Phase Identification of Crystal Precipitated from Molten CaO–SiO₂–FeO_x–P₂O₅ Slag by High Temperature In-situ X-ray Diffraction

Abstract

For the first time, we have succeeded in directly identifying the crystalline phase precipitated from the fully liquid slag of the CaO–SiO₂–FeO_x–P₂O₅ system by high-temperature in-situ X-ray diffraction analysis. Dephosphorization from molten iron can be promoted by 2CaO·SiO₂ precipitates in molten P₂O₅-containing slag because they form a solid solution with 3CaO·P₂O₅. Knowledge of the crystal structure of the 2CaO·SiO₂ precipitate is important because it strongly affects the phosphorus solubility. Although it is believed that the α phase of the 2CaO·SiO₂–3CaO·P₂O₅ solid solution precipitates from the molten slag, the crystal structure of the precipitate has not been identified because the crystal structure of the 2CaO·SiO₂ compound rapidly changes by phase transition when cooled from high temperature. In this study, slag samples were aerodynamically levitated and completely melted by laser heating under an Ar atmosphere, and then the diffraction patterns were obtained by transmitting a high-energy and high-intensity X-ray beam into the molten slag. We verified that the α-2CaO·SiO₂ phase precipitated as the primary crystal phase from molten slag containing 10–30 mass% FeO_x and 5 mass% P₂O₅, whereas nagelschmidtite precipitated for the molten slag with high P₂O₅ content. The α-2CaO·SiO₂ precipitates contained much higher FeO_x content than the reported solubility limit, which was supported by the diffraction angles positively deviated from those of the FeO_x-free α phase in the CaO–SiO₂–P₂O₅ system and chemical analysis of the quenched slag sample. This excess FeO_x solute may influence the phosphorus distribution in the α-2CaO·SiO₂ precipitates.

1. Introduction

Dephosphorization is essential in iron and steelmaking because even a small amount of phosphorus remaining in cast iron significantly degrades the mechanical properties. At present, dephosphorization of molten iron is performed by oxidizing the phosphorus solute by injected oxygen gas and distributing phosphorus oxide (e.g., P₂O₅) into the molten slag. A lot of fundamental studies have been performed to understand the dephosphorization mechanisms, as well as to evaluate the phosphorus partitioning in various types of slag.^{1,2,3,4,5,6,7,8,9,10,11,12,13)} Since the use of calcium fluoride as an additive was restricted, dephosphorization by multi-phase slag (solid–liquid coexistent state) has been performed.^{14,15,16,17,18,19,20,21,22,23)} In this case, the dicalcium silicate (2CaO·SiO₂, C2S) solid phase precipitates in the molten slag and phosphorus oxide is known to selectively dissolve into the C2S compound as tricalcium phosphate (3CaO·P₂O₅, C3P) to form a solid solution. Thus, the phosphorus solubility in the C2S precipitates is of great importance to enhance the dephosphorization efficiency.

It is well known that the C2S compound exhibits different polymorphs. The α (hexagonal) and α′ (orthorhombic) phases are stable at high temperature, the β (monoclinic) phase is a metastable phase, and the γ (orthorhombic) phase is stable at low temperature, all of which consist of isolated Ca²⁺ and SiO₄⁴⁻ tetrahedrons.^24,25,26,27) The reported phase diagram of the C2S–C3P pseudo-binary system²⁸⁾ indicates that the phosphorus solubility limit strongly depends on the crystal structure of the C2S host compound. In particular, the α phase forms a complete solid solution in the C2S–C3P system.

Many studies on the dephosphorization behavior of the molten slag including C2S precipitates have been performed,^{15,16,17,18,19,20,21,22,23)} and it is believed that the C2S precipitates exist as the α phase and dissolve a lot of phosphorus in the slag. However, the crystal structure of the α phase has not been confirmed. Because the crystal structure of the C2S compound with low phosphorus solubility rapidly transforms during cooling from high temperature,²⁹⁾ it is almost impossible to verify the high-temperature phase in the solidified slag when it is analyzed at room temperature.

