P content in the metallic Fe is given in Figure 2 (solid square points). With 5 and 6 wt pct C additions, the P content is less than 0.01 wt pct. With 8 wt pct C addition, the P content increases to more than 2 wt pct. Clearly, the reduction of P-containing compounds and the dissolution of P in Fe are closely related to the amount of C addition, i.e., the higher the carbon addition, the higher the concentration of phosphorus in the extracted Fe. At 5 and 6 wt pct C additions, the low P contents suggest that P-containing compounds cannot be reduced or volatilized after reduction. Morita et al.[11] have carried out carbothermic reduction experiments of steelmaking slag at over 1873 K (1600 °C) using microwave processing, and proved that Fe x O is more easily reduced than P oxides. Maruoka et al.[19] studied the distribution ratio L p of P between metallic Fe and slag after the reduction of iron ore by CO, and observed that P is reduced and dissolves in the metallic phase when the Fe x O content in the slag is less than 10 wt pct.

Fig. 2
figure 2

P content in the metallic Fe and P distribution ratio L p

Most P in BOF slag is incorporated with Ca2SiO4 (C2S) to form a C2SiO4-Ca3(PO4)2 (C2S-C3P) solid solution.[7] The effect of basicity on the stable forms of complex phosphate ions has been studied by Cho et al., and [\( {\text{PO}}_{4}^{3 - } \)] was concluded as the stable form in highly basic BOF slag.[20] The structure of C2S-C3P is similar to that of pure C2S, with part of tetrahedral [\( {\text{SiO}}_{4}^{4 - } \)] groups substituted by tetrahedral [\( {\text{PO}}_{4}^{3 - } \)] groups. To maintain the charge balance, the substitution combines with less Ca2+ than that in the pure C2S.[21] During the reduction of the C2S-C3P solid solution, the P-containing compounds react with C and/or CO to break the P-O chemical bonds. Thereafter, the released free Ca ions can combine with silicate and form Ca3SiO4 (tricalcium silicate, C3S). Therefore, with the reduction of P from the BOF slag, the C3S content is expected to increase. The reduction of P-rich compounds and the formation of C3S can be written as reactions (1-1) through (1-4):

$$ 5{\text{C}} + 3{\text{Ca}}_{2} {\text{SiO}}_{4} + {\text{Ca}}_{3} \left( {{\text{PO}}_{4} } \right)_{2} = 3{\text{Ca}}_{3} {\text{SiO}}_{5} + 5 {\text{CO(g)}} + 2[{\text{P}}], $$
(1-1)
$$ 5{\text{C}} + 3{\text{Ca}}_{2} {\text{SiO}}_{4} + {\text{Ca}}_{3} \left( {{\text{PO}}_{4} } \right)_{2} = 3{\text{Ca}}_{3} {\text{SiO}}_{5} + 5{\text{CO(g)}} + {\text{P}}_{2} ({\text{g}}), $$
(1-2)
$$ 5{\text{CO}} + 3{\text{Ca}}_{2} {\text{SiO}}_{4} + {\text{Ca}}_{3} \left( {{\text{PO}}_{4} } \right)_{2} = 3{\text{Ca}}_{3} {\text{SiO}}_{5} + 5{\text{CO}}_{2} ({\text{g}}) + 2[{\text{P}}], $$
(1-3)
$$ 5{\text{CO}} + 3{\text{Ca}}_{2} {\text{SiO}}_{4} + {\text{Ca}}_{3} \left( {{\text{PO}}_{4} } \right)_{2} = 3{\text{Ca}}_{3} {\text{SiO}}_{5} + 5{\text{CO}}_{2} ({\text{g}}) + {\text{P}}_{2} ({\text{g}}). $$
(1-4)

The P distribution ratio L p is defined as

$$ L_{\text{p}} = \frac{{({\text{pct}}\;{\text{P}})}}{{[{\text{pct}}\;{\text{P]}}}}, $$
(2)

where (P) and [P] represent the wt pct of P, respectively in slag and metal. Lee and Fruehan[22] suggested that L p can be estimated as

$$ \log \;L_{\text{P}} = {-} 12.24 + \frac{2000}{T} + 2.5\log ({\text{pct}}\;{\text{FeO}}) + 6.65\log \;B^{*} + 0.13[{\text{pct}}\;{\text{C}}], $$
(3)

where T is the absolute temperature (K); (pct FeO) and [pct C] are the wt pct of FeO in the liquid slag and C in the metallic Fe, respectively; and B* is the “weighted basicity” defined as B* = [(pct CaO) + 0.8(pct MgO)]/[pct SiO2 + (pct Al2O3) + 0.8(pct P2O5)]. [pct C] was measured by WDS, and B* was calculated according to the slag composition, and (pct FeO) was calculated using FactSage 7.0. For the calculation, all the iron is assumed to be hematite (Fe2O3). FactPS and FToxid databases were applied. Then, the P distribution ratio can be obtained. T, [pct C], (pct FeO), B*, and the calculated values of L p with Eq. [3] are listed in Table III. Log L p is given as a function of carbon addition in Figure 2 denoted by the solid triangles.

