Effect of CaF₂ on Viscosity and Refining Ability of Highly Basic Slags for Duplex Stainless Steel

Abstract

To optimize CaF₂ content in highly basic CaO-18%Al₂O₃-SiO₂-10%MgO-CaF₂ (%CaO/%SiO₂=6, denoted as C/S=6) refining slags used for the production of Al-killed duplex stainless steel with high cleanliness demand, the effect of CaF₂ content on the viscosity and refining ability of the slags were studied and compared with typical CaF₂-free highly basic CaO-30%Al₂O₃-SiO₂-10%MgO (C/S=6) slag. The effect of CaF₂ addition in decreasing slag viscosity becomes less obvious with increasing temperature and CaF₂ content. When CaF₂ content exceeds 10%, slag viscosity only marginally decreases with further increasing CaF₂ content. Both monoxide-CaO and monoxide-MgO phases are precipitated in all the CaF₂-bearing slags. CaF₂ addition slightly increases monoxide-MgO precipitation, but dramatically decreases monoxide-CaO precipitation. Viscosities of the CaF₂-bearing slags were also theoretically calculated and good agreement with the measured values was observed. Moreover, the 6% CaF₂-bearing slag has very close viscosities above 1833 K but much lower viscosities below 1833 K, compared with the CaF₂-free highly basic slag. Further evaluation of the 6% CaF₂-bearing slag on steel cleanliness confirms that 6% CaF₂ addition is sufficient for the highly basic CaO-18%Al₂O₃-SiO₂-10%MgO-CaF₂ (C/S=6) slag. The mechanism of CaF₂ in decreasing the viscosity of CaF₂-bearing slags was discussed from the viewpoints that CaF₂ behaves as a network breaker and that CaF₂ suppresses the precipitation of solid phases. The first aspect was identified to play a much greater role in decreasing slag viscosity.

1. Introduction

Duplex stainless steel consists of roughly equal proportion of austenite and ferrite phases in their microstructures, exhibiting roughly twice the strength compared to austenitic stainless steel,¹⁾ and excellent corrosion resistance, particularly chloride stress corrosion and chloride pitting corrosion.²⁾ However, due to the presence of two phases which have very different deformation behaviors,³⁾ duplex stainless steel has inherently poor thermo-plasticity, often leading to the formation of edge cracks in steel plates or surface cracks in steel round bars under hot working conditions.⁴⁾ Non-metallic inclusions and impurity elements such as sulfur have very negative impact on steel properties, such as ductility, toughness, fatigue strength and corrosion resistance.^5,6) Therefore, Al deoxidation instead of Si deoxidation is currently used to achieve very high cleanliness in the production of duplex stainless steel.

The chemical composition of refining slag has significant effect on steel cleanliness. Extensive studies have shown that highly basic refining slag favors deoxidation,⁷⁾ desulfurization⁸⁾ as well as the transformation of irregular Al₂O₃ inclusions to globular CaO–MgO–Al₂O₃ inclusions.^7,9) Yoon et al.¹⁰⁾ reported that control of the %CaO/%Al₂O₃ ratio at 1.7–1.8 in highly basic refining slag is the most effective way to remove Al₂O₃ inclusions from bearing steel as the refining slag has low melting temperature and good fluidity. Similar %CaO/%Al₂O₃ ratios were also obtained by other researchers.^11,12) However, for stainless steel production including duplex stainless steel, the above optimal %CaO/%Al₂O₃ ratio in highly basic refining slag cannot be reached without Al₂O₃ addition as dissolved oxygen content in liquid stainless steel after decarburization via AOD (argon oxygen decarburization) or VOD (vacuum oxygen decarburization) process is much lower than that in carbon steel via BOF (basic oxygen furnace) process. The low dissolved oxygen content in liquid stainless steel limits the addition of Al deoxidizer and thus the Al₂O₃ content in refining slag. Therefore, fluorspar (CaF₂) is commonly added to decrease the viscosity of highly basic refining slag for the production of duplex stainless steel. A low slag viscosity enhances mass transfer in the slag phase, thus promoting the adsorption of non-metallic inclusions^13,14) and desulfurization.^8,15) Therefore, effect of CaF₂ content on the viscosity of some metallurgical slags has been extensively investigated. Wu et al.¹⁶⁾ found that addition of less than 5% CaF₂ effectively decreases the viscosity of highly basic CaO–Al₂O₃–SiO₂–MgO (C/S=1.95–8.0) slags containing 25% Al₂O₃ especially at low temperatures. Kim et al.¹⁷⁾ reported that the effect of CaF₂ addition on the viscosity of highly basic CaO–Al₂O₃–SiO₂–MgO (C/S=5.8–8.1) slags containing 26%–31% Al₂O₃ becomes less discernible when CaF₂ content is higher than 10%. Park et al.¹⁸⁾ compared the effect of CaF₂ and Al₂O₃ on the viscosity of low basicity CaO-SiO₂-10%MgO-CaF₂/Al₂O₃ slags and found that CaF₂ is more effective in decreasing viscosity of the slags. Pereira et al.¹⁹⁾ reported that addition of 2%–5% CaF₂ in low basicity slags (C/S=1.62–2.26) containing 7%–13% Al₂O₃ decreases the viscosity of the slags from 0.45 Pa·s to 0.10 Pa·s. Deng et al.²⁰⁾ revealed that CaF₂ addition not only decreases the viscosity of low basicity SiO₂–CaO–Al₂O₃–MgO silicate melts also changes the crystallization behavior from spherical diopside to plate-like anorthite. Wang et al.²¹⁾ observed that the viscosity of CaO-SiO₂-Al₂O₃-based low basicity mold fluxes decreases continuously with increasing CaF₂ content from 3% to 11%. Similar results were also obtained for low basicity mold fluxes by other researchers.^22,23,24) Park et al.²⁵⁾ revealed that CaF₂ addition effectively decreases the viscosity of CaO-SiO₂-10%/40%MnO slags, especially when the MnO content is low. A similar conclusion by Kim et al.²⁶⁾ was also obtained for BaO–SiO₂–MnO–MgO slags containing different MnO contents.