Recently, by a unique approach where high-temperature in-situ X-ray diffraction was performed by transmitting high-energy X-rays into the aerodynamically levitated C2S–C3P melt, we directly verified that crystallization of the liquid phase of the C2S–C3P system forms the α phase as the primary crystal.²⁹⁾ In contrast, the molten slag made in the dephosphorization process is a multicomponent system consisting of not only CaO, SiO₂, and P₂O₅, but also FeO_x and other minor components. It should be noted that the C2S compound formed in cement clinker or liquid slag accepts to some extent various types of doping elements, including Fe²⁺ and Fe³⁺, which affect the stable C2S polymorph.^{30,31,32,33,34,35,36)} Additionally, once the solubility limit is achieved, a ternary intermediate compound with the C2S-related composition (e.g., CaFeSiO₄, kirschsteinite) coexists with the liquid slag.³⁷⁾ These features indicate that the phase relationship between the liquid slag and the C2S precipitates cannot be simply predicted from the C2S–C3P system.

The aim of this study is to directly identify the crystal structure of the precipitated phase in the molten slag of the CaO–SiO₂–FeO_x–P₂O₅ system related to the dephosphorization process. In particular, we aim to verify whether the α phase of dicalcium silicate including phosphorus or another phase precipitates as the primary crystal from the molten FeO_x-containing slag. To accomplish this aim, we applied high-temperature in-situ X-ray diffraction to the FeO_x-containing slag, which was levitated and completely melted under an inert atmosphere in a sealed chamber.

2. Experimental

2.1. Sample Preparation

It has been reported that in the CaO–SiO₂–FeO_x system at metallic iron saturation, the C2S compound is the primary crystal at a CaO/SiO₂ mass ratio of 1.5.³⁷⁾ To investigate the FeO_x content dependence of the crystal structure of the C2S precipitate, three slag compositions were selected at a fixed CaO/SiO₂ mass ratio of 1.5 and a fixed P₂O₅ content of 5 mass% but varying the FeO content between 10 and 30 mass%. To investigate the P₂O₅ content dependence of the precipitate phase, two slag compositions were selected at a fixed FeO content of 10 mass% but varying the P₂O₅ content between 5 and 15 mass%. The chemical compositions of the slag samples are summarized in Table 1. These slag samples were prepared in the following way. First, the FeO compound was prepared by annealing a mixture of metallic iron and Fe₃O₄ reagent powder (special grade, Fuji-Film Wako Chemicals Co., Ltd., Osaka, Japan) at 1173 K in an iron crucible under a dehydrated Ar gas atmosphere overnight followed by cooling the material in vacuum. Second, the CaO compound was prepared by sintering CaCO₃ reagent (special grade, Fuji-Film Wako Chemicals Co., Ltd.) at 1223 K in an alumina crucible overnight for CO₂ emission and then cooling the resulting material in vacuum. Third, SiO₂ (quartz) and 3CaO·P₂O₅ reagents (special grade, Fuji-Film Wako Chemicals Co., Ltd.) and the prepared FeO and CaO powders were homogeneously mixed to satisfy the desired slag composition, and the mixture was pre-melted in a molybdenum crucible to make a slag sample in an induction furnace under a dehydrated Ar gas atmosphere. The pre-melted slag sample was then crushed into a powder and homogeneously mixed with 1 mass% of metallic iron powder to satisfy the metallic iron saturation condition. Finally, each slag sample was obtained by pressing the mixture into a piece with a diameter of about 1.5 mm.

Table 1. Chemical compositions of the slag samples (mass%).

