EffEct of caf 2 SuBStItutIon wIth tio 2 on thE cryStallIzatIon charactErIStIcS of low-fluorIdE Slag for ElEctroSlag rEMEltIng

The effect of CaF 2 substitution with TiO 2 on the crystallization characteristics of low-fluoride slag was studied using differential scanning calorimetry (DSC) combined with XRD and SEM-EDS analysis. The effective activation energy for crystallization of the slag was evaluated. The results showed that the liquidus temperature of the slag increased unnoticeably with increasing TiO 2 content. Increasing TiO 2 addition from 4.3 wt% to 13.0 wt% decreased the undercooling of the slag and enhanced the crystallization ability of the slag. There is no change in the types and precipitation sequence of the crystalline phase in the slag with different TiO 2 contents during continuous cooling. The crystalline phases were Ca 12 Al 14 O 32 F 2 , CaTiO 3 , MgO, and CaF 2 . The first and second crystallization phase were Ca 12 Al 14 O 32 F 2 and CaTiO 3 , respectively. The dominant crystalline phase was faceted Ca 12 Al 14 O 32 F 2 crystals. The morphology of CaTiO 3 crystal changed from needle-like to blocky with increasing TiO 2 content. The MgO crystal was with little blocky morphology, and the needle-like CaF 2 distributed among CaTiO 3 crystal. The precipitated amount of both MgO and CaF 2 was very small. The effective activation energy for Ca 12 Al 14 O 32 F 2 formation decreased with increasing TiO 2 content in the slag, indicating that TiO 2 enhanced the crystallization tendency of the slag.


Introduction
Electroslag remelting (ESR) allows a good control of the cleanliness and solidification structure of the ingot [1]. Slag plays an important role in the ESR process, including generating Joule heat for melting electrode, removing harmful elements and non-metallic inclusions, etc [2][3][4]. The ESR slag contains a large amount of fluoride, such as 70%CaF 2 -30%A1 2 O 3 , 60%CaF 2 -20%CaO-20%A1 2 O 3 , 40%CaF 2 -30%CaO-30%A1 2 O 3 , etc., in order to decrease melting temperature and viscosity of the slag. However, the volatilization of fluoride from molten slag melt during ESR process has always been a big obstacle because it causes environmental pollution, corrosion of plant equipment, and the chemical composition change of the slag [5][6]. Furthermore, the chemistry change of slag caused by volatilization of fluoride generally reduces the stability of operating practice and degrade the quality of as-ingot [7].
Only a few studies regarding the development of low-fluorine or fluorine-free ESR slag have been reported. Narita et al. [8] investigated the application of CaO-Al 2 O 3 -based slag in ESR, and found that there was no trouble during the operation when using CaO-Al 2 O 3 -based slags with no fuming and good electric stability. Liang et al. [9] showed that the power consumption could decrease about 30% by using 49.5%CaO-43.7%Al 2 O 3 -6.8%SiO 2 slag and 52.0%CaO-41.2%Al 2 O 3 -6.8%SiO 2 in comparison to 60%CaF 2 -20%CaO-20%A1 2 O 3 slag, and reduce fluoride pollution as well. Mao et al. [10] reported that the CaO-Al 2 O 3 -MgO fluorine-free slag can guarantee the quality of G20CrNi2MoA as-ingot and improve environmental pollution. Rehak et al. [11] found that using fluorinefree CaO-Al 2 O 3 slag in the ESR production of Alkilled steel could decrease the amount of oxide inclusions. Although several applications of lowfluorine and fluorine-free ESR slag have been performed in industrial trial, the widely used ESR slag is still high-fluorine slag which has not been completely substituted by low-fluorine or fluorine-free ESR slag. This is because the application of these slags still generates several problems, such as the difficulties in liquid slag starting, controlling of the as-cast surface quality and high working voltage, etc., which
The main idea for the development of low-fluorine or fluorine-free ESR slag was to control the CaO/Al 2 O 3 ratio and substitute a small amount of MgO or SiO 2 for CaF 2 . In the ESR production of Ticontaining steel and alloy, TiO 2 is always added to the ESR slag to minimize the titanium loss [14,15]. Many researches on the low-fluorine and fluorine-free metallurgical slag suggested that TiO 2 has the similar effect as CaF 2 on the thermo-physical properties of the slag, such as reducing the viscosity and increasing electrical conductivity of slag [16][17][18]. Moreover, TiO 2 is more difficult to volatilize than CaF 2 . It provides guidance to develop low-fluorine ESR slag by substituting TiO 2 for CaF 2 in the ESR production of Ti-containing steel and alloy.
The previous studies [19][20][21] demonstrated that the crystallization characteristic of ESR slag was a vital factor in determining the current distribution in mold and surface quality of as-cast ingot. The present work was carried out to study the influence of substitution of CaF 2 with TiO 2 on the crystallization characteristic of low-fluoride ESR slag. The differential scanning calorimetry (DSC) was employed to study the crystallization behavior of the slag. Continuous cooling transformation (CCT) diagrams of crystalline phases were plotted based on the DSC curves. The types and morphology of the crystalline phases were identified by XRD and SEM-EDS analysis. In addition, the effective activation energy for crystallization of the slag was evaluated based on DSC thermal data.

