Dissolution Behavior of Alumina-Based Inclusions in CaF 2 -Al 2 O 3 -CaO-MgO-SiO 2 Slag Used for the Electroslag Metallurgy Process

: Removal of non-metallic inclusions to CaF 2 -based slag is one of the most important functions of electroslag remelting. In this work, the dissolution behavior for alumina-based inclusions in CaF 2 -Al 2 O 3 -CaO-MgO-SiO 2 slag has been investigated. Results indicate that the diffusion or permeability capacity of slag components into alumina particles is F − , Ca 2+ , Si 4+ , Mg 2+ , from strongest to weakest, for CaF 2 -Al 2 O 3 -CaO-MgO-SiO 2 slag. Alumina inclusions react with F − in liquid slag at ﬁrst and then react with CaO to form x CaO- y Al 2 O 3 system. Subsequently, MgO substitutes for CaO to form a MgO-Al 2 O 3 system layer surrounding the other product and reactant, and then enters the liquid slag. CaF 2 can improve the dissolution capacity of slag to alumina inclusions. A complex region was formed between alumina-based particles and the slag, with different areas dominated by CaF 2 , CaO-Al 2 O 3 , CaO-SiO 2 and MgO-Al 2 O 3 . The dissolution process of alumina particles in slag is different from the formation of compound inclusions originated from the Al-O deoxidization reaction.


Introduction
Electroslag remelting has been widely used for the manufacture of high quality steel and alloys. This process creates a sound product with a homogeneous compositions and a low concentration of very small inclusions [1]. As is well known, the removal of non-metallic inclusions from the consumable electrode is one of the most important characteristics for the electroslag remelting process. The process of removing inclusions mainly happens at the tip of the electrode, and involves the absorption and dissolution of inclusions in fluoride-containing slag [2,3]. Although the removal process of inclusions has been studied during electroslag remelting, the mechanism for their absorption and dissolution in slag is not clear. Alumina-based inclusions are typically the most common inclusions in Al-killed steel [4,5]. Therefore, the investigation of the dissolution of alumina inclusions in fluoride slag is very significant.
Many studies of the dissolution behavior of alumina inclusions in slag systems [6][7][8][9] including CaO-Al 2 O 3 -SiO 2 melts [7,8], CaO-Al 2 O 3 -SiO 2 -MgO [9], and Al 2 O 3 -CaO-FeO x -SiO 2 [10] have been carried out. They had found that the alumina dissolution rate was affected by the rotation speed of the rod, the temperature and the viscosity of the slag [6,11]. Yan et al. [9] found that the rate-limiting step of Al 2 O 3 dissolution is mass transfer through the boundary layer and established a range of effective diffusion coefficients. Bui et al. [12] investigated the effect of the alumina content and solid phase in molten flux on the dissolution of alumina. In addition, they indicated that an intermediate compound played an important role in alumina dissolution in slag. Park et al. [8] investigated the interfacial reaction between alumina refractory and CaO-CaF 2 -SiO 2 -Al 2 O 3 -MgO-MnO slag and the impact of the CaO/SiO 2 ratio on the spinel solution. Some investigations [13][14][15] have studied alumina solubility and diffusion in fluoride electrolytes. Frazer et al. [13] studied alumina particle solubility and diffusion behavior in NaF-AlF 3 and NaF-AlF 3 -CaF 2 systems and estimated the diffusion coefficient. Yang et al. [15] measured alumina solubility in a range of KF-AlF 3 -based low-temperature electrolyte systems.
In recent years, Monaghan and Chen [16], Fox and Valdez et al. [17], Verhaeghe and Liu et al. [18], Liu and Verhaeghe et al. [19], and Orrling and Sridhar et al. [20] employed laser scanning confocal microscopy (CSLM) to investigate the dissolution of inclusions in slag. CSLM enables the in-situ observation of the dissolution of inclusions. A transparent liquid slag is an important requirement for these types of investigations; however, most fluoride slags are not transparent. Therefore, CSLM cannot be used to study the dissolution of inclusions in most fluoride-containing slags. Though Sridhar et al. [21] and Yu et al. [12] investigated the behavior of alumina in Na 2 O-CaO-Al 2 O 3 -SiO 2 -CaF 2 , it has a low fluoride content in the slag and the situation is different for fluoride-containing slags. Examples cannot be found in the literature for the dissolution of alumina inclusions in the fluoride-containing slag used in the electroslag remelting process. The dissolution/removal/modification of inclusions exists during electroslag remelting. However, dissolution is the first process, which is the important premise of removal and modification. Therefore, in this paper, the absorption and dissolution mechanisms of alumina inclusions in fluoride-containing slag have been investigated.