Name	CaO	SiO2	FeO	P2O5
A	51	34	10	5
B	45	30	20	5
C	39	26	30	5
D	48	32	10	10
E	45	30	10	15

2.2. High-Temperature In-situ X-ray Diffraction

To obtain clear diffraction patterns of the fully liquid slag and precipitate at high temperature while avoiding contamination from the container material, we combined the high-energy and high-intensity X-ray source at SPring-8 and an aerodynamic levitation apparatus. A schematic diagram of the sealed chamber for aerodynamic levitation is shown in Fig. 1. The chamber was installed at the BL08W beamline of SPring-8, where a fixed-energy coherent X-ray light source (E = 115.6 keV, λ = 0.1078 Å, ΔE/E ≈ 10⁻³, where E and λ are the X-ray energy and wavelength) provided by using a superconducting wiggler and high photon flux (1.0 × 10¹⁵ photons·s⁻¹) was used.³⁸⁾ This enabled time-resolved diffraction measurements to be performed with high resolution to accurately detect precipitation and the phase transition events in the molten slag sample.

Fig. 1.

Schematic diagram of X-ray diffraction of an aerodynamically levitated sample in the chamber.

In the aerodynamic levitator (Fig. 1), a piece of the slag sample was first attached to the nozzle made of copper (gas inlet size of 1.0 mm diameter), and then the inner chamber atmosphere was reduced to vacuum (below 0.3 Pa) and replaced by highly pure Ar gas (99.9996% purity). Next, the sample was levitated by a highly pure Ar gas jet stream from the bottom of the nozzle and irradiated by a CO₂ gas laser generated by a Synrad Firestar t-100 laser oscillator (10.6 μm laser wavelength, 2.2 mm laser diameter, and 100 W average output power) from above through a reflecting mirror, a collecting lens, and a transmissive ZnSe window glass to heat to 2673 K for complete melting. The above aerodynamic levitation and laser heating were performed under a highly pure Ar gas atmosphere at ambient pressure in the sealed chamber, where additional oxidation of iron oxide during the experiment was minimized. The sample temperature was controlled externally by adjusting the laser power, and measured by a pyrometer placed outside of the chamber window glass, where calibration was performed to reproduce the crystallization temperatures of alumina (2346 K) and pure C2S (2416 K) as reference materials. The details of the calibration test are given in our previous paper.²⁹⁾ We then confirmed that the observed phase transition temperatures of the P₂O₅-free CaO–SiO₂–FeO_x slag corresponded to the thermodynamic prediction by the FactSage 7.3 program with the use of the latest Fact Oxide database.³⁹⁾

High-temperature in-situ X-ray diffraction was performed by transmitting a high-energy and high-intensity X-ray beam to the levitated molten slag through a transmissive Kapton polyimide film layer, and the diffracted light was collected by a two-dimensional amorphous Si flat panel detector at 1.5 s intervals. The X-ray beam size was 3 mm (vertical) by 1 mm (horizontal), so the X-ray radiation area covered almost the whole sample. The beam center position and the camera distance were calibrated from the diffraction pattern of the CeO₂ reference material. The obtained high-resolution two-dimensional area-detector image was converted to one-dimensional patterns using the PIXIA program in the Orochi program suite developed by Dr. Satoshi Tominaka at the National Institute for Materials Science, Japan.⁴⁰⁾

After the above experiments, several slag samples containing the C2S precipitate were quenched by shutting off the irradiating laser power (cooling rate on the order of −1000 K·s⁻¹). The cross-sectional microstructures of the quenched samples were observed by a JEOL JSM-5600 scanning electron microscope (JEOL Ltd., Tokyo, Japan) using acceleration voltage of 15 kV and the chemical compositions in local areas were analyzed by energy dispersive X-ray analysis (EDX).