Table III Calculated P Distribution Ratio L p

By considering the mass balance of P and assuming that no P is evaporated during the reduction, the P content in the metal can be calculated, as shown in the Figure 2 (solid circle points). The P content in the metallic Fe matches well with the experimental observation where the P-rich phase changes from the slag (mainly C2S-C3P) to metallic Fe with the increasing C addition (Figure 1). With the increasing C addition, the measured and calculated P contents in the metallic Fe increase considerably, but the measured increase is less than the calculated one, and this trend becomes more significant with higher amount of carbon addition. The reason is probably that some P evaporates after the reduction.

The consumption of C by the reduction of iron oxides can be estimated via the reaction (4):

$$ 3{\text{C}} + {\text{Fe}}_{2} {\text{O}}_{3} = 3{\text{CO}} + 2{\text{Fe}} . $$
(4)

To simplify the calculation, all iron oxides are assumed to be hematite (Fe2O3). According to the reaction (4), the molar ratio of C to Fe2O3 is 3 in order to completely reduce the Fe2O3. In the present experiments, 5, 6, 7, and 8 wt pct C were used based on the Fe2O3 content in the master slag. In addition to the reduction of iron oxides, the C can also reduce other oxides, such as P-containing compounds and Mn-containing compounds. For the current slag, under the condition of 6 wt pct C addition, the P content in the metallic Fe is less than 0.01 wt pct and the purity of the extracted Fe is more than 98 wt pct, implying a promising metallic product. FactSage calculation suggests that Fe3P and Fe3C precipitate at higher C additions (8 wt pct). Fe3P increases upon increasing the temperature (precipitating temperature is 1713 K (1440 °C)), but Fe3C decreases with the increasing temperature (disappearing temperature is 1833 K (1560 °C)). To estimate Fe reduction ratio, mass balance calculation based on the measured values was carried out by considering following assumptions: Fe is in metal, RO (5 wt pct C addition) and periclase phases (6 to 8 wt pct C additions); all MgO is in the periclase phase; Mn is in metallic and RO phases. Also, the Fe reduction ratio can be assessed by FactSage calculation. Figure 3 shows the measured and calculated (by FactSage) Fe recovery values. Upon increasing the C additions from 5 to 7 wt pct, the Fe reduction ratio increases rapidly, but increasing C additions further has little effect. Under 5 and 6 wt pct C additions, in addition to the metallic Fe and the periclases, Fe elements also distribute in free lime, which could explain the overestimation of the measured Fe reduction ratio.

Fig. 3
figure 3

Measured and calculated Fe reduction ratios with different C additions

Therefore, to minimize the P content in the metallic Fe, while maximizing the Fe reduction ratio from the present BOF slag at 1873 K, the optimized molar ratio of carbon to iron oxides is suggested to be 3.

Figure 4 shows the morphologic evolution of Fe particles reduced by 7 wt pct C, respectively, with 0, 5, and 10 wt pct Al2O3 additions. With the increasing Al2O3 addition, the size of the particles changes significantly. The particles which attach to the amorphous phase are larger (typical diameter more than 10 µm) than the ones which are located in the solid solutions (typical diameter less than 10 µm). The crystals in the amorphous were probably precipitated during water quenching.