Materials and Methods
The experiment has been carried out using a resistance furnace (Ferrous Metallurgy Institute of Northeastern University, Shenyang, China) as shown in Figure 1. The heating unit is made of molybdenum disilicide. B-type thermocouple has been adopted for measuring the furnace temperature.
compound played an important role in alumina dissolution in slag. Park et al. [8] investigated the interfacial reaction between alumina refractory and CaO-CaF2-SiO2-Al2O3-MgO-MnO slag and the impact of the CaO/SiO2 ratio on the spinel solution.
Some investigations [13][14][15] have studied alumina solubility and diffusion in fluoride electrolytes. Frazer et al. [13] studied alumina particle solubility and diffusion behavior in NaF-AlF3 and NaF-AlF3-CaF2 systems and estimated the diffusion coefficient. Yang et al. [15] measured alumina solubility in a range of KF-AlF3-based low-temperature electrolyte systems.
In recent years, Monaghan and Chen [16], Fox and Valdez et al. [17], Verhaeghe and Liu et al. [18], Liu and Verhaeghe et al. [19], and Orrling and Sridhar et al. [20] employed laser scanning confocal microscopy (CSLM) to investigate the dissolution of inclusions in slag. CSLM enables the in -situ observation of the dissolution of inclusions. A transparent liquid slag is an important requirement for these types of investigations; however, most fluoride slags are not transparent. Therefore, CSLM cannot be used to study the dissolution of inclusions in most fluoride-containing slags. Though Sridhar et al. [21] and Yu et al. [12] investigated the behavior of alumina in Na2O-CaO-Al2O3-SiO2-CaF2, it has a low fluoride content in the slag and the situation is different for fluoride-containing slags. Examples cannot be found in the literature for the dissolution of alumina inclusions in the fluoride-containing slag used in the electroslag remelting process. The dissolution/removal/modification of inclusions exists during electroslag remelting. However, dissolution is the first process, which is the important premise of removal and modification. Therefore, in this paper, the absorption and dissolution mechanisms of alumina inclusions in fluoride-containing slag have been investigated.

Materials and Methods
The experiment has been carried out using a resistance furnace (Ferrous Metallurgy Institute of Northeastern University, Shenyang, China) as shown in Figure 1. The heating unit is made of molybdenum disilicide. B-type thermocouple has been adopted for measuring the furnace temperature.  The material preparations before experiment are as follows. The bulk slag materials used in the experiment are composed of analytical reagents and were weighed using a high-precision electronic balance. The principal component content for each slag material is larger than 98.5%. The particle size is close to 0.15 mm. Slags charged in an alumina crucible were calcined at 1123 K for 4 h after thoroughly mixing in an agate mortar. The alumina crucible was then stored in a desiccator before the experiment. Experiments were carried out with two kinds of slag mixtures, including 50%CaF2-20%Al2O3-20%CaO-5%MgO-5%SiO2 (Slag S1) and 70%CaF2-30%Al2O3 (Slag S2). The dimensions of alumina-based particle are 600-800 μm. Most alumina particles were subspheroidal in shape. X-ray diffraction (XRD) was used to analyze the phase components in the alumina-based as The material preparations before experiment are as follows. The bulk slag materials used in the experiment are composed of analytical reagents and were weighed using a high-precision electronic balance. The principal component content for each slag material is larger than 98.5%. The particle size is close to 0.15 mm. Slags charged in an alumina crucible were calcined at 1123 K for 4 h after thoroughly mixing in an agate mortar. The alumina crucible was then stored in a desiccator before the experiment. Experiments were carried out with two kinds of slag mixtures, including 50%CaF 2 -20%Al 2 O 3 -20%CaO-5%MgO-5%SiO 2 (Slag S1) and 70%CaF 2 -30%Al 2 O 3 (Slag S2). The dimensions of alumina-based particle are 600-800 µm. Most alumina particles were subspheroidal in shape. X-ray diffraction (XRD) was used to analyze the phase components in the alumina-based as shown in Figure 2. In