3. Results

3.1. Identification of the Phase Precipitated from Molten Slag with Different FeO Contents (P₂O₅ = 5 mass%)

First, precipitation from fully liquid slag A (FeO = 10 mass%) was detected by in-situ X-ray diffraction. The typical temperature profile for precipitation from the liquid slag is shown in Fig. 2. After complete melting at above 2273 K, the sample temperature was gradually decreased (cooling rate of about −5 K·s⁻¹) until precipitation was detected. When precipitation was detected, the sample was cooled around 500 K below the liquidus temperature estimated from the phase diagram of the CaO–SiO₂–FeO_x system at metallic iron saturation calculated by FactSage 7.3.³⁹⁾ This was probably because inhomogeneous nucleation was suppressed for the levitated sample without the container. Unlike crystallization of the C2S–C3P melt,²⁹⁾ no specific temperature increase indicated that latent heat generation was detected. Figure 3 shows two-dimensional diffraction images of slag A detected at periods marked as (a)–(d) in Fig. 2. In the fully liquid state, a halo pattern was observed (Fig. 3(a)), whereas several fragments of Debye–Scherer rings were detected after precipitation (Figs. 3(b)–3(d)). This indicated that polycrystalline grains of limited orientation range precipitated in the slag. Nevertheless, high-resolution diffraction images were successfully obtained for the precipitate in the liquid slag at high temperature.

Fig. 2.

Temperature profile of precipitation from liquid slag A (FeO = 10 mass%, P₂O₅ = 5 mass%).

Fig. 3.

Two-dimensional diffraction images when precipitation occurred in slag A: (a) fully liquid state and (b)–(d) precipitation. (Online version in color.)

The results of phase identification of the precipitate from slag A using the converted one-dimensional diffraction pattern and the original two-dimensional image (Fig. 3(d)) are shown in Fig. 4. The one-dimensional diffraction pattern of the precipitate well matched the reference pattern of pure α-C2S reported by Mumme et al.²⁶⁾ rather than other compound references,^{25,27,41,42,43)} although each observed diffraction angle was positively shifted from the α-C2S reference. Thus, it was verified that the α-C2S compound precipitated as the primary crystal from liquid slag A. The positive shifts of the diffraction angles were attributed to dopant contamination, such as Fe²⁺, in the C2S crystal structure, which will be discussed later.

Fig. 4.

Results of phase identification of the precipitate from slag A: (a)–(d) converted one-dimensional diffraction patterns and (e) original two-dimensional image of Fig. 3(d). (Online version in color.)

The two-dimensional diffraction images and corresponding one-dimensional diffraction patterns of the precipitates from liquid slag B (FeO = 20 mass%) and C (FeO = 30 mass%) are shown in Fig. 5. In both cases, many diffraction spots formed several Debye–Scherer rings in the two-dimensional images after precipitation. However, unlike the case of slag A, the Debye-Scherer rings were not clearly observed. The corresponding one-dimensional patterns were in agreement with the reference pattern of α-C2S with positive shifts of the diffraction angles, where those at the first strong peaks were not clearly recognized. Therefore, up to 30 mass% FeO, the α-C2S phase was confirmed to be the primary crystal phase precipitated from the molten slag.

Fig. 5.

Diffraction patterns of the precipitates from liquid slags (a) B (FeO = 20 mass%, P₂O₅ = 5 mass%) and (b) C (FeO = 30 mass%, P₂O₅ = 5 mass%). (Online version in color.)

3.2. Precipitation from Molten Slags with Different P₂O₅ Contents (FeO = 10 mass%)

In this section, the results of precipitation from molten slag D (P₂O₅ = 10 mass%) and E (P₂O₅ = 15 mass%) are presented. The temperature profile and one-dimensional diffraction patterns for slag D when precipitation occurred from the undercooled liquid are shown in Fig. 6. Only a few diffraction peaks were observed in the halo patterns, although several more peaks were detected a few seconds after precipitation occurred. All of these diffraction peaks agreed with the reference pattern of α-C2S while positive shifts were observed. In contrast, unique precipitation behavior was observed for slag E with the highest P₂O₅ content (Fig. 7). Many diffraction peaks were observed after precipitation, and their peak intensities gradually increased with time. Most of these diffraction peaks were identified as the nagelschmidtite group (Ca₇(PO₄)₂(SiO₄)₂ is one of the members), where the peak positions were estimated from Ca₅Na₂(PO₄)₄ as another group member,⁴⁴⁾ while a few minor peaks were determined to be either silicocarnotite (Ca₅(PO₄)₂SiO₄)⁴¹⁾ or α-C2S.²⁶⁾ This precipitation behavior was reproduced in several runs of slag E. Therefore, it was concluded that nagelschmidtite was the primary crystal precipitated from this liquid slag composition, whereas silicocarnotite and α-C2S were secondary or metastable phases.