Currently, effect of CaF₂ on the viscosity of highly basic CaO–Al₂O₃–SiO₂–MgO–CaF₂ slags with C/S>4 and Al₂O₃ content in the order of 15%–20%, which could be used for the production of Al-killed duplex stainless steel with high cleanliness demand, has not been reported yet in the open literature. As already well known, excessive CaF₂ addition in metallurgical slags not only accelerates refractory corrosion²⁷⁾ but also causes serious environmental issues.²⁸⁾ Therefore, in this work, we studied the effect of CaF₂ content on the viscosity and refining ability of highly basic CaO-18%Al₂O₃-SiO₂-10%MgO-CaF₂ (C/S=6) slags and compared with the typical CaF₂-free highly basic CaO-30%Al₂O₃-SiO₂-10%MgO (C/S=6) slag which is commonly used for the production of high cleanliness carbon steel, aiming at optimizing CaF₂ content. The temperature range for viscosity measurement is 1753–1873 K. Moreover, solid precipitation in molten slags was mostly neglected in previous studies on slag viscosity, even for highly basic slags.^16,17) In this work, we also focused on the effect of CaF₂ on the precipitation behaviors of the highly basic slags and further made clear the extent of solid precipitation in affecting slag viscosity.

2. Experimental Methods

2.1. Procedures

2.1.1. Viscosity Measurement

The rotating-cylinder method was employed in this work to measure slag viscosity. The schematic of the experimental set-up is shown in Fig. 1. A rotational viscometer (Brookfield DV2T) covered with a sealed chamber was combined with a MoSi₂ resistance-heated vertical tube furnace (inner diameter: 55 mm; length: 420 mm). A Mo spindle with tapered ends (bob diameter: 17 mm; bob length: 35 mm; taper degree: 60°; shaft diameter: 3 mm) was connected to the viscometer via an alumina tube. The furnace temperature was calibrated in the temperature range of 1673–1923 K using a Type-B thermocouple (Pt-30%Rh/Pt-6%Rh) inserted from the top of the furnace tube.

Fig. 1.

Schematic of the experimental set-up for viscosity measurement. (Online version in color.)

Slag samples for viscosity measurement were prepared by mixing reagent-grade CaO, SiO₂, Al₂O₃, MgO and CaF₂ substances. All the pure oxides were tested for high-temperature volatilization by heating them at 1473 K for one hour and the weight losses of the oxides were negligible. Therefore, each oxide mixture of 65 gram was loaded directly into a cylindrical Mo crucible (outer diameter: 35 mm; inner diameter: 30 mm; height: 62 mm) without further high-temperature treatment to minimize F loss. The Mo crucible containing the oxide mixture was fixed tightly in the center of the furnace with three vertical Mo rods, which were connected to the top lid of the furnace, to prevent the crucible from rotating. Before heating the furnace, air in the furnace tube and in the chamber covering the viscometer was purged with a vacuum pump. Then, high purity Ar gas at a flow rate of 0.05 L·min⁻¹ was introduced into the furnace. The heating rates below and above 1473 K were 15 K∙min⁻¹ and 6 K∙min⁻¹, respectively. After holding the furnace at 1873 K for 10 min, the spindle was lowered into the molten slag to start viscosity measurement. The tip of the bob was located at 10 mm above the bottom of crucible and about 5 mm of the shaft was immersed in the slag. Viscosity measurement was performed at every 10 K interval during cooling at 5 K∙min⁻¹ until rotational torque reached around 90% of the maximum value. For each measurement, the furnace was first held at the target temperature for 5 min to stabilize slag temperature. Then, viscosity data was recorded every 10 s during the duration of 3 min.

The initial chemical compositions of the slag samples are given in Table 1. For all slags, the binary basicity C/S is 6 and MgO content is 10%. Moreover, for SF1, SF2, SF3 and SF4 slags, Al₂O₃ content keeps constant at 18% and CaF₂ content increases from 6% to 12%. The SA slag, which is commonly used for Al-killed carbon steel, has 30% Al₂O₃ but no CaF₂. For high measurement accuracy, the viscometer requires a minimum torque, around 10% of the maximum value. Therefore, for SF1, SF2, and SA slags, the rotation speed of the spindle was 20 rpm. For SF3 and SF4 slags with higher CaF₂ contents, the rotation speed was 30 rpm.

Table 1. The chemical compositions of the slags and the flow activation energies.

Slag No.	CaO	SiO2	MgO	Al2O3	CaF2	C/S	EηH	EηL
	(mass%)						(kJ∙mol−1)
SF1	56.6	9.4	10.0	18.0	6.0	6	874.0	365.4
SF2	54.9	9.1	10.0	18.0	8.0	6	226.9	398.7
SF3	53.1	8.9	10.0	18.0	10.0	6	98.2	417.3
SF4	51.4	8.6	10.0	18.0	12.0	6	77.3	345.5
SA	51.4	8.6	10.0	30.0	0.0	6	224.3	564.9

2.1.2. Preparation of Water-quenched Slags

To reveal the high-temperature precipitation of the SF1 and SA slags, water-quenched slags were prepared. The slags (3 g) each wrapped with Mo foil were together suspended in a MoSi₂ resistance-heated vertical tube furnace (inner diameter: 90 mm) with Mo wires at 1853 K, followed by holding there for 30 min under flowing Ar atmosphere (0.8 L·min⁻¹). Then, the temperature was decreased to 1773 K at a cooling rate of 10 K∙min⁻¹. 60 min later, the slags were rapidly taken out and quenched in water.

2.1.3. Refining Experiments

To evaluate the effect of the 6% CaF₂-bearing slag (SF1) on steel cleanliness, deoxidation experiments were performed on a laboratory scale using the above vertical tube furnace. 700 gram of 2507 duplex stainless steel was loaded into a MgO crucible (inner diameter: 55 mm; height: 70 mm) and then placed in the tube furnace at room temperature. The chemical composition of the duplex stainless steel is given in Table 2. Before heating up, air in the furnace tube was purged with a vacuum pump. Then, purified Ar and N₂ gases at respective flow rates of 0.6 L·min⁻¹ and 0.4 L·min⁻¹ were introduced into the furnace. After homogenizing for 30 min at 1853 K, a water-quenched steel sample was taken with a quartz tube (inner diameter: 4 mm) for measuring the initial oxygen content. Then, 35 gram of refining slag (SF1 and SA) was dropped to the molten steel through a small hole in the lid. At 15 min and 30 min after slag addition, water-quenched steel samples were taken with the quartz tubes inserted approximately in the middle of the liquid steel. 35 min later, power supply to the furnace was switched off. At high temperature range above 1673 K, the cooling rate was recorded at around 10 K∙min⁻¹. Note that the refining slags were prepared by mixing reagent-grade CaO, SiO₂, Al₂O₃, MgO and CaF₂ substances, followed by heating the oxide mixtures at 1473 K for one hour for complete dehydration.

Table 2. The chemical composition of 2507 duplex stainless steel before re-melting, mass%.