this example, the amount of Al 2 O 3 is about 87.3% according to the XRD analysis. The other phases present are 8.5% CaAl 12 O 19 , 3.57% MgAl 2 O 4 and 1.4% SiO 2 , respectively. shown in Figure 2. In this example, the amount of Al2O3 is about 87.3% according to the XRD analysis. The other phases present are 8.5% CaAl12O19, 3.57% MgAl2O4 and 1.4% SiO2, respectively. The detailed experiment was carried out according to the following procedure. At first, slag about 400 g in weight surrounded by a 0.3-mm-thick molybdenum film from the side and bottom was placed into a graphite crucible. Graphite crucible is 40 mm and 70 mm in inner diameter and height. The power was switched on after the graphite crucible had been placed in the furnace chamber. The furnace was heated to 1823 K at a rate of 6 K/min and then kept this temperature for 10 min. Then, about 15 g of the alumina-based particle were introduced into the liquid slag pool through hollow alumina tube of 20 mm in inner diameter as shown in Figure 1. After 10 s from alumina-based particle addition, an alumina crucible tied on a long molybdenum wire was introduced into liquid slag and then slag specimen was taken from the liquid slag pool. Immediately, the liquid slag in alumina crucible was quenched on a large steel block, on which had a cylindrical recess to keep the slag cylindrical. And then more specimens were taken from the liquid slag pool in order to investigate the solubility situation of alumina-based inclusion in slag. The argon with the 99.99% in purity, used as the shielding gas, was introduced into the chamber at 3 L/min by inlet pipeline from the top when the chamber temperature rose up to 573 K. The shielding gas was kept this rate until the experiment had been finished.

Characteristics of Alumina-Based Inclusions in Slag
The slag specimens obtained in the experiment were then ground and polished using standard metallography procedures. Some alumina-based inclusions were found in the slag specimens. Figure 3a shows the morphology of an alumina particle in solidified slag. It can be seen that the alumina particles do not dissolve completely. The color of the bulk slag and alumina particle is obviously different as observed by SEM. A clear boundary between the inclusion and slag can be observed due to different chemical compositions.
The white line of dashes in Figure 3a, a typical dissolution boundary between the inclusion and slag S1, has been enlarged in Figure 3b. The yellow line in Figure 3b is an obvious transition region, which is different in color from the other part. The characteristics of the boundary layer are similar for other particles found in slag specimens. Moreover, the inclusion particle dissolves with the duration of the reaction time and the boundary layer shifts to the inner part of the particle by analyzing the slag specimens at different times. The detailed experiment was carried out according to the following procedure. At first, slag about 400 g in weight surrounded by a 0.3-mm-thick molybdenum film from the side and bottom was placed into a graphite crucible. Graphite crucible is 40 mm and 70 mm in inner diameter and height. The power was switched on after the graphite crucible had been placed in the furnace chamber. The furnace was heated to 1823 K at a rate of 6 K/min and then kept this temperature for 10 min. Then, about 15 g of the alumina-based particle were introduced into the liquid slag pool through hollow alumina tube of 20 mm in inner diameter as shown in Figure 1. After 10 s from alumina-based particle addition, an alumina crucible tied on a long molybdenum wire was introduced into liquid slag and then slag specimen was taken from the liquid slag pool. Immediately, the liquid slag in alumina crucible was quenched on a large steel block, on which had a cylindrical recess to keep the slag cylindrical. And then more specimens were taken from the liquid slag pool in order to investigate the solubility situation of alumina-based inclusion in slag. The argon with the 99.99% in purity, used as the shielding gas, was introduced into the chamber at 3 L/min by inlet pipeline from the top when the chamber temperature rose up to 573 K. The shielding gas was kept this rate until the experiment had been finished.

