Co-modification and Crystalline-control of Ti-bearing Blast Furnace Slags

Abstract

The present paper investigated how the SiO₂ and B₂O₃ additions influenced the crystallization behaviors of Ti-bearing blast furnace (Ti-BF) slags with a purpose of recycling the titanium. Ti-BF slags were co-modified by different SiO₂ and B₂O₃ contents. Isothermal experiments were carried out using a Single Hot Thermocouple Technique (SHTT) to in-situ determine the crystal formation, and accordingly Time Temperature Transformation (TTT) curves were established. The results showed that the small amount of B₂O₃ addition can efficiently promote the transformation of primary crystalline phase from perovskite to rutile, whereas the precipitation of perovskite was suppressed. The kinetics of precipitation of the crystalline phase was explored and the results indicated that the growth of rod shape rutile was 1-D with the rate-controlling step of interfacial reaction; whereas the precipitated perovskite presented a 3-D growth style. With the increase of holding time, the nucleation rate of the formed crystals including rutile and perovskite became smaller.

1. Introduction

Titanium resource is abundant in China and 91% (wt.% TiO₂) of that is located in Sichuan province, South China, in the form of vanadium titanium-magnetite ores.¹⁾ Conventionally, blast furnace (BF) process is used to treat these ores, through which the valuable elements (Fe, V et al.) are extracted, while titanium element is left into the slags, resulting in Ti-bearing blast furnace (Ti-BF) slags. These Ti-BF slags, with the TiO₂ content of 22–25 wt.%,²⁾ are typical residues generated and important secondary resources for titanium extraction. However, most of these slags are not treated reasonably and disposed in the slag yard, which causes a number of problems including resource waste, water contamination, air pollution and landscape transformation.^3,4) Nowadays, South China has produced more than 70 million tons of Ti-BF slags and the production rate is still up to 3 million tons per year.⁵⁾ In order to reduce the environmental pollution and extract the titanium element, selective crystallization and phase separation (SCPS) method has been proposed and extensively studied in the past years.^{1,2,5,6,7,8,9)} The whole process of SCPS method can be divided into several steps. Firstly, the Ti-BF slags are modified by additives, such as CaO, SiO₂ or P₂O₅, which are likely to enhance the precipitation of the Ti-enriched phases. Secondly, the cooling processes of the Ti-BF slags are reasonably controlled in advance to fully enrich titanium element into selected crystalline phases, such as perovskite,^1,2,6,8) rutile^5,7,8) or anosovite.⁹⁾ Then the Ti-enriched phases are separated from the glassy slags using gravity separation or flotation separation method. The separated rutile from slags can be further treated through hydrochloric acid leaching method for TiO₂ white pigment production or used as raw materials in titanium metallurgy.^10,11,12,13)

The most important considerations of SCPS method are the selection and formation of Ti-enriched phases. The primary crystalline phase of the industrial Ti-BF slags is perovskite, the crystallization behaviors of which have been studied by many researchers.^6,7,8,9) Compared with perovskite, rutile showed a better separation property because of its rod shape structure and high density, up to 4.2–4.3 g/cm³.⁵⁾ The previous study showed that SiO₂ can be used to modify the slags and substantially promote the precipitation of rutile with the optimum SiO₂ content of 35 wt.%.⁹⁾ The original basicity ((CaO/SiO₂, wt.%)) of Ti-BF slags is 1.2, and consequently this will require large amounts of SiO₂ addition and cause a lot of energy consumption. It is necessary to explore a novel way to modify the Ti-BF slags to obtain the Ti-enriched phases. B₂O₃ is a typical acid oxide and it is reasonable to conjecture that B₂O₃ could promote the precipitation of rutile with the previous reports that B₂O₃ could enhance the growth of rutile in the solution to some extent.^14,15) Therefore, the present study was motivated that the small addition of B₂O₃ can remarkably modify the basicity of the Ti-BF slags instead of SiO₂. In other words, the Ti-bearing BF slags were co-modified by SiO₂ and B₂O₃. In this study, a series of Ti-BF slags containing different levels of B₂O₃ with different basicity were designed and the crystallization behaviors of these slags were analyzed to investigate the co-modification of SiO₂ and B₂O₃. Single Hot Thermocouple Technique (SHTT) was used to carry out the investigation and visualize phase change in the slag melts, through which Time Temperature Transformation (TTT) curves were established.

2. Materials and Methods

2.1. Sample Preparation

In the present study, two series of Ti-BF slags were designed with the range of basicity 0.5–0.7 and B₂O₃ content 0–7 wt.%, as shown in Table 1. Analytical reagent (AR) oxides were used to prepare the samples and these oxides were mixed thoroughly to form a homogeneous blend. Then the mixture was placed in a Pt crucible (Φ40×45×H40 mm) and pre-melted in a tube furnace at 1500°C under Ar atmosphere for 2 h to homogenize the chemical compositions. The liquid slags were then rapidly quenched by cold water to form amorphous phase. The obtained solid slags were dried at 120°C for 12 h, crushed and grinded to 300 meshes for SHTT experiments. To confirm the amorphous phase of the slags, X-ray diffraction (XRD) analysis was performed and the results of slag series A as an example are clearly shown in Fig. 1.