C	Si	Mn	Cr	Ni	Mo	P	S	Al	Cu	N
0.03	0.49	0.71	25.68	6.40	3.27	0.03	0.001	0.055	0.24	0.27

2.2. Post-test Characterization

Several pieces of the water-quenched slags were mounted in epoxy resin, mechanically ground and polished, dried and finally coated with carbon for compositional and microstructural analysis with an electron probe micro-analyzer (FEG-EPMA JXA-8530F) equipped with EDS and WDS spectrometers.

Three small pieces (around 1 g) were cut from each steel sample for total oxygen content measurement with a LECO combustion analyzer using the inert gas fusion method. The steel samples taken at 30 min after slag addition were further prepared for the characterization of non-metallic inclusions in polished cross sections with a high-resolution scanning electron microscope (SEM, Quanta 250 FEG). For measuring the planar size and number density of the inclusions, successive SEM micrographs were taken at a magnification of 500. The micrographs were then processed with an image analyzer. The planar size of a single inclusion is characterized by its equivalent diameter, which is defined as the diameter of a circle with the same area as the inclusion. Non-metallic inclusions larger than 0.5 μm in diameter were counted in. For each sample, over 150 inclusions were measured.

3. Experimental Results

3.1. Effect of CaF₂ Content on Slag Viscosity

Figure 2 plots the viscosities of the slags with different CaF₂ contents as a function of temperature ranged from 1753–1873 K. The viscosities measured by Song et al.²⁹⁾ for the 55%CaO-30%Al₂O₃-10%SiO₂-5%MgO slag, which has a composition close to the present SA slag (51.43%CaO-30%Al₂O₃-8.57%SiO₂-10%MgO), were also included in the figure for comparison. The two slags have very close viscosities at temperatures above 1843 K, indicating that the viscosities measured in the present work are quite reliable. The distinct viscosities below 1843 K may be attributed to the different basicity and MgO content. As expected, for all the slags, the viscosities decrease with increasing temperature. This trend, however, becomes less significant at high temperatures above 1853 K, except for the slag containing 6% CaF₂. CaF₂ addition significantly decreases slag viscosity. The phenomenon could be observed more clearly in Fig. 3, where the effect of CaF₂ contents on slag viscosities at different temperatures is shown. The effect of CaF₂ addition in decreasing slag viscosity becomes less obvious with increasing temperature and CaF₂ content. At 1773 K, slag viscosity substantially decreases from 0.47 Pa∙s to 0.14 Pa∙s with increasing CaF₂ content from 6% to 12%. At 1873 K, however, slag viscosity only slightly decreases from 0.09 Pa∙s to 0.04 Pa∙s. This phenomenon is consistent with the report by previous researchers.^22,25,30) Moreover, when CaF₂ content exceeds 10%, slag viscosity only marginally decreases with further increasing CaF₂ content, especially at high temperatures. Such a phenomenon was also reported by other researchers but for different slag systems.^17,22,30)

Fig. 2.

Viscosities of the CaO-SiO₂-18%(30%)Al₂O₃-10%MgO-CaF₂ (C/S=6) slags as a function of temperature at different CaF₂ contents. The error bars represent the standard deviations. (Online version in color.)

Fig. 3.

Effect of CaF₂ contents on the viscosities of the CaO-SiO₂-18%Al₂O₃-10%MgO-CaF₂ (C/S=6) slags at different temperatures. (Online version in color.)

Still in Fig. 2, the SF4 slag containing 18% Al₂O₃ and 12% CaF₂ has much lower viscosity than the SA slag containing 30% Al₂O₃ especially at low temperatures, indicating that CaF₂ is much more effective than Al₂O₃ in lowering slag viscosity. This finding is consistent with the study by Park et al.¹⁸⁾ Moreover, all the slags containing CaF₂ content above 8% have much lower viscosity than the SA slag. The viscosity curve of the SF1 slag containing 18% Al₂O₃ and 6% CaF₂ intersects twice with the SA slag at temperatures above 1833 K, where the two slags have close viscosities. Below 1833 K, however, the SF1 slag has much lower viscosities than the SA slag. The SA slag is a typical CaO–Al₂O₃ based highly basic refining slag in carbon steel production.¹⁰⁾ Thus, from the viewpoint of slag viscosity, 6% CaF₂ might be sufficient for the current CaF₂-bearing highly basic refining slag with 18% Al₂O₃. This will be further evaluated experimentally in section 3.4.

Figure 4 shows the viscosities (η) of the slags in natural logarithmic form (lnη) versus reciprocal temperature (10000/T) at different CaF₂ contents. For all slags, the data points could be accurately fitted with two linear segments by the method of least-squares within the investigated temperature range and the transition temperatures are roughly around 1853 K. Above 1853 K, lnη increases slowly with 10000/T except for the SF1 slag, indicating that the slag viscosities are not very sensitive with high temperatures. Below 1853 K, lnη increases rapidly with 10000/T. These results are consistent with those in Fig. 2. The phenomenon that viscosity lnη versus 1/T could be fitted with several linear segments was also reported for other slag systems.^31,32,33) The linear relationship between lnη and 1/T indicates that the dependence of slag viscosity on temperature could be described with the Arrhenius viscosity model expressed with the below equation.³⁴⁾   

$$\eta ={\eta }_{0}exp\left(\frac{E_{\eta }}{RT}\right)$$(1)

where η₀ is the pre-exponential constant, E_η is the activation energy, R is the gas constant and T is the absolute temperature. The activation energy represents the energy barrier for viscous flow and could be calculated from the slope of the fitted line on a plot of lnη versus 1/T. Note that the activation energy is temperature dependent.³⁵⁾ Only within a narrow temperature range, the activation energy could be regarded as constant. This explains the variation of the activation energy over the wide temperature range in this work. The activation energies for the high and low temperature ranges were listed in Table 1 as E_ηH and E_ηL, respectively. As seen, at high temperatures, the activation energy decreases with increasing CaF₂ content, again indicating that CaF₂ addition favors slag flowing. At low temperatures, the activation energies of the CaF₂-bearing slags fluctuate slightly. The SA slag has the highest activation energy.

Fig. 4.

Natural logarithm of the viscosities for the CaO-SiO₂-18%(30%)Al₂O₃-10%MgO-CaF₂ (C/S=6) slags versus reciprocal temperature. (Online version in color.)