Fig. 4
figure 4

Morphology of metallic Fe reduced by 7 wt pct C and (a) 0 wt pct Al2O3; (b) 5 wt pct Al2O3; (c) 10 wt pct Al2O3 additions

In order to understand the formation and growth mechanism, the Fe particles in each case were processed by ImageJ, with the aid of which the two-dimensional size distribution can be obtained. The particles with a size less than 1 µm are neglected to avoid errors due to the limitation of the software. In order to eliminate the arbitrariness caused by the size of bins defined manually, the population density function (PDF)[23,24] is adopted. The frequency in PDF is defined as the normal frequency divided by the bin width and has length−4 units. The functional PDF is applied to compare the size distributions between different samples (Figure 5). In each case, the Fe particle size varies greatly, and most of the Fe particles are less than 10 µm. The population density decreases markedly with the size increment. The largest particle size is increased from 16.5 to 78.5 and 127.5 µm with the increasing Al2O3 addition from 0 to 5 and 10 wt pct. Furthermore, the population density of smaller-sized particles (diameter less than 10 µm) decreases with the increasing Al2O3 addition, indicating that the number of smaller particles is decreased upon Al2O3 addition. The large range of Fe particle size under higher Al2O3 addition implies a more inhomogeneous size distribution that was induced by Al2O3 addition.

Fig. 5
figure 5

Population density functions of Fe particles under 0, 5, and 10 wt pct Al2O3 additions. Largest particle size is from 16.5 to 78.5 and 127.5 µm under 0, 5, and 10 wt pct Al2O3 additions, respectively

The particles were divided into two classes: one less than 10 µm and the other more than 10 µm. Figure 6 shows the average diameter of the Fe particles and the liquid fraction with respect to different Al2O3 additions. With size less than 10 µm, no significant size difference can be detected. With size being more than 10 µm, however, the size varies significantly with Al2O3 additions. Once a stable nucleus of a Fe particle is formed, it grows by diffusion of growth units of Fe atoms (reduced by carbon) from the bulk of the slag to the surface of the Fe particle (volume diffusion), followed by incorporation of the Fe atoms into the units (surface kinetics). Most of the metallic Fe is believed to be reduced from the Fe-containing phase RO (FeO-based solid solution) and the C2AF (brownmillerite) phases. As shown in Figure 7(a), two kinds of RO microstructures in the master bulk slag are found: the smaller-sized RO dispersed in the free lime and the larger-sized RO separated from other phases.

Fig. 6
figure 6

Average diameter of metallic Fe and liquid fraction with 0, 5, and 10 wt pct Al2O3 additions under 7 wt pct C: (a) particles less than 10 µm; (b) particles more than 10 µm. The error bars represent the standard error of the mean

Fig. 7
figure 7

(a) Microstructures of the master bulk slag; (b) Microstructure of the reduced slag with 10 wt pct Al2O3 addition

The Fe particles less than 10 µm (circles in Figure 7(b)) are considered to be formed from the dispersed RO in the lime (circles in Figure 7(a)). Their average size is independent of Al2O3 additions and is quite homogeneous (see the standard deviation indicated by the error bar in Figure 6(a)). FactSage calculation suggests that at 1873 K (1600 °C), the liquid fractions under 7 wt pct C addition are 33.5, 62.6, and 77.5 pct with 0, 5, and 10 wt pct Al2O3 additions (Figure 6). Thus, the Fe particles grow in the mixture of liquid and solids. As the Fe particles are always located in the solid-solution phase, their growth is probably controlled by solid diffusion. Due to the small diffusion coefficient of Fe in the slag (10−10 to 10−11 m2/s),[25] diffusion-controlled growth is limited to an extreme small scale. Therefore, the size of the smaller-sized Fe particles (<10 µm) is comparable with the dispersed RO in the master slag.

The larger Fe particles (>10 µm) are believed to be formed from the larger-sized RO and C2AF phases. After Fe has been extracted, the melting temperature of the Fe-deficient slag is significantly increased. Figure 4 indicates that Al2O3 addition enlarges the liquid fraction in the slag, resulting in more amorphous phase. Due to the fact that the larger Fe particles are located in the amorphous area which corresponds to the liquid fraction at the experimental temperature, the growth of Fe particles is probably enhanced by the liquid slag at high temperature. The liquid fraction at 1873 K (1600 °C) was calculated by FactSage 7.0 (Figure 6). The increase of the average Fe diameter shows a similar trend as the increase of the liquid fraction of the slag at 1873 K (1600 °C), implying that the particles sizes are indeed increased by the liquid slag. There are two possible reasons for the liquid-induced growth: more Fe particle collisions can be caused by the flow of the liquid slag, and the diffusion and convection of Fe are improved in the liquid phase. The presence of the large and inhomogeneously distributed Fe particles caused by external Al2O3 additions (as given in Figure 5) is therefore attributed to the local inhomogeneous distribution of the liquid slag in the mixture at high temperature. Although a more quantitative understanding of Fe particle growth is needed for the commercial fine Fe preparation, based on the above findings, however, it can be concluded that the Fe particle purity, size, and size distribution can be manipulated through the process control with carbothermic reduction parameters, such as controlling the carbon and slag modifier additions and temperature.