Sample Preparation
Slag samples were prepared using reagent-grade powders of CaF 2 , CaCO 3 , Al 2 O 3 , MgO, Li 2 CO 3 , and TiO 2 powder. CaCO 3 powder was calcined at 1323 K for 10 h in muffle furnace to produce CaO. The thoroughly mixed powder was pre-melted at 1773 K (1500 °C) in a platinum crucible for 5 min, and subsequently the liquid sample was quenched in iced water and crushed. The chemical composition of the slag was analyzed using inductively coupled plasmaoptical emission spectroscopy. The fluorine content of pre-melted slag was determined by ion-selective electrode method. The chemical composition of the pre-melted slag is shown in Table 1.

DSC Measurement
The crystallization characteristics of the slag were investigated by DSC (Netzsch STA449F3; Netzsch Instrument Inc., Germany) measurements in Ar gas atmosphere at a flow rate of 70 mL/min. For each DSC measurement, approximately 50 mg of sample was heated at the heating rate of 30 K/min from room temperature up to 1773 K (1500 °C) in a platinum crucible and held for 1 min to eliminate bubbles and homogenize its chemical composition. Then, the liquid sample was cooled at a constant cooling rate (10 K/min, 15 K/min, 20 K/min and 25 K/min) to the room temperature.

SEM-EDS and XRD analysis
The slag samples after DSC measurements were mounted with epoxy resin and polished. Then, a thin platinum film was coated onto the cross section of the polished sample to enhance the sample electric conductivity. The morphology and composition of crystalline phases in slag samples after DSC measurements were determined by the SEM-EDS analysis.
Considering that the slag samples after DSC measurements were too small to determine crystalline phase by XRD analysis, a series of cooling experiments were carried out to determine the types of the crystalline phases in slag samples corresponding to each exothermic peak of the DSC curves. Approximately 2 g of slag was put in a platinum crucible at 1773 K (1500 °C) for 1 minute, then cooled at the cooling rate of 10 K/min to the target temperature and held for 1 h. The slag sample was quenched by ice water and the crystalline phases in the slag were analyzed by the XRD analysis. Figure 1 shows the DSC curves of the slag at four different cooling rates, i.e., 10 K/min, 15 K/min, 20 K/min, and 25 K/min, respectively. The exothermic peak on DSC curves indicates the crystalline phase formation event. Four exothermic peaks are observed on the DSC curves of the three slag samples at various cooling rates, which are designated as P1, P2, P3, and P4, respectively. It should be noted that the intensity of P3 and P4 on the DSC curves are much lower than those of P1 and P2 that the peaks are difficult to detect on the DSC curves, especially in Figure 1 Figure 1(b) shows that the third exothermic peak overlaps partially with the fourth exothermic peak at various cooling rates, indicating that the crystalline phase corresponding to P4 started to precipitate before the finish of the crystalline phase formation corresponding to P3. Furthermore, the intensity of first exothermic peak is much higher than the other three exothermic peaks for the three slag samples at various cooling rates, suggesting that the peak P1 should represent the dominate crystalline phase in the slag samples. Figure 2 shows the XRD patterns of the slag samples quenched at the different temperatures. Four crystalline phases were identified at 1000 °C in each slag sample by XRD, i.e., Ca 12 Al 14 O 32 F 2 , CaTiO 3 , MgO, and CaF 2 . There is no change in the types of the crystalline phase in the studied slag samples with different TiO 2 contents during continuous cooling. From Figure 2(c) to 2(e), the first and second exothermic peaks on the DSC curves for slag C represent the precipitation of Ca 12 Al 14 O 32 F 2 and CaTiO 3 during the continuous cooling of the slag, respectively. The third exothermic peak and the fourth exothermic peak on the DSC curves represent the precipitation of MgO and CaF 2 for the three slag samples at various cooling rates. According to the morphology of the crystallization phase shown in Section 3.3, it can be deduced that increasing TiO 2 content from 4.3 wt% to 13.0 wt% in the slag has no influence on the precipitation sequence of the crystalline phase. Therefore, it was concluded that the first and second exothermic peaks on the DSC curves for the slag correspond to the precipitation of Ca 12 Al 14 O 32 F 2 and CaTiO 3 , respectively. It should be noted that the XRD peaks of MgO and CaF 2 are much weaker than those of Ca 12 Al 14 O 32 F 2 , CaTiO 3 in Figure  3(a) to 3(c). In consideration that the precipitated amount of the MgO and CaF 2 are much lower in the slag and the peak P3 overlaps partially with peak P4 in Figure 2(b), the XRD identification was not carried out to identify the precipitation order of MgO and CaF 2 in the slag.