3.2. Precipitation Behaviors of the Slags

Normally, solid phases will be precipitated from molten slag during cooling, thus increasing slag viscosity. Therefore, precipitation behaviors of the SF1 slag and the SA slag, which have very close viscosities at high temperatures, were investigated both theoretically and experimentally for comparison. Figure 5 presents the relative amounts of equilibrium phases in the slags as a function of temperature, calculated with the thermodynamic software FactSage 7.2 using the FToxid and FactPS databases. For the SF1 slag, two solid phases, i.e. monoxide-CaO phase (denoted as m-CaO) containing more than 96.7% CaO and monoxide-MgO phase (denoted as m-MgO) containing more than 99.2% MgO, already present at 1873 K, coexisting with the liquid oxyfluoride phase (55.7%CaO-22.1%Al₂O₃-11.6%SiO₂-3.2%MgO-7.4%CaF₂). The presence of the m-CaO and m-MgO phases indicates that CaO and MgO contents in the SF1 slag are oversaturated. The m-CaO and m-MgO phase fractions are 11.7% and 7.1% at 1873 K, respectively. Their phase fractions slightly increase in a near-linear manner upon cooling from 1873 K until around 1598 K, where Ca₁₂Al₁₄F₂O₃₂ and Ca₂SiO₄ phases are precipitated. For the SA slag, only m-MgO phase, accounting for 4.0%, equilibrates with the liquid oxide phase (53.6%CaO-31.3%Al₂O₃-8.9%SiO₂-6.2%MgO) at 1873 K. Also, the m-MgO phase fraction slightly increases upon cooling from 1873 K. Ca₂SiO₄ and Ca₃Al₂O₆ phases are precipitated at 1692 K and 1638 K, respectively. Compared with the SF1 slag, no m-CaO phase is precipitated in the SA slag. Also, m-MgO fraction in the SA slag is approximately 2.8% lower within the temperature range from 1773 K to 1873 K. Therefore, the SA slag contains approximately 15.0% more liquid phase than the SF1 slag at temperatures above 1773 K.

Fig. 5.

Precipitation behaviors of the slags with temperature. (a) SF1 slag; (b) SA slag. (Online version in color.)

To verify the above thermodynamic calculations, precipitation behaviors of the SF1 and SA slags were investigated experimentally. Figure 6 shows the microstructures of the SF1 and SA slags which were water-quenched from 1773 K. Table 3 shows the chemical compositions of all phases observed in the slags. In the SF1 slag, the m-CaO and m-MgO phases were identified, exhibiting block-like morphology. Their chemical compositions are very closed to the calculated values. Also, a white dendritic-like phase (indicated with +3) was occasionally observed in the matrix, indicating that this phase might be formed during cooling. Composition analysis shows that the dendritic-like phase contains high amounts of SiO₂ and CaF₂. This is different from the thermodynamic calculation in Fig. 5(a), where SiO₂ and CaF₂ components appear in two separate phases, i.e. the Ca₂SiO₄ and Ca₁₂Al₁₄F₂O₃₂ phases. In the SA slag, only m-MgO phase was identified. This is consistent with the thermodynamic calculation in Fig. 5(b).

Fig. 6.

Microstructure of the slags quenched at 1773 K. (a) SF1 slag; (b) SA slag. (Online version in color.)

Table 3. Chemical compositions of the phases indicated by plus sign in Fig. 6 and their phase fractions, mass%.

Phases	CaO	SiO2	MgO	Al2O3	CaF2	Note	Measured fraction	Calculated fraction
1	95.3	1.4	1.6	1.7	–	m-CaO	11.3	12.7
2	0.9	–	99.1	–	–	m-MgO	7.4	7.9
3	62.8	17.1	2.0	14.6	3.5	Nonequilibrium	/	/
4	56.3	12.0	3.3	21.5	6.9	Matrix	81.3	79.4
5	51.9	9.8	6.7	31.6	–	Matrix	96.5	94.6
6	0.7	–	99.3	–	–	m-MgO	3.5	5.4

According to the lever rule, phase fractions of the m-CaO, m-MgO and matrix phases in the SF1 and SA slags could be determined using the measured (CaO) and MgO contents in relevant phases listed in Table 3. Note that for the SA slag, only MgO contents in relevant phases are required. Both the calculated and measured phase fractions are given in Table 3 for comparison. As seen, the measured values are quite close to the calculated values, indicating that the thermodynamic calculations are reliable at least for the high-temperature precipitates.

Figure 7 plots the calculated fractions of m-CaO and m-MgO phases versus CaF₂ content at 1773 K and 1873 K. Note that both m-CaO and m-MgO phases are precipitated in all the CaF₂-bearing slags. CaF₂ slightly promotes m-MgO precipitation but dramatically suppresses m-CaO precipitation. Take 1873 K as an example, increasing CaF₂ content from 6% to 12%, m-MgO fraction linearly increases from 7.1% to 8.0%, while m-CaO fraction linearly decreases from 11.7% to 4.6%. Temperature has insignificant effect on the precipitation of the m-CaO and m-MgO phases. Compared with 1873 K, m-CaO and m-MgO fractions at 1773 K are approximately 1.5% and 0.7% higher at all CaF₂ contents.

Fig. 7.

Effect of CaF₂ contents on the phase fractions of the CaO-SiO₂-18%Al₂O₃-10%MgO-CaF₂ (C/S=6) slags at different temperatures. (Online version in color.)

3.3. Modelling of Slag Viscosity

The viscosities of the CaF₂-bearing slags were calculated with the aid of the thermodynamic software FactSage 7.2 using the Equilib and Viscosity modules, and compared with the measured values. The Viscosity module directly relates the viscosity to the structure of the melt, and the structure is calculated from the thermodynamic description of the melt using the Modified Quasichemical Model.³⁶⁾ Note that the Viscosity module is applicable only for single-phase liquid slags. Presence of solid particles significantly affects the viscosity of the liquid phase containing the solid particles.³⁷⁾ Therefore, many equations relating the viscosities of pure liquid phase and solid-liquid mixture phases have been proposed.³⁸⁾ According to Kondratiev et al.³⁹⁾ who evaluated a number of viscosity correlations for partially crystalized slags, the Einstein-Roscoe equation⁴⁰⁾ expressed below provides good fit into experimental results.   

$$\eta ={\eta }_{\text{L}}{(1-mf)}^{-n}$$(2)

where η is the viscosity of the mixed solid-liquid phases, η_L is the viscosity of the liquid phase, f is the particle volume fraction, m and n are empirical parameters which equal to 1.35 and 2.5, respectively, for rigid spheres of equal size.³⁹⁾ According to Kondratiev et al.,³⁹⁾ m (2.04) and n (1.29) obtained by fitting predictions into experimental data provide good fit for slags with volume fraction of solids up to 30 vol%. Therefore, 2.04 and 1.29 for m and n respectively were used in the present calculations. To calculate the viscosities of the CaF₂-bearing slags containing solid precipitates, chemical composition of the liquid slag phase and mass fraction of the solid precipitates at investigated temperatures were first calculated with the Equilib module in the FactSage 7.2 using the FToxid and FactPS databases. Then, viscosity of the liquid slag phase was calculated with the Viscosity module. Finally, the Einstein-Roscoe equation was used to calculate the viscosity of the partially crystalized slags. For simplicity, the volume fraction of the precipitates in the CaF₂-bearing slags is assumed to be equal to their mass fraction. Moreover, the performance of the viscosity modelling through FactSage and Einstein-Roscoe equation was evaluated by the mean deviation between measured and calculated viscosities, as expressed below.³⁶⁾   

$$\mathrm{\Delta }=\frac{1}{X}\sum \frac{\left|{\eta }_{\text{Md}}-{\eta }_{\text{Cd}}\right|}{{\eta }_{\text{Md}}}\cdot 100\text{\%}$$(3)

where Δ is the mean deviation between measured and calculated viscosities, η_Md and η_Cd are measured and calculated viscosities respectively, X is the total number of data pairs for each slag. In this work X equal to 11 as 11 data pairs from 1773 K to 1873 K were compared for each slag.