Characteristics of Alumina-Based Inclusions in Slag
The slag specimens obtained in the experiment were then ground and polished using standard metallography procedures. Some alumina-based inclusions were found in the slag specimens. Figure 3a shows the morphology of an alumina particle in solidified slag. It can be seen that the alumina particles do not dissolve completely. The color of the bulk slag and alumina particle is obviously different as observed by SEM. A clear boundary between the inclusion and slag can be observed due to different chemical compositions.
The white line of dashes in Figure 3a, a typical dissolution boundary between the inclusion and slag S1, has been enlarged in Figure 3b. The yellow line in Figure 3b is an obvious transition region, which is different in color from the other part. The characteristics of the boundary layer are similar for other particles found in slag specimens. Moreover, the inclusion particle dissolves with the duration of the reaction time and the boundary layer shifts to the inner part of the particle by analyzing the slag specimens at different times.  Table 1. It can be seen from the results that point P1 represents the bulk slag and point P4 is the inclusion body. Points P2 and P3 are in the boundary layer between the bulk slag and alumina-based particle. A higher Mg concentration exists at point P2, which may be due to the diffusion of Mg 2+ in slag from the liquid slag to the alumina particle. High contents of Mg, Al and O at point P2 may be attributed to the formation of the MgO-Al2O3 system. Line scanning has been done in order to investigate the dissolution mechanism of alumina inclusions in fluoride-containing slag as shown in Figure 4. The regions b, c, d and e are proposed as the dissolution boundary because of the rapid decrease or increase in F, Ca, O and Al. This proves that inclusion dissolution and ion diffusion in liquid slag occur simultaneously.   Table 1. It can be seen from the results that point P1 represents the bulk slag and point P4 is the inclusion body. Points P2 and P3 are in the boundary layer between the bulk slag and alumina-based particle. A higher Mg concentration exists at point P2, which may be due to the diffusion of Mg 2+ in slag from the liquid slag to the alumina particle. High contents of Mg, Al and O at point P2 may be attributed to the formation of the MgO-Al 2 O 3 system. Line scanning has been done in order to investigate the dissolution mechanism of alumina inclusions in fluoride-containing slag as shown in Figure 4. The regions b, c, d and e are proposed as the dissolution boundary because of the rapid decrease or increase in F, Ca, O and Al. This proves that inclusion dissolution and ion diffusion in liquid slag occur simultaneously.  Table 1. It can be seen from the results that point P1 represents the bulk slag and point P4 is the inclusion body. Points P2 and P3 are in the boundary layer between the bulk slag and alumina-based particle. A higher Mg concentration exists at point P2, which may be due to the diffusion of Mg 2+ in slag from the liquid slag to the alumina particle. High contents of Mg, Al and O at point P2 may be attributed to the formation of the MgO-Al2O3 system. Line scanning has been done in order to investigate the dissolution mechanism of alumina inclusions in fluoride-containing slag as shown in Figure 4. The regions b, c, d and e are proposed as the dissolution boundary because of the rapid decrease or increase in F, Ca, O and Al. This proves that inclusion dissolution and ion diffusion in liquid slag occur simultaneously.  More slag samples at different times were analyzed. Figure 5 shows one of the results after about 20 s of the addition of particle, which is similar to the results in Figure 4. More slag samples at different times were analyzed. Figure 5 shows one of the results after about 20 s of the addition of particle, which is similar to the results in Figure 4.

Dissolution Behavior of Alumina Inclusions in Slag
It can be seen from Figures 4 and 5 that region f can be regarded as the inclusion body because of the constant Al content and lower F, Ca, Mg and Si levels. Region a is the bulk slag. Region e is the innermost reaction layer, in which higher F and Ca have been revealed. It indicates that the diffusion or penetration rate of Ca 2+ and F − toward the inclusion is faster than other components in the liquid slag. It can be inferred from the results obtained in Figures 3-5 that the following reaction has happened in this region.