Crystallization temperature and crystallization ability of the slag
The onset temperature of the exothermic peak on the DSC curve can be determined as the precipitation temperature of crystalline phase, at which the crystalline phase starts to precipitate [22]. Figure 3 shows the CCT diagrams of the crystalline phases in the slag constructed based on the records on the DSC curves at various cooling rates to clarify the crystallization tendency of the studied slag. It can be found that the precipitation temperatures of the crystalline phases in each slag decrease with increasing cooling rate due to stronger driving force needed to promote the nucleation in the molten slag with increasing the cooling rate [23,24]. Therefore, the precipitation temperature decreases with increasing cooling rate to provide larger undercooling for the nucleation.
Based on the DSC curves of the slag samples, the CCT diagrams for first and second crystalline phase in the slag samples were constructed, as shown in Figure  4. The precipitation temperature of the first crystalline phase during cooling process represents the crystallization temperature of the slag. It is clear from Figure 4(a) that the precipitation temperature of the first crystalline phase Ca 12 Al 14 O 32 F 2 increases with increasing the TiO 2 content in slag, indicating that the addition of TiO 2 enhances the tendency of crystallization. As shown in Figure 4(b), the precipitation temperature of the second crystalline phase CaTiO 3 in slag A is much lower than those in slag B and slag C. And increasing TiO 2 content from Figure 1. DSC curves of the slag at various cooling rates 8.6 wt% to 13.0 wt% has no obvious effect on the precipitation temperature of CaTiO 3 . This result is consistent with the finding reported by others [25]. The slag with a higher crystallization temperature easily forms a thick slag skin and unstable heat flux across the slag skin. In this case, it provides an unsound condition for the surface quality of remelted ingot [20,27]. Therefore, it is not suitable to substitute excessive TiO 2 for CaF 2 in the low-fluoride slag for electroslag remelting.   The liquidus temperature of slag is the temperature at which solid slag is completely transformed into liquid, which can be obtained by extrapolating the crystallization temperatures of slag at the various cooling rates to the value at the cooling rate of 0 K/min [20]. The extrapolated liquidus temperature of the studied slag is 1719 K (1446 °C), 1720 K (1447 °C), and 1727 K (1454 °C), respectively. It suggested that the increasing of the liquidus temperatures of the studied slag was not very noticeable with increasing TiO 2 addition. The undercooling can be applied to evaluate the crystallization ability of the slag. The calculated undercoolings of the slag at various cooling rates are presented in Figure 5. It is noted that the undercooling of slag decreases with increasing TiO 2 addition in the slag, indicating that the crystallization ability of the slag was enhanced with increasing TiO 2 content from 4.3 wt% to 13.0 wt% in the slag.