Figure 8 shows the calculated viscosities of the CaF₂-bearing slags in the temperature range of 1773–1873 K, together with the measured values for comparison. As seen, the calculated viscosities, which slowly increase with decreasing temperature, are much less sensitive with temperature for all slags, compared with the measured viscosities. Moreover, for the SF1 slag, the measured viscosities are larger than the calculated viscosities at all investigated temperatures except 1873 K, and the difference increases with decreasing temperature. On the contrary, for the SF4 slag, the measured viscosities are smaller than the calculated viscosities at all temperatures except 1773 K, and the difference decreases with decreasing temperature. For the SF2 and SF3 slags, the calculated viscosity curves intersect with the measured viscosity curves around 1833 K and 1803 K respectively, making the calculated values much closer to the measured viscosities compared with the SF1 and SF4 slags.

Fig. 8.

Comparison of calculated and measured viscosities of the CaF₂-bearing slags in the temperature range of 1773–1873 K. (Online version in color.)

Figure 9 shows the mean deviation between the calculated and measured viscosities of the CaF₂-bearing slags. The mean deviation first sharply decreases then gradually increases with increasing CaF₂ content. The SF1 slag containing 6% CaF₂ has the largest mean deviation of 38.4%, while the SF2 slag containing 8% CaF₂ has the smallest mean deviation of 20.2%. The mean deviations in this work are in good agreement with the report by Rocha et al.³⁶⁾ that the mean deviation between the calculated viscosities through the FactSage software and the measured viscosities lies in the range of 13.31%–37.53% for 162 slags in total. Note that in the viscosity calculations by Rocha et al.³⁶⁾ solid precipitates in all slags were not considered and thus no viscosity correlations for partially crystalized slags were used. According to Jung,³⁷⁾ viscosity values for the same samples when measured by various research groups in the world can easily differ by 20%–50%. The mean deviations between the calculated and measured viscosities for all slags are within this range. Therefore, it could be concluded that the calculated viscosities agree well with the measured values for the current slags.

Fig. 9.

The mean deviation between the calculated and measured viscosities of the CaF₂-bearing slags.

3.4. Evaluation of the 6% CaF₂-bearing Slag on Steel Cleanliness

The effect of the SF1 slag containing 6% CaF₂ on the cleanliness of Al-killed 2507 duplex stainless was evaluated on a laboratory scale and compared with the SA slag. Total oxygen content in steel is often used to represent steel cleanliness in terms of oxide inclusions. Therefore, variations of total oxygen contents in 2507 duplex stainless steel refined with SF1 and SA slags were measured before and after slag addition and shown in Fig. 10. Before slag addition, the two liquid steel have very close initial oxygen contents of around 50 ppm. After slag addition, total oxygen contents decrease with time for both slags. In the final ingots, total oxygen contents are 24 ppm and 26 ppm respectively for the SF1 and SA slags, showing no significant difference. Therefore, the two slags have almost the same deoxidation ability.

Fig. 10.

Evolution of total oxygen contents in 2507 duplex stainless steel refined with SF1 and SA slags after slag addition. (Online version in color.)

Two types of non-metallic inclusions with very distinct morphologies were observed at 30 min after slag addition for both the SF1 and SA slags, i.e. square/rhombus and near-spherical morphologies, as representatively shown in Fig. 11. Square/rhombus inclusions dominate largely the morphologies. The different morphologies indicate their different chemical compositions. Figure 12 shows the chemical compositions of the non-metallic inclusions at 30 min after slag addition. Note that ten square/rhombus inclusions and ten near-spherical inclusions were analyzed for each sample. The square/rhombus inclusions mainly consist of Al₂O₃ and MgO, while the near-spherical inclusions consist of Al₂O₃, MgO and CaO. Average MgO content in the non-metallic inclusions when refined with the SF1 slag is approximately 5% higher, compared with the SA slag. The transformation of Al₂O₃ inclusions into Al₂O₃–MgO and Al₂O₃–MgO–CaO inclusions is a common phenomenon in Al-killed steel refined with highly basic slag even without direct Mg and Ca treatment.⁴¹⁾ Formation of the Al₂O₃–MgO and Al₂O₃–MgO–CaO inclusions arises from the simultaneous reduction of MgO and CaO in highly basic slag by soluble Al, supplying soluble Mg and Ca into liquid steel. Then, the soluble Mg and Ca react with Al₂O₃ inclusions, forming Al₂O₃–MgO and Al₂O₃–MgO–CaO inclusions.

Fig. 11.

Representative non-metallic inclusions in 2507 duplex stainless steel at 30 min after slag addition. (a) square/rhombus morphology; (b) near-spherical morphology.

Fig. 12.

The chemical compositions of analyzed non-metallic inclusions in 2507 duplex stainless steel at 30 min after slag addition. (Online version in color.)

Table 4 lists the summary of the characteristics of the non-metallic inclusions in steel samples taken at 30 min after slag addition. As seen, the two slags result in slightly different inclusion characteristics. The steel refined with the SF1 slag has a slightly larger mean diameter but a smaller number density than the SA slag. For both slags, the largest inclusions observed are smaller than 5 μm in size. Figure 13 gives the size distribution of the non-metallic inclusions in the same steel samples. The steel refined with the SF1 slag has slightly fewer fine inclusions but more coarse inclusions. Specifically, for the SF1 slag, 63.7% inclusions are below 2 μm and 12.3% inclusions are over 3 μm. For the SA slag, non-metallic inclusions falling into the above two size groups are 75.8% and 4.0%, respectively.

Table 4. Characteristics of non-metallic inclusions in 2507 duplex stainless steel samples taken at 30 min after the addition of SF1 and SA slags.