The reaction product of Reaction (2) could then react with the alumina inclusion to form a CaO-Al2O3 system, as can be proved by the CaF2-Al2O3 binary phase diagram. The molar ratio of CaO and Al2O3 mainly depends on the content of each component, which can also be clarified by the CaO-Al2O3 phase diagram as shown in Figure 6 [22]. In addition, the existence of F − can improve the slag dissolution capacity to non-metallic inclusions, as can be proved by the findings of Cho and Fan [23]. They investigated the diffusion

Dissolution Behavior of Alumina Inclusions in Slag
It can be seen from Figures 4 and 5 that region f can be regarded as the inclusion body because of the constant Al content and lower F, Ca, Mg and Si levels. Region a is the bulk slag. Region e is the innermost reaction layer, in which higher F and Ca have been revealed. It indicates that the diffusion or penetration rate of Ca 2+ and F − toward the inclusion is faster than other components in the liquid slag. It can be inferred from the results obtained in Figures 3-5 that the following reaction has happened in this region.
The reaction product of Reaction (2) could then react with the alumina inclusion to form a CaO-Al 2 O 3 system, as can be proved by the CaF 2 -Al 2 O 3 binary phase diagram. The molar ratio of CaO and Al 2 O 3 mainly depends on the content of each component, which can also be clarified by the CaO-Al 2 O 3 phase diagram as shown in Figure 6 [22]. More slag samples at different times were analyzed. Figure 5 shows one of the results after about 20 s of the addition of particle, which is similar to the results in Figure 4.

Dissolution Behavior of Alumina Inclusions in Slag
It can be seen from Figures 4 and 5 that region f can be regarded as the inclusion body because of the constant Al content and lower F, Ca, Mg and Si levels. Region a is the bulk slag. Region e is the innermost reaction layer, in which higher F and Ca have been revealed. It indicates that the diffusion or penetration rate of Ca 2+ and F − toward the inclusion is faster than other components in the liquid slag. It can be inferred from the results obtained in Figures 3-5 that the following reaction has happened in this region.
The reaction product of Reaction (2) could then react with the alumina inclusion to form a CaO-Al2O3 system, as can be proved by the CaF2-Al2O3 binary phase diagram. The molar ratio of CaO and Al2O3 mainly depends on the content of each component, which can also be clarified by the CaO-Al2O3 phase diagram as shown in Figure 6 [22]. In addition, the existence of F − can improve the slag dissolution capacity to non-metallic inclusions, as can be proved by the findings of Cho and Fan [23]. They investigated the diffusion In addition, the existence of F − can improve the slag dissolution capacity to non-metallic inclusions, as can be proved by the findings of Cho and Fan [23]. They investigated the diffusion Metals 2016, 6, 273 6 of 8 dissolution of alumina in various steelmaking slags. The addition of CaF 2 in steelmaking slag can improve the dissolution rate of alumina particles in slag [23,24].
Higher Al and Ca but lower F, Mg and Si have been obtained for region d, implying that calcium aluminates could be present. The main component in region b is the MgO-Al 2 O 3 system; in addition, there is a small fraction of CaO-Al 2 O 3 spinel. The following reactions maybe happen from region c to region b.
Region c appears to be a transition region between the calcium aluminate-dominated region d and the spinel-dominated region b. Within region c there is an additional trend of an increase of the calcium and silicon levels, with a corresponding drop in aluminum, implying the formation of calcium silicates.
However, it is very complex for region c, in which the change trend is similar for Ca and Si, which first increase and then decrease. Al decreases first and then increases. It can be only inferred that CA6(CaO·6Al 2 O 3 ), CA2(CaO·2Al 2 O 3 ), CA(CaO·Al 2 O 3 ), C12A7(12CaO·7Al 2 O 3 ) and C3A(3CaO·Al 2 O 3 ) may be present [22]. The content of CaO decreases; in contrast, Al 2 O 3 increases at the same time in region c from the alumina particle to bulk slag direction. The reaction can be expressed by Reaction (2).