SEM-EDS observation of the crystals in the slag
The morphologies and compositions of crystalline phases in the slag after DSC experiments were identified by SEM-EDS analysis. The types of the crystalline phases can be determined by combining the SEM-EDS with XRD results. Figures 6 to 9 show the SEM images of the slag after DSC measurements at the cooling rate of 10 K/min. Four crystalline phases were found in the slag, agreeing with DSC and XRD results. The element mappings of slag C shown in Figure 9 present the major element composition of the four crystalline phases. The dominant crystalline phase in the slag was faceted Ca 12 Al 14 O 32 F 2 crystal, taking up most of the crystalline fraction in the slag, which suggested that the crystallization of Ca 12 Al 14 O 32 F 2 crystal was controlled by interfacial reaction. Under this condition, viscosity had no vital influence on the crystallization of Ca 12 Al 14 O 32 F 2 crystal. Therefore, TiO 2 addition increasing activity values of some main components (CaO, Al 2 O 3 , and CaF 2 ) during Ca 12 Al 14 O 32 F 2 formation was the main factor in the enhancement of slag crystallization ability with increasing TiO 2 content. This is similar to the role of Li 2 O on the crystallization behavior of CaO-SiO 2based mold flux [25], but different from the effect of TiO 2 on the crystallization behavior of CaO-Al 2 O 3 -based mold flux when the crystallization behavior of slag is controlled by element diffusion [28]. The morphology of CaTiO 3 crystal changed from needle-like to blocky with increasing TiO 2 content. The CaTiO 3 crystal with nonfaceted morphology indicated that viscosity was the main factor determining the crystallization and the crystallization of CaTiO 3 crystal was controlled by the diffusion of species from bulk melts to crystal-melt interface. The MgO crystal was with little blocky morphology, and the CaF 2 crystal with needle-like morphology distributed among CaTiO 3 crystal. The precipitated amount of MgO and CaF 2 were both very small, which were in agreement with the DSC results.

Effective activation energy for crystallization of the slag
The effective activation energy for crystallization is an important parameter to reflect the crystallization tendency of the slag [17,27]. Many methods have been put forward to determine the effective activation energy for non-isothermal crystallization kinetic analysis [29][30][31][32]. Friedman method [33], a most widely used method for non-crystallization kinetics analysis during cooling process, was employed to estimate the effective activation energy for crystallization of the slag. The equation of Friedman method is expressed as follows: (1) where α is crystallization fraction, t, T α and E c(α) are the crystallization time, the crystallization temperature and the effective activation energy for crystallization at a given crystallization fraction α, respectively. The crystallization fraction α and its corresponding temperature T α can be evaluated from the DSC thermal data, as shown in Figure 10. The crystallization fraction α can be obtained by the ratio of A T /A, where T O and T E are the onset temperature and end temperature of exothermic peak, A T and A are the partial area at the temperature of T α and the total area of exothermic peak, respectively.
The effective activation energy for Ca 12 Al 14 O 32 F 2 crystallization in the slag are presented in Figure 11. It can be seen that there is a notable decrease on the effective activation energy for Ca 12 Al 14 O 32 F 2 crystallization with increasing TiO 2 content in the slag, indicating that TiO 2 enhanced the crystallization tendency for the slag to crystallize, which was consistent with the DSC results.
The effective activation energy for Ca 12 Al 14 O 32 F 2

conclusions
The effect of CaF 2 substitution with TiO 2 on the crystallization characteristics of low-fluoride slag was studied using differential scanning calorimetry (DSC) combined with the XRD and SEM-EDS analysis. The effective activation energy for crystallization of the slag was evaluated. The conclusions are summarized as follows: (1) The liquidus temperatures of the slag did not have noticeable increase with increasing TiO 2 addition. The undercooling of slag decreased with increasing TiO 2 addition from 4.3 wt% to 13.0 wt%, indicating that the crystallization ability of the slag was enhanced with increasing TiO 2 content in the slag. It is not suitable to substitute excessive TiO 2 for CaF 2 in the low-fluoride slag for electroslag remelting.
(2) There is no change in the types and precipitation sequence of the crystalline phase in the slag with different TiO 2 contents during continuous cooling. The crystalline phases were Ca 12 Al 14 O 32 F 2 , CaTiO 3 , MgO, and CaF 2 . The first and second crystallization phase were Ca 12 Al 14 O 32 F 2 and CaTiO 3 , respectively.
(3) The dominant crystalline phase in the slag was faceted Ca 12 Al 14 O 32 F 2 crystals. The morphology of CaTiO 3 crystal changed from needle-like to blocky with increasing TiO 2 content. The MgO crystal formed with little blocky morphology, and the CaF 2 crystal with needle-like morphology distributed among CaTiO 3 crystal. The precipitated amount of both MgO and CaF 2 were very small.
(4) The effective activation energy for Ca 12 Al 14 O 32 F 2 formation decreased with increasing TiO 2 content in the slag, indicating that TiO 2 enhanced the crystallization tendency of the slag.  Figure 11. Dependence of the effective activation energy for crystallization on the crystallization fraction