Sample No.	Mean diameter	Standard deviation	Largest inclusions observed	Area fraction	Number density
	(μm)	(μm)	(μm)	(%)	(mm−2)
SF1-3	1.69	0.98	4.7	0.018	59
SA-3	1.42	0.61	4.4	0.017	83

Fig. 13.

Comparison of inclusion size distributions at 30 min after slag addition in 2507 duplex stainless steel refined with SF1 and SA slags. (Online version in color.)

4. Discussion

In this work, we observed that CaF₂ addition obviously decreases the viscosities of the highly basic refining slags especially at low temperatures. This effect, however, weakens greatly when CaF₂ content exceeds 10%. Such a phenomenon is consistent with previous studies.^{16,17,18,19,20,21,22,23,24,25,26)} Physical properties of slags including viscosity intrinsically depend on ionic structure of molten slags. Therefore, to clarify the mechanism of fluorine (F) ions in decreasing slag viscosity, various spectroscopic instruments, such as nuclear magnetic resonance (NMR) spectroscopy,⁴²⁾ Raman spectroscopy,^25,30) X-ray photoelectron spectroscopy (XPS)^22,43) and Fourier transform infrared (FTIR) spectroscopy,^44,45) as well as molecular dynamics simulation method^45,46,47) have been used to explore the effect of F ions on the structural changes of molten slags. In SiO₂ and Al₂O₃ containing slags, silicate and aluminate networks exist. Regarding the effect of F ions on silicate network, currently there are two opposite opinions. Kiczenski et al.,⁴²⁾ Hayashi et al.,⁴³⁾ Sasaki et al.⁴⁸⁾ and other researchers^45,46) claimed that F ions preferentially coordinate to Ca or Na cations but not Si ions, forming Ca-2F or Na-F complex. Consequently, polymerization of the slags is enhanced. To decrease the viscosity of the slags, F ions act as a diluent. On the contrary, Kim et al.,^22,23) Park et al.,^25,30) and Deng et al.²⁰⁾ reported that F ions coordinate to Si ions, thus decreasing the polymerization of the slags by breaking silicate network. Kalisz⁴⁹⁾ pointed out that the different behaviors of F ions may be due to the different chemical compositions of the slags under consideration, mainly the basicity of the slags and the presence of other structure modifiers. However, opposite conclusions regarding the polymerization of F-containing slags with very close basicity were obtained, when analysed with different methods, i.e. XPS⁴³⁾ and Raman spectroscopy.²²⁾ Moreover, we noticed that different interpretation methods regarding Raman spectra were used by different researchers,^25,30,45,48) and thus leading to opposite conclusions. For example, Sasaki et al.⁴⁸⁾ found no new bands appear in the Raman spectra after NaF addition and concluded that F ions may not coordinate to Si ions. An opposite conclusion was obtained by Park et al.^25,30) after analysing the Q₃/Q₂ ratio (Si₂O₅-sheet for Q₃ and SiO₃-chain for Q₂) versus CaF₂ content. If the later interpretation method was adopted in the study by Sasaki et al.,⁴⁸⁾ the conclusion might be reversed, based on Sasaki’s result in Fig. 13 where distribution of Q_i as a function of NaF was given. Currently, information on F coordination to Ca/Na or Si ions are indirectly obtained for all spectroscopic instruments. The second claim that F ions coordinate to Si ions seems more successful. This is because that the variations of bridging, non-bridging and free oxygen ions from XPS spectra²²⁾ as well as Q₃/Q₂ ratio from Raman spectra^25,30) versus CaF₂ content are in good agreement with that of viscosity. In other words, the second claim could well explain why effect of CaF₂ on slag viscosity becomes less obvious at high contents.

Regarding the effect of F ions on aluminate network, Park et al.⁴⁴⁾ concluded that F ions decrease the polymerization of CaO–Al₂O₃–CaF₂ slags by forming [AlO_nF_4-n] units via analyzing the FTIR spectra of the slags. Such a conclusion was also obtained by Zhang et al.⁴⁶⁾ via molecular dynamics simulation. In the present work, aluminate network may dominate as Al₂O₃ content is much higher than SiO₂. However, according to the work by Kim et al.¹⁷⁾ who analyzed the Raman spectra of low-silica highly basic aluminosilicate slags, the composition of which is very close to that in the present work, the polymerized aluminate network is slightly modified by CaF₂. The decreased viscosity is mainly due to the liberation of [SiO₄] (Q₀) units from the [AlO₄] aluminate network. At high CaF₂ content F ions simply substitute for the non-bridging O ions in [AlO₄] unit. This explains why slag viscosity only marginally decreases when CaF₂ content exceeds 10% in the present work.

CaF₂ addition was observed to suppress solid precipitation in this work. As mentioned above, presence of solid particles significantly affects the viscosity of the liquid phase containing the solid particles.³⁷⁾ Therefore, to identify which aspect plays a dominant role in decreasing the viscosity of current CaF₂-bearing highly basic slags, η_Lx/η_L6 and (1−2.04f_x)^−1.29/(1−2.04f₆)^−1.29 ratios were calculated and plotted in Fig. 14. η_Lx and η_L6 are the viscosities of the liquid phases in CaF₂-bearing slags with CaF₂ contents at a given value and at 6%, respectively. f_x and f₆ are the fractions of solid precipitates in CaF₂-bearing slags with CaF₂ contents at a given value and at 6%, respectively. The values of η_Lx and η_L6 could be obtained via Eq. (2) using measured viscosities and calculated solid fractions through the FactSage software. Therefore, from the η_Lx/η_L6 ratio, we could see the extent of CaF₂ in decreasing the viscosity of liquid phase at a given CaF₂ content, compared with 6% CaF₂. Similarly, from the (1−2.04f_x)^−1.29/(1−2.04f₆)^−1.29 ratio, we could see the extent of CaF₂ in decreasing slag viscosity by suppressing solid precipitation at a given CaF₂ content, compared with 6% CaF₂. As seen in Fig. 14, η_Lx/η_L6 ratio significantly decreases with increasing CaF₂ content, especially at low temperature of 1773 K. On the contrary, (1−2.04f_x)^−1.29/(1−2.04f₆)^−1.29 ratio slightly decreases with increasing CaF₂ content. Moreover, (1−2.04f_x)^−1.29/(1−2.04f₆)^−1.29 ratio is insensitive to temperature. Therefore, we could conclude that CaF₂ addition obviously decreases slag viscosity mainly because CaF₂ behaves as a network breaker, not because CaF₂ suppresses solid precipitation. For all the five slags, the fraction of solid precipitation increases only slightly upon cooling from 1873 K to the temperature where massive precipitation occurs (see Figs. 5 and 7). This may be the reason why the dependence of slag viscosity on temperature could be described with the Arrhenius viscosity model, which is mainly applicable for pure liquid slag.