Effect of Al2O3 on Solidification Microstructure During Reduction

To optimize the microstructure of the slag after Fe extraction, the effects and Al2O3 on slag microstructure and mineralogy were investigated. Figure 8 shows the XRD patterns for the tested samples under fixed 7 wt pct C, with 0, 5, and 10 wt pct Al2O3 additions. The main crystalline phases are found to be metallic Fe, C2S, and C3S phases. The JCPDS 37-1497 database[26] was employed for free lime identification. The theoretical interplanar distances of 2.4059 and 1.7008 Å correspond, respectively, to 2θ at 37.34 and 53.86 deg using Cu Kα radiation. It can be seen that at the 2θ positions of 37.62 and 54.40 deg, which we consider to be the characteristic positions of lime for our samples; the slags with 0 and 5 wt pct Al2O3 additions present a peak, while the one with 10 wt pct Al2O3 addition shows no peak. The free lime can be completely removed by more Al2O3 additions. The shift of 2θ degree in our sample is believed to be caused by the incorporation of Fe2+, Mn2+, and Mg2+, diameters of which are smaller than Ca2+. Thereafter, the practical interplanar distance is slightly decreased, so that the corresponding 2θ degree is shifted to a bigger 2θ degree. The intense decrease of Fe peaks indicate that the quantity of metallic Fe is decreased with Al2O3 addition. The possible reason is that Fe particle size is increased by adding Al2O3.

Fig. 8
figure 8

Comparison of XRD patterns for samples with 0, 5, and 10 wt pct Al2O3 additions, respectively, and 7 wt pct C addition

Figure 9 shows the effect of Al2O3 additions on the microstructural modification of BOF slag reduced by 7 wt pct C. The composition of each phase was measured by WDS. Through linking each phase detected by XRD, its mineralogy can be identified. It is evident that most of the Fe particles are spherical. C3S typically has a bigger size (>30 µm) than other phases, indicating that it might be the first crystal that precipitates, having the longest time to grow. No free lime can be observed in the case of 10 wt pct Al2O3 addition, which is consistent with the XRD pattern.

Fig. 9
figure 9

Microstructural modification by Al2O3 addition (7 wt pct C): (a): 0 wt pct Al2O3, (b): 5 wt pct Al2O3, (c) 10 wt pct Al2O3

There is a phase containing most of the elements in the slag and cannot be linked to the composition of any crystalline phase. Its unique morphology is like a wall separating each crystal. It is therefore considered to be amorphous. In each sample, more than five points of the amorphous phases were measured. It was found that the CaO content increases from 39.9 to 46.3 wt pct upon increasing the Al2O3 addition from 5 to 10 wt pct, suggesting that the elimination of free lime is caused by dissolving more CaO into the amorphous phase (i.e., Al2O3 addition increases lime solubility of the slag).

Effect of Combination of SiO2 and Al2O3 on Solidification Microstructure During Reduction

As described above, with 5 wt pct Al2O3 and 7 wt pct C additions, free lime cannot be eliminated (see Figure 8). Figure 10 shows the phase identification for the slags with combined SiO2 and Al2O3 additions. An extra 1 wt pct SiO2 addition is able to remove the free lime completely. C3S, which is a favored phase for cement production,[27] can be formed under a combination of 5 wt pct Al2O3 and no more than 4 wt pct SiO2, as shown in Figure 10(a). Yet with 5 wt pct SiO2 + 5 wt pct Al2O3 (five SiO2 curves in Figure 10(a)) or 10 wt pct Al2O3 additions (Figure 10(b)), no C3S phase has been precipitated. The decrease of peaks at approximately 34 deg in Figure 10(a) indicates the decrease of C3S by adding SiO2. By referring to the EPMA observation, the peaks at around 40 deg in Figure 10(b) is considered as the overlap of C2S and C3S. Therefore, with the increasing SiO2, C3S is decreased, while C2S is increased due to the decrease of basicity B:

$$ B = \frac{{{\text{CaO}}\;({\text{wt}}\;{\text{pct}})}}{{{\text{SiO}}_{2} \;({\text{wt}}\;{\text{pct}}) + {\text{Al}}_{2} {\text{O}}_{3} \;({\text{wt}}\;{\text{pct}})}}. $$
(5)
Fig. 10
figure 10

Phase modification by different additions of SiO2 associated with (a) 7 wt pct C + 5 wt pct Al2O3 additions and (b) 7 wt pct C + 5 wt pct Al2O3

According to Eq. [5], B > 1.98 is a tentative criterion to precipitate C3S (at visible extent). On the other hand, B < 2.59 is required to avoid any free lime. Therefore, the optimized basicity to maintain C3S and to remove free lime ranges from 1.98 to 2.59.