Fig. 6.

Precipitate from liquid slag D (FeO = 10 mass%, P₂O₅ = 10 mass%): (a) temperature profile when precipitation occurred and (b)–(e) one-dimensional diffraction patterns. (Online version in color.)

Fig. 7.

Precipitate from liquid slag E (FeO = 10 mass%, P₂O₅ = 15 mass%): (a) temperature profile when precipitation occurred and (b)–(e) one-dimensional diffraction patterns. (Online version in color.)

4. Discussion

In this section, we discuss Fe²⁺ contamination in the α-C2S compound precipitated from the molten slag.

Comparison of the one-dimensional diffraction patterns of the α-C2S precipitates from the liquid slags with a fixed P₂O₅ content but different FeO contents (slags A, B, and C) and that of the α-C2S precipitate from the C2S-C3P melt²⁹⁾ is shown in Fig. 8. Regardless of FeO content in the slag, the diffraction angles of the precipitates in FeO_x-containing slags A–C positively shifted to comparable degrees from those of the pure α-C2S reference²⁶⁾ and the precipitate from the FeO_x-free C2S–C3P melt. This tendency suggested Fe²⁺ contamination in the crystal structure of the α-C2S precipitate in the slag, because Ca²⁺ site substitution by the Fe²⁺ ion with a smaller ionic radius⁴⁵⁾ can decrease the lattice size of the α-C2S crystal, which leads to positive shifts of the diffraction angles. In addition, particularly for cases of slag B and C, several diffraction peaks were not clearly observed. This may partially attribute to the limited resolution of two-dimensional diffraction detection, but probably indicate that the crystal structure of the α-C2S precipitated from liquid slag was in a disordered state by random substitution of Ca²⁺ sites by Fe²⁺ ions.

Fig. 8.

Comparison of the one-dimensional diffraction patterns of the α-2CaO·SiO₂ precipitates from liquid slags A–C and that precipitated from the 2CaO·SiO₂–3CaO·P₂O₅ melt.²⁹⁾ (Online version in color.)

To analyze the detailed composition of the precipitate, slag sample B in the liquid state was quenched after holding the sample with α-C2S precipitation for 10 s at high temperature by stopping laser heating. The temperature profile and the change in the one-dimensional diffraction pattern when the sample was quenched are shown in Fig. 9. Because the containerless aerodynamic levitation was used, the sample was kept fully liquid state and prevented from the precipitation while it was cooled 600 K below the liquidus temperature estimated from the phase diagram of the CaO–SiO₂–FeO_x system at metallic iron saturation (Fig. 9(b)). Then, the precipitation of α-C2S phase from the undercooled liquid slag was observed (Fig. 9(c)). The α-C2S phase remained in the quenched sample, where each diffraction angle positively shifted because of shrinking of the lattice constant (Fig. 9(d)). In addition, several diffraction peaks corresponding to calcio-olivine²⁵⁾ and wustite⁴⁶⁾ were detected.

Fig. 9.

(a) Temperature profile and (b)–(d) one-dimensional diffraction patterns of slag B when quenched after precipitation. (Online version in color.)

The cross-sectional microstructure of the quenched sample is shown in Fig. 10 and the analyzed chemical compositions of the marked spots are summarized in Table 2. The average slag composition analyzed for the whole sample (1.5 mm in diameter) is also presented as “Bulk” in this table. It was confirmed that the analyzed bulk composition was identical to the desired composition for slag B. In Fig. 10, the following three different morphologies were identified. The matrix in gray contrast (spots 1–3) had a composition in which the molar ratio of (Ca + Fe) / (Si + P) was 1.9, which is close to the olivine composition. As compared with the average bulk composition, this matrix phase contained higher FeO but lower P₂O₅. The fine dispersed particles with bright contrast were identified as wustite (FeO_x). Therefore, it is suggested that the liquid matrix was partially decomposed into wustite and olivine phases when quenched. The dendritic morphology with dark contrast (spots 4–6) was identified as the α-C2S precipitate, because the molar ratio of (Ca + Fe) / (Si + P) was equal to 2.0 and the P₂O₅ content was locally higher than the P₂O₅ contents in the matrix. However, the phosphorus partition ratio between the precipitates and the matrix was lower than that reported for the solidified slag.^15,16,17) Surprisingly, these α-C2S precipitates included as much as 10 mass% FeO, which was lower than the average bulk composition but much higher than the reported solubility limit in C2S–C3P precipitates in the solidified slag.^{15,16,17,18,19,20,21,22,23)}