Fig. 14.

Effect of CaF₂ content on the ratio of η_Lx/η_L6 or (1−2.04f_x)^−1.29/(1−2.04f₆)^−1.29. (Online version in color.)

The SF1 slag containing 6% CaF₂ has very close viscosities above 1833 K but much lower viscosities below 1833 K, compared with the CaF₂-free SA slag. Further evaluation of the SF1 slag on steel cleanliness confirms that 6% CaF₂ is sufficient for the CaF₂-bearing highly basic slag. Moreover, the fraction of m-MgO precipitation is around 4% higher in the SF1 slag, indicating lower MgO solubility in the SF1 slag. Therefore, the driving force for the corrosion of MgO-based refractory may be much lower in the SF1 slag. The SF1 slag also has a high amount of m-CaO precipitation over 10%. This greatly deteriorates slag fluidity. Therefore, basicity of the SF1 slag should be optimized to suppress m-CaO precipitation.

5. Conclusions

In this work, CaF₂ content in highly basic CaO-18%Al₂O₃-SiO₂-10%MgO-CaF₂ (C/S=6) refining slag used for Al-killed duplex stainless steel with high cleanliness demand was optimized mainly from the view point of slag viscosity. The slag with optimal CaF₂ content was further evaluated on precipitation behaviors upon cooling and steel cleanliness. The results obtained are summarized as follows:

(1) CaF₂ addition obviously decreases slag viscosity. This effect, however, becomes less obvious with increasing temperature and CaF₂ content. When CaF₂ content exceeds 10%, slag viscosity only marginally decreases with further increasing CaF₂ content. Moreover, CaF₂ is much more effective in decreasing slag viscosity than Al₂O₃.

(2) Thermodynamic calculations show that both monoxide-CaO and monoxide-MgO phases are precipitated in all the CaF₂-bearing slags. With increasing CaF₂ content from 6% to 12%, CaF₂ addition slightly increases monoxide-MgO precipitation fraction, at 1873 K for example, from 7.1% to 8.0%, but dramatically decreases monoxide-CaO precipitation fraction from 11.7% to 4.6%. The calculated results were verified with measured values for the 6% CaF₂-bearing slag at 1773 K and good agreement was observed.

(3) Viscosities of the CaF₂-bearing slags were theoretically calculated. The mean deviation between the calculated and measured viscosities varies between 20.2% and 38.4%, indicating good agreement.

(4) The 6% CaF₂-bearing slag has very close viscosities above 1833 K but much lower viscosities below 1833 K, compared with the CaF₂-free CaO-30%Al₂O₃-SiO₂-10%MgO (C/S=6) slag. Further evaluation of the 6% CaF₂-bearing slag on steel cleanliness confirms that 6% CaF₂ is sufficient for the CaF₂-bearing highly basic slag.

(5) CaF₂ not only behaves as a network breaker also suppresses solid precipitation, thus significantly decreasing slag viscosity. The first aspect was identified to play a much greater role in decreasing slag viscosity.

Acknowledgement

This work was financially supported by National Natural Science Foundation of China (No. 51904067), China Postdoctoral Science Foundation (No. 2019M651127) and Fundamental Research Funds for the Central Universities (NO. N2025039).