Figure 11 presents the combined effect of SiO2 and Al2O3 additions on slag microstructural modification for the reduction test with 7 wt pct C addition. No free lime is observed in the test, which agrees well with the XRD patterns (see Figure 10). The composition of each phase was characterized by WDS analysis. The crystals could be identified by their XRD patterns. By comparing samples with the same level of SiO2 addition (Figures 11(a) and (d), (b) and (e), (c) and (f)), the amorphous phase is found to increase with the increasing Al2O3 addition. By comparing the amorphous phase under the same level of Al2O3 addition (Figures 11(a) through (c), and (d) through (e)), the amorphous phase is decreased with the increasing SiO2 additions.

Fig. 11
figure 11

Microstructural modification due to SiO2 and Al2O3 additions with 7 wt pct C addition. “mSnA” represents m wt pct SiO2 and n wt pct Al2O3 addition

For a quantitative understanding of the effect of combined SiO2 and Al2O3 additions on the mineral modification associated with the reduction, the composition of each microstructure was measured using WDS analysis. To ensure the accuracy, every microstructure was measured in more than 10 points. A mass balance calculation of Al2O3, CaO, and SiO2 was carried out using the overall composition and the composition of each phase to evaluate the fraction of different phases. Specifically, Al2O3, CaO, and SiO2 are assumed to be in the amorphous, C3S, and C2S phases, respectively. Thus, the Eqs. [6-1] through [6-3] can be derived.

$$ m_{{{\text{Al}}_{2} {\text{O}}_{3} }} = m_{\text{Am}} \times w_{{{\text{A}}_{1} }} + m_{{{\text{C}}_{3} {\text{S}}}} \times w_{{{\text{A}}_{2} }} + m_{{{\text{C}}_{3} {\text{S}}}} \times w_{{{\text{A}}_{3} }} , $$
(6-1)
$$ m_{\text{CaO}} = m_{\text{Am}} \times w_{{{\text{C}}_{1} }} + m_{{{\text{C}}_{3} {\text{S}}}} \times w_{{{\text{C}}_{2} }} + m_{{{\text{C}}_{2} {\text{S}}}} \times w_{{{\text{C}}_{3} }} , $$
(6-2)
$$ m_{{{\text{SiO}}_{2} }} = m_{\text{Am}} \times w_{{{\text{S}}_{1} }} + m_{{{\text{C}}_{3} {\text{S}}}} \times w_{{{\text{S}}_{2} }} + m_{{{\text{C}}_{2} {\text{S}}}} \times w_{{{\text{S}}_{3} }} , $$
(6-3)

where \( m_{{{\text{Al}}_{2} {\text{O}}_{3} }} , \) m CaO, and \( m_{{{\text{SiO}}_{2} }} \) are the overall mass percentages of the Al2O3, CaO, and SiO2, respectively; m Am, \( m_{{{\text{C}}_{3} {\text{S}}}} , \) and \( m_{{{\text{C}}_{2} {\text{S}}}} \) are the mass percentages of amorphous, C3S, and C2S phases, respectively; w A, w C, and w S are the mass fractions of Al2O3, CaO, and SiO2, respectively, in which the subscript indicates the mass fractions in amorphous (1), C3S (2), and C2S (3) phases. In this way, the mineral modification by the combined addition of SiO2 and Al2O3 can be obtained. The main minerals are presented in Figure 12.

Fig. 12
figure 12

Modification of (a) amorphous, (b) C3S, and (c) C2S distribution by the combined addition of SiO2 and Al2O3