Results show only an increase in the MgO level in region b, with the level in other regions not going above the bulk slag levels. The present results indicate that an alumina inclusion will initially be attacked by calcium fluoride-based slag. For the experimental time used, complex behavior was observed in the region between the bulk alumina particle and the slag. After the initial attack area, there were regions dominated by calcium aluminates, calcium silicates and spinel, indicating a progression of reactions.
This behavior is different from that in conventional steelmaking, with alumina inclusions generated by Al deoxidation reacting with MgO to form MgO·Al 2 O 3 at first, and then the Ca element substitution for the Mg element in MgO·Al 2 O 3 inclusions to form CaO·Al 2 O 3 outside of the MgO·Al 2 O 3 core. The dissolution mechanism of alumina inclusions in CaF 2 -Al 2 O 3 -CaO-MgO-SiO 2 slag can be illustrated as in Figure 7.  [23,24]. Higher Al and Ca but lower F, Mg and Si have been obtained for region d, implying that calcium aluminates could be present. The main component in region b is the MgO-Al2O3 system; in addition, there is a small fraction of CaO-Al2O3 spinel. The following reactions maybe happen from region c to region b.
Region c appears to be a transition region between the calcium aluminate-dominated region d and the spinel-dominated region b. Within region c there is an additional trend of an increase of the calcium and silicon levels, with a corresponding drop in aluminum, implying the formation of calcium silicates.
However, it is very complex for region c, in which the change trend is similar for Ca and Si, which first increase and then decrease. Al decreases first and then increases. It can be only inferred that CA6(CaO·6Al2O3), CA2(CaO·2Al2O3), CA(CaO·Al2O3), C12A7(12CaO·7Al2O3) and C3A(3CaO·Al2O3) may be present [22]. The content of CaO decreases; in contrast, Al2O3 increases at the same time in region c from the alumina particle to bulk slag direction. The reaction can be expressed by Reaction (2).
Results show only an increase in the MgO level in region b, with the level in other regions not going above the bulk slag levels. The present results indicate that an alumina inclusion will initially be attacked by calcium fluoride-based slag. For the experimental time used, complex behavior was observed in the region between the bulk alumina particle and the slag. After the initial attack area, there were regions dominated by calcium aluminates, calcium silicates and spinel, indicating a progression of reactions.
This behavior is different from that in conventional steelmaking, with alumina inclusions generated by Al deoxidation reacting with MgO to form MgO·Al2O3 at first, and then the Ca element substitution for the Mg element in MgO·Al2O3 inclusions to form CaO·Al2O3 outside of the MgO·Al2O3 core. The dissolution mechanism of alumina inclusions in CaF2-Al2O3-CaO-MgO-SiO2 slag can be illustrated as in Figure 7.   Figure 8 shows a SEM-SE image of the dissolution boundary between the inclusion and slag S2 as well as the line-scanning results of the boundary layer by SEM-SBSE. It can be seen that the color of the bulk slag and alumina particles is obviously different and the boundary is very clear. Similar results have been obtained, but it is easier than the dissolution of inclusions in slag S1. Only CaO·6Al2O3 exists in the dissolution process, which is from Reactions (1) and (2) because enough Al and less Ca exist in this region for slag S2. It can further illustrate that the resource of Si in region c of Figure 4 is from the diffusion of SiO2 from the slag into the inclusions instead of the SiO2 in the inclusions itself because no Si has been found in Figure 8, based on the same alumina inclusion particles in the experiment.  Figure 8 shows a SEM-SE image of the dissolution boundary between the inclusion and slag S2 as well as the line-scanning results of the boundary layer by SEM-SBSE. It can be seen that the color of the bulk slag and alumina particles is obviously different and the boundary is very clear. Similar results have been obtained, but it is easier than the dissolution of inclusions in slag S1. Only CaO·6Al 2 O 3 exists in the dissolution process, which is from Reactions (1) and (2) because enough Al and less Ca exist in this region for slag S2. It can further illustrate that the resource of Si in region c of Figure 4 is from the diffusion of SiO 2 from the slag into the inclusions instead of the SiO 2 in the inclusions itself because no Si has been found in Figure 8, based on the same alumina inclusion particles in the experiment.