Fig. 10.

Cross-sectional micrograph of slag sample B quenched after precipitation. (Online version in color.)

Table 2. Results of chemical analysis of the local areas marked in Fig. 10.

Spot	Analyzed composition in mass%				Molar ratio (Ca+Fe)/(Si+P)	Phase
	CaO	SiO2	FeO	P2O5
1	38.8	32.5	26.6	2.1	1.9	L or Olivine
2	39.5	32.7	25.7	2.0	1.9	L or Olivine
3	37.7	33.6	26.8	1.9	1.8	L or Olivine
4	52.8	27.4	12.1	7.7	2.0	α-C2S
5	53.9	27.2	11.0	8.0	2.0	α-C2S
6	54.4	27.6	11.1	6.9	2.0	α-C2S
Bulk*)	45.4	29.3	20.6	4.7	2.0	–

*  - Average slag composition analyzed for whole sample area (diameter: 1.5 mm).

The above results indicated that the α-C2S precipitates in the liquid slag included plenty of Fe and P as solute elements but they still kept the molar ratio of (Ca + Fe) / (Si + P) equal to 2.0. This suggests that the α-C2S precipitates maintain charge balances in their crystal structures by substituting two SiO₄^4– tetrahedrons with two PO₄^3– tetrahedrons with one vacancy in Ca²⁺ site,²⁹⁾ and by substituting part of Ca²⁺ sites with Fe²⁺ ions. However, the α-C2S precipitates were temporarily oversaturated with FeO. The excess Fe²⁺ ion contamination at least causes shrinkage in the lattice size and may even make the α-C2S crystal structure disordered state, when it is randomly substituted for Ca²⁺ sites.

5. Conclusions

We have performed high-temperature in-situ X-ray diffraction measurements to directly identify the crystal structure of the phase precipitated from the molten slag of the CaO–SiO₂–FeO_x–P₂O₅ system, where the CaO/SiO₂ mass ratio was fixed at 1.5. The combined use of the aerodynamic levitation technique under an inert atmosphere and high-energy coherent X-ray imaging by the superconducting wiggler at SPring-8 enabled the precipitation behavior of the molten slag to be successfully determined, and the high-resolution diffraction patterns of the precipitates were evaluated. The following conclusions were drawn:

- When the FeO content was in the range 10–30 mass% at a fixed P₂O₅ content of 5 mass%, the α-2CaO·SiO₂ phase was identified as the primary crystal phase precipitated from the liquid slag.

- When the P₂O₅ content was in the range 5–10 mass% at a fixed FeO content of 10 mass%, the α-2CaO·SiO₂ phase remained the primary crystal. However, at 15 mass% P₂O₅, the nagelschmidtite (Ca₇(PO₄)₂(SiO₄)₂) intermediate compound alternatively precipitated from the liquid slag. In this case, α-2CaO·SiO₂ and silicocarnotite (Ca₅(PO₄)₂SiO₄) possibly precipitated as secondary solid phases.

- It is suggested that the α-2CaO·SiO₂ precipitate in the molten slag included not only phosphorus, but also an excess amount of iron because of the following reasons. First, the one-dimensional diffraction patterns of the α-2CaO·SiO₂ precipitates in the molten slag samples commonly shifted in the positive direction from those obtained for the α phase precipitated from the 2CaO·SiO₂–3CaO·P₂O₅ melt, which was attributed to Fe²⁺ contamination in the α-2CaO·SiO₂ crystal structure. Second, chemical analysis of the microstructure of the quenched slag sample revealed that phosphorus was selectively partitioned, but iron was also contained in the α-2CaO·SiO₂ precipitates and the Fe content greatly exceeded the reported solubility limit.