Table 1. Chemical compositions of modified Ti-BF slags (wt.%).

Sample	Basicity	CaO	SiO2	Al2O3	MgO	TiO2	B2O3
A1	0.5	19	37	12	7	25	0
A2	0.5	18	36	12	7	24	3
A3	0.5	17	35	11	7	23	7
B1	0.7	23	33	12	7	25	0
B2	0.7	22	32	12	7	24	3
B3	0.7	21	31	11	7	23	7

Fig. 1.

XRD results of the pre-melted slags (Series A).

2.2. Observation of Crystallization Behaviors

Observation of crystallization behaviors including crystal formation and growth in the slag melts was carried out using SHTT, which combined the advantages of in-situ optical observation and the low inertial of the system. The working mechanism of SHTT has been described in detail elsewhere^16,17) and was briefly introduced here. As shown in Fig. 2, the heating and cooling of slag samples were conducted by a Pt-Rh thermocouple (B-type), the temperature of which was controlled by the computer program. The images of the crystallization behaviors in the slag melts were captured by the microscope equipped with a video camera. Before the isothermal experiments, the temperature of thermocouple was calibrated using pure K₂SO₄ with a constant melting point of 1067°C. Around 10 mg slag sample was mounted on the top of the thermocouple in ambient air, heated to 1500°C and held for 120 s to eliminate the bubbles and homogenize the chemical compositions. Then the liquid slags were rapidly quenched at a cooling rate of 50°C/s to a given temperature and held for a long time for the observation of crystallization. During this process, the crystalline evolution with time in the slag melts was clearly observed and the sample images were instantaneously captured by the video camera.

Fig. 2.

Schematic diagram of the SHTT instrument.

2.3. Analysis of Crystalline Phases

The crystalline phase in the slag samples was identified by X-Ray Powder Diffraction (XRD) analysis. The sample size after SHTT experiments was not large enough to perform XRD tests, so more pre-melted glassy slag samples were placed in a platinum crucible in a tube furnace, fully melted at 1500°C, rapidly cooled to a given temperature and held at this temperature for 2 h. Then the samples after crystallization were rapidly quenched by water. The samples obtained this way were crushed into fine powders and examined by XRD technique at room temperature. The microstructures of the crystalline phases obtained by SHTT were observed by Scanning Electron Microscope (SEM) with Back Scattered Electron (BSE) and the chemical compositions were verified by the Energy-dispersive X-Ray spectroscopy (EDS) tests.

3. Results and Discussion

3.1. Effect of B₂O₃ on TTT Curves

To construct TTT curves, isothermal experiments with different holding temperatures in the range of 850–1300°C were performed. Generally, TTT curves are composed of a series of incubation time with varying temperature. As the liquid slags are rapidly quenched to the holding temperature, it takes some time to crystalize, i.e., incubation time. Incubation time is an important parameter to characterize the crystallization ability and a shorter incubation time is generally associated with a stronger crystallization tendency. In this study, the incubation time at each temperature was measured three times and the average value was taken in order to reduce the error of measurements.

Figure 3 illustrated the TTT curves in this study. As can be observed, these TTT curves showed a similar shape, i.e., double “C” from high temperatures to low temperatures, which was supposedly the occurrence of two different crystallization events. As the holding temperature was above the break temperature, there was no crystal precipitated. With the holding temperature decreasing lower than the break temperature, crystallization occurred in the melts. For slag series A, rod shape crystal was the primary crystalline phase and first observed to form below break temperature in slag melts, as shown in Fig. 4(a). With further decreasing temperature, the crystals with different structures, such as cloud and dendrite shape, started to form and coexisted in the slag melts. As for slag series B, the primary crystalline phase below break temperature was dendrite shape crystal for samples B1 and B2, as shown in Fig. 4(b), whereas the rod shape crystal was formed in sample B3 with 7 wt.% B₂O₃ content. Similarly, several crystalline phases coexisted in the melts as the temperature further decreased. Figure 5 summarizes the varying trend of the break temperature and regions that the crystals were formed for different samples. It can be observed that the break temperature of slags which rod shape crystal was precipitated became higher with the increase of B₂O₃ content, which indicated that the rod shape crystal was enhanced by B₂O₃, whereas the precipitation of dendrite shape crystal was suppressed by B₂O₃ addition. It is also noted that the incubation time firstly decreased, and then increased with decreasing temperature. The reason caused is that the undercooling degree of slag melts increased with the decrease of temperature, resulting in an increase of crystallization ability. Meanwhile, the viscosity of slag melts correspondingly increased with decreasing temperature, which resulted in the decrease of diffusion of ions in the slag melts and consequently a decrease the crystallization ability. These two opposite effects resulted in the “C” shape curve. As the temperature further decreased, the similar variation tendency with the first “C” shape appeared in the crystallization temperature region.