The amount of the amorphous phase (Figure 12(a)) increases significantly upon increasing the amount of Al2O3 addition from 5 to 10 wt pct, which is in reasonable agreement with Figure 11. Meanwhile, the amount of amorphous phase decreases with SiO2 addition, specifically for 10 wt pct Al2O3 addition. Due to the fact that the metallic Fe particle size is decreased by lowering the amorphous fraction, the Fe size is expected to decrease with SiO2 addition. The amount of C3S decreases upon increasing the amount of either Al2O3 or SiO2 addition (see Figure 12(b)) due to the decrease of basicity. By adding 10 wt pct Al2O3, C3S is completely eliminated. On the contrary, the amount of C2S increases, upon the additions of both Al2O3 and SiO2 (see Figure 12(c)). The increase in the amount of C2S phase resulting from the addition of SiO2 might be attributed to the decrease of basicity; consequently, more amount of C2S was precipitated from the molten slag at high temperature. The changes in the amounts of C2S and C3S upon SiO2 addition at the addition level of 5 wt pct Al2O3 are more significant than the corresponding changes at the addition level of 10 wt pct Al2O3. According to Eq. [5], the basicity change by SiO2 addition at a lower level of Al2O3 addition is more sensitive than that at a higher level of Al2O3 addition. The change of basicity level could explain the behavior of C2S and C3S precipitations.

The ternary phase diagram CaO-SiO2-Al2O3 was calculated at 1873 K with FactSage 7.0 and is shown in Figure 13. To find the position indicating the present composition in the phase diagram, the slag was simplified to a ternary system of CaO-SiO2-Al2O3, and accordingly, the composition is normalized. The calculated phase diagram is shown, where “+” and “*” represent slags with additions of 5 and 10 wt pct Al2O3, respectively. The solid point shows the composition of the original slag. By adding 5 and 10 wt pct Al2O3, as indicated by the arrows, the composition moves closer to the liquid area (the liquidus is highlighted by the thick curve in Figure 13). Furthermore, the dashed lines S and A, respectively, represent the modification direction by SiO2 (S) and by Al2O3 (A) additions. With the SiO2 addition, the modified slag moves through region 3 (liquid + C2S + C3S) to region 2 (liquid + C2S), which indicates that SiO2 addition decreases C3S and increases C2S. Meanwhile, with Al2O3 addition, the modified slag moves through region 3 (liquid + C2S + C3S), region 2 (liquid + C2S), and then to region 1 (liquid). It demonstrates that with limited Al2O3 addition, C3S can decrease and C2S can increase, but further additions of Al2O3 decreases C2S and results in a fully liquid slag. Therefore, the amount of each phase can be controlled by controlling the addition amounts of SiO2 and Al2O3.

Fig. 13
figure 13

The ternary phase diagram for CaO-SiO2-Al2O3 as calculated at 1873 K (1600 °C) using FactSage 7.0. The lines indicated by A and S represent the compositional modifications due to Al2O3 and SiO2 additions, respectively

Conclusions

The valorization of a typical BOF slag with high basicity was studied by carbothermic reduction and slag microstructural modification through combined SiO2 and Al2O3 additions. The reductions of Fe oxides and P-containing compound were discussed, and optimized process conditions to recover high-grade metallic iron (>98 wt pct) were proposed. The formation and growth of the extracted metallic Fe were observed. Due to their excellent electronic, magnetic, and chemical properties, the micron-sized iron particles may have high- value-added applications in preparing magnetorheological fluids, metal-injected molding, and microwave absorption materials.[28] The effects of SiO2 and Al2O3 on the slag microstructure after solidification were investigated. The formation of the amorphous phases, and C2S and C3S crystals, can be controlled by suitable additions, implying potential applications as cementitious substitutions. The main conclusions can be summarized as follows:

  1. (1)

    With the increasing C addition, the P-rich phase changes from slag to metal. By controlling C addition, it is possible to avoid contamination of metallic Fe by P during carbothermic reduction. It is suggested to keep the molar ratio of C to iron oxides as 3:1, in order to minimize the P content in the metallic Fe;

  2. (2)

    the smaller metallic Fe particles (<10 µm) are formed from the dispersed RO in the lime, and the larger ones (<10 µm) are formed from the bigger-sized RO and C2AF. The growth of smaller particles is dominated by solid diffusion, and the growth of the larger particles is influenced by liquid fraction of the slag at high temperature (liquid-induced flow, diffusion, and convection). Based on the findings in this study, it can be concluded that the Fe particle purity, size, and size distribution can be manipulated through the process control with carbothermic reduction parameters, such as carbon and slag modifier additions and temperature;

  3. (3)

    the optimized basicity to precipitate C3S crystal and remove free lime (for high value-added application of the slag) should be controlled in the range from 1.98 to 2.59;

  4. (4)

    Al2O3 addition enlarges the liquid fraction at high temperature and consequently decreases the viscosity of the current slag; SiO2 addition effectively stabilizes free lime and promotes the formation of the C2S phase. Depending on the application purpose, the microstructure and mineralogy of the slag product can be tailored through the combined effects of Al2O3 and SiO2 additions.