Acknowledgments

This study was financially supported by a 26th ISIJ Research Promotion Grant (Iron and Steel Institute of Japan). The results of high-temperature in-situ X-ray diffraction analysis were obtained through successful beamline applications 2018B1140 and 2019A1444 supported by SPring-8 (Hyogo, Japan). We thank Dr. Koji Ohara (Japan Synchrotron Radiation Research Institute, Japan) for assistance with the in-situ X-ray diffraction measurements. We thank Tim Cooper, PhD, from Edanz Group (www.edanzediting.com/ac) for editing a draft of this manuscript.

References

• 1)   S.  Kitamura,  T.  Kitamura,  E.  Aida,  R.  Sakomura,  T.  Kaneko and  T.  Nuibe: ISIJ Int., 31 (1991), 1329.
• 2)   N.  Sano,  F.  Tsukihashi and  A.  Tagaya: ISIJ Int., 31 (1991), 1345.
• 3)   J. J.  Pak and  R. J.  Fruehan: Metall. Trans. B, 22 (1991), 39.
• 4)   S.  Nakamura,  F.  Tsukihashi and  N.  Sano: ISIJ Int., 33 (1993), 53.
• 5)   P.  Wei,  M.  Sano,  M.  Hirasawa and  K.  Mori: ISIJ Int., 33 (1993), 479.
• 6)   O. I.  Ostrovski,  Y. I.  Utochkin,  A. V.  Pavlov and  R. A.  Akberdin: ISIJ Int., 34 (1994), 849.
• 7)   A.  Hernandez,  A.  Romero,  F.  Chavez,  M.  Angeles and  R. D.  Morales: ISIJ Int., 38 (1998), 126.
• 8)   A.  Sobandi,  H. G.  Katayama and  T.  Momono: ISIJ Int., 38 (1998), 781.
• 9)   B. J.  Monaghan,  R. J.  Pomfret and  K. S.  Coley: Metall. Mater. Trans. B, 29 (1998), 111.
• 10)   W. H.  Van Niekerk and  R. J.  Dippenaar: Metall. Mater. Trans. B, 29 (1998), 147.
• 11)   Y.  Kobayashi,  N.  Yoshida and  K.  Nagai: ISIJ Int., 44 (2004), 21.
• 12)   G.  Li,  T.  Hamano and  F.  Tsukihashi: ISIJ Int., 45 (2005), 12.
• 13)   Y.  Kaida,  M.  Hasegawa,  M.  Iwase,  Y.  Kikuchi and  K.  Wakimoto: Metall. Mater. Trans. B, 36 (2005), 43.
• 14)   T.  Hamano,  M.  Horibe and  K.  Ito: ISIJ Int., 44 (2004), 263.
• 15)   R.  Inoue and  H.  Suito: ISIJ Int., 46 (2006), 174.
• 16)   H.  Suito and  R.  Inoue: ISIJ Int., 46 (2006), 180.
• 17)   R.  Inoue and  H.  Suito: ISIJ Int., 46 (2006), 188.
• 18)   T.  Hamano,  S.  Fukagai and  F.  Tsukihashi: ISIJ Int., 46 (2006), 490.
• 19)   S.  Fukagai,  T.  Hamano and  F.  Tsukihashi: ISIJ Int., 47 (2007), 187.
• 20)   F.  Pahlevani,  S. Y.  Kitamura,  H.  Shibata and  N.  Maruoka: ISIJ Int., 50 (2010), 822.
• 21)   X.  Yang,  H.  Matsuura and  F.  Tsukihashi: ISIJ Int., 50 (2010), 702.
• 22)   M.  Hasegawa,  Y.  Kashiwaya and  M.  Iwase: High Temp. Mater. Process., 31 (2012), 421.