References

• 1)   M. C.  Tanzi,  S.  Farè and  G.  Candiani: Foundations of Biomaterials Engineering, Academic Press, London, (2019), 253. https://doi.org/10.1016/B978-0-08-101034-1.00003-7
• 2)   L. F.  Garfias-Mesias,  J. M.  Sykes and  C. D. S.  Tuck: Corros. Sci., 38 (1996), 1319. https://doi.org/10.1016/0010-938X(96)00022-4
• 3)   A.  Iza-Mendia,  A.  Pinol-Juez,  J. J.  Urcola and  I.  Gutierrez: Metall. Mater. Trans. A, 29 (1998), 2975. https://doi.org/10.1007/s11661-998-0205-z
• 4)   Y.  Zhao,  Y.  Wang,  S.  Tang,  W.  Zhang and  Z.  Liu: J. Mater. Process. Technol., 266 (2019), 246. https://doi.org/10.1016/j.jmatprotec.2018.11.010
• 5)   P. A.  Thornton: J. Mater. Sci., 6 (1971), 347. https://doi.org/10.1007/PL00020378
• 6)   M. A.  Baker and  J. E.  Castle: Corros. Sci., 34 (1993), 667. https://doi.org/10.1016/0010-938X(93)90279-P
• 7)   Z.  Deng,  M.  Zhu,  B.  Zhong and  Y.  Dai: J. Iron Steel Res. Int., 20 (2013), 21. https://doi.org/10.1016/S1006-706X(13)60051-5
• 8)   J. G.  Kang,  J. H.  Shin,  Y.  Chung and  J. H.  Park: Metall. Mater. Trans. B, 48 (2017), 2123. https://doi.org/10.1007/s11663-017-0948-2
• 9)   M.  Jiang,  X.  Wang,  B.  Chen and  W.  Wang: ISIJ Int., 50 (2010), 95. https://doi.org/10.2355/isijinternational.50.95
• 10)   B.-H.  Yoon,  K.-H.  Heo,  J.-S.  Kim and  H.-S.  Sohn: Ironmaking Steelmaking, 29 (2002), 214. https://doi.org/10.1179/030192302225004160
• 11)   W.-J.  Ma,  Y.-P.  Bao,  M.  Wang and  D. W.  Zhao: Ironmaking Steelmaking, 41 (2014), 26. https://doi.org/10.1179/1743281212Y.0000000096
• 12)   Y.  Ji,  C.  Liu,  Y.  Lu,  H.  Yu,  F.  Huang and  X.  Wang: Metall. Mater. Trans. B, 49 (2018), 3127. https://doi.org/10.1007/s11663-018-1397-2
• 13)   J.-Y.  Choi,  H.-G.  Lee and  J.-S.  Kim: ISIJ Int., 42 (2002), 852. https://doi.org/10.2355/isijinternational.42.852
• 14)   M.  Valdez,  G. S.  Shannon and  S.  Sridhar: ISIJ Int., 46 (2006), 450. https://doi.org/10.1179/1743281212Y.0000000096
• 15)   A. P. L.  Berger,  R.  de Oliveira Pezzin,  F. F.  Grillo,  E.  Junca,  H. S.  Furtado and  J. R.  de Oliveira: J. Mater. Res. Technol., 8 (2019), 4651. https://doi.org/10.1016/j.jmrt.2019.08.009
• 16)   L.  Wu,  J.  Gran and  D.  Sichen: Metall. Mater. Trans. B, 42 (2011), 928. https://doi.org/10.1007/s11663-011-9546-x
• 17)   T. S.  Kim and  J. H.  Park: ISIJ Int., 54 (2014), 2031. https://doi.org/10.2355/isijinternational.54.2031
• 18)   J. H.  Park and  D. J.  Min: J. Non-Cryst. Solids, 337 (2004), 150. https://doi.org/10.1016/j.jnoncrysol.2004.03.109
• 19)   J. A. M.  Pereira,  V. C.  da Rocha,  A.  Yoshioka,  W. V.  Bielefeldt and  A. C. F.  Vilela: Mat. Res., 21 (2018), e20180296. https://doi.org/10.1590/1980-5373-mr-2018-0296
• 20)   L.  Deng,  X.  Zhang,  M.  Zhang and  X.  Jia: J. Non-Cryst. Solids, 500 (2018), 310. https://doi.org/10.1016/j.jnoncrysol.2018.08.018
• 21)   X.  Wang,  H.  Jin,  L.  Zhu,  Y.  Xu,  R.  Liu,  Z.  Piao and  S.  Qu: Metals, 9 (2019), 871. https://doi.org/10.3390/met9080871
• 22)   H.  Kim and  I.  Sohn: ISIJ Int., 51 (2011), 1. https://doi.org/10.2355/isijinternational.51.1
• 23)   G. H.  Kim and  I.  Sohn: J. Am. Ceram. Soc., 102 (2019), 6575. https://doi.org/10.1111/jace.16526
• 24)   J.  Yang,  J.  Zhang,  O.  Ostrovski,  C.  Zhang and  D.  Cai: Metall. Mater. Trans. B, 50 (2019), 1766. https://doi.org/10.1007/s11663-019-01579-z
• 25)   J. H.  Park,  K. Y.  Ko and  T. S.  Kim: Metall. Mater. Trans. B, 46 (2015), 741. https://doi.org/10.1007/s11663-014-0269-7
• 26)   T. S.  Kim and  J. H.  Park: J. Am. Ceram. Soc., 102 (2019), 4943. https://doi.org/10.1111/jace.16363
• 27)   J. S.  Han,  J. G.  Kang,  J. H.  Shin,  Y.  Chung and  J. H.  Park: Ceram. Int., 44 (2018), 13197. https://doi.org/10.1016/j.ceramint.2018.04.145
• 28)   K. T.  Semrau: J. Air Pollut. Control Assoc., 7 (1957), 92. https://doi.org/10.1080/00966665.1957.10467795
• 29)   M.  Song,  Q.  Shu and  D.  Sichen: Steel Res. Int., 82 (2011), 260. https://doi.org/10.1002/srin.201000150
• 30)   K. Y.  Ko and  J. H.  Park: ISIJ Int., 53 (2013), 958. https://doi.org/10.2355/isijinternational.53.958
• 31)   Z.  Huaiwei,  S.  Fei,  S.  Xiaoyan,  Z.  Bo and  H.  Xin: Metall. Mater. Trans. B, 43 (2012), 1046. https://doi.org/10.1007/s11663-012-9698-3
• 32)   Y.  Harada,  H.  Yamamura,  Y.  Ueshima,  T.  Mizoguchi,  N.  Saito and  K.  Nakashima: ISIJ Int., 58 (2018), 1285. https://doi.org/10.2355/isijinternational.ISIJINT-2018-009
• 33)   M. S.  Seo and  I.  Sohn: J. Am. Ceram. Soc., 102 (2019), 6275. https://doi.org/10.1111/jace.16456
• 34)   H. R.  Shaw: Am. J. Sci., 272 (1972), 870. https://doi.org/10.2475/ajs.272.9.870
• 35)   D. B.  Dingwell: Treatise on Geophysics, Vol. 2, ed. by G. Schubert, Elsevier, Amsterdam, (2015), 479. https://doi.org/10.1016/B978-0-444-53802-4.00043-9
• 36)   V. C.  da Rocha,  M. L.  da Silva,  W. V.  Bielefeldt and  A. C. F.  Vilela: Rem: Int. Eng. J., 71 (2018), 243. https://doi.org/10.1590/0370-44672017710029
• 37)   I.-H.  Jung: Calphad, 34 (2010), 332. https://doi.org/10.1016/j.calphad.2010.06.003
• 38)   Z.  Liu,  L.  Pandelaers,  B.  Blanpain and  M.  Guo: Metall. Mater. Trans. B, 49 (2018), 2469. https://doi.org/10.1007/s11663-018-1374-9
• 39)   A.  Kondratiev and  E.  Jak: Metall. Mater. Trans. B, 32 (2001), 1027. https://doi.org/10.1007/s11663-001-0091-x
• 40)   R.  Roscoe: Br. J. Appl. Phys., 3 (1952), 267. https://doi.org/10.1088/0508-3443/3/8/306
• 41)   J. H.  Park and  H.  Todoroki: ISIJ Int., 50 (2010), 1333. https://doi.org/10.2355/isijinternational.50.1333
• 42)   T. J.  Kiczenski and  J. F.  Stebbins: J. Non-Cryst. Solids, 306 (2002), 160. https://doi.org/10.1016/S0022-3093(02)01157-2
• 43)   M.  Hayashi,  N.  Nabeshima,  H.  Fukuyama and  K.  Nagata: ISIJ Int., 42 (2002), 352. https://doi.org/10.2355/isijinternational.42.352
• 44)   J. H.  Park,  D. J.  Min and  H. S.  Song: ISIJ Int., 42 (2002), 38. https://doi.org/10.2355/isijinternational.42.38
• 45)   S.  He,  S.  Wang,  B.  Jia,  M.  Li,  Q.  Wang and  Q.  Wang: Metall. Mater. Trans. B, 50 (2019), 1503. https://doi.org/10.1007/s11663-019-01547-7
• 46)   X.  Zhang,  C.  Liu and  M.  Jiang: ISIJ Int., 60 (2020), 2176. https://doi.org/10.2355/isijinternational.ISIJINT-2020-002
• 47)   Y.  Sasaki,  M.  Iguchi and  M.  Hino: ISIJ Int., 47 (2007), 638. https://doi.org/10.2355/isijinternational.47.638
• 48)   Y.  Sasaki,  H.  Urata and  K.  Ishii: ISIJ Int., 43 (2003), 1897. https://doi.org/10.2355/isijinternational.43.1897
• 49)   D.  Kalisz: Arch. Metall. Mater., 59 (2014), 149. https://doi.org/10.2478/amm-2014-0236