• 23)   X.  Gao,  H.  Matsuura,  I.  Sohn,  W.  Wang,  D. J.  Min and  F.  Tsukihashi: Mater. Trans., 54 (2013), 544.
• 24)   Y. J.  Kim,  I.  Nettleship and  W. M.  Kriven: J. Am. Ceram. Soc., 75 (1992), 2407.
• 25)   S.  Udagawa,  K.  Urabe,  M.  Natsume and  T.  Yano: Cem. Concr. Res., 10 (1980), 139.
• 26)   W.  Mumme,  L.  Cranswick and  B.  Chakoumakos: Neues Jahrb. Mineral. Abh., 170 (1996), 171.
• 27)   H.  Saalfeld and  K. H.  Klaska: Z. Kristallogr., 155 (1981), 65.
• 28)   W.  Fix,  H.  Heymann and  R.  Heinke: J. Am. Ceram. Soc., 52 (1969), 346.
• 29)   M.  Suzuki,  S.  Nakano,  H.  Serizawa and  N.  Umesaki: ISIJ Int., 60 (2020), 1127. https://doi.org/10.2355/isijinternational.ISIJINT-2019-449
• 30)   N. L.  Bowen,  J. F.  Schairer and  E.  Posnjak: Am. J. Sci., 26 (1933), 193.
• 31)   K. J. D.  Mackenzie: J. Mater. Sci., 17 (1982), 1834.
• 32)   C. J.  Chan,  W. M.  Kriven and  J. F.  Young: J. Am. Ceram. Soc., 75 (1992), 1621.
• 33)   M. Y.  Benarchid,  A.  Diouri,  A.  Boukhari,  J.  Aride and  I.  Elkhadiri: Mater. Chem. Phys., 94 (2005), 190.
• 34)   Y. L.  Chen,  P. H.  Shih,  L. C.  Chiang,  Y. K.  Chang,  H. C.  Lu and  J. E.  Chang: J. Hazard. Mater., 170 (2009), 443.
• 35)   M.  Suzuki,  N.  Umesaki,  T.  Okajima and  T.  Tanaka: J. Am. Ceram. Soc., 99 (2016), 3151.
• 36)   K. O.  Jang,  X.  Ma,  J.  Zhu,  H.  Xu,  G.  Wang and  B.  Zhao: Metall. Mater. Trans. B, 48 (2017), 1547.
• 37)   A.  Muan and  E. F.  Osborn: Phase Equilibria among Oxides in Steelmaking, Addison-Wesley Publishing Company, Reading, MA, (1965), 113.
• 38)   K.  Ohara,  S.  Tominaka,  H.  Yamada,  M.  Takahashi,  H.  Yamaguchi,  F.  Utsuno,  T.  Umeki,  A.  Yao,  K.  Nakada,  M.  Takemoto,  S.  Hiroi,  N.  Tsuji and  T.  Wakihara: J. Synchrotron Radiat., 25 (2018), 1627.
• 39)  CRCT/GTT: FactSage, http://factsage.com, (accessed 2020-02-07).
• 40)   S.  Tominaka,  H.  Yamada,  S.  Hiroi,  S. I.  Kawaguchi and  K.  Ohara: ACS Omega, 3 (2018), 8874.
• 41)   B.  Dickens and  W. E.  Brown: Tschermaks Mineral. Petrogr. Mitt., 16 (1971), 1.
• 42)   Y.  Ohashi and  L. W.  Finger: Am. Mineral., 63 (1978), 274.
• 43)   S.  Saburi,  I.  Kusachi,  C.  Henmi,  A.  Kawahara,  K.  Henmi and  I.  Kawada: Mineral. J., 8 (1976), 240.
• 44)   G.  Celotti and  E.  Landi: J. Eur. Ceram. Soc., 23 (2003), 851.
• 45)   R. D.  Shannon: Acta Crystallogr. A, 32 (1976), 751.
• 46)   H.  Fjellvåg,  F.  Grønvold,  S.  Stølen and  B. C.  Hauback: J. Solid State Chem., 124 (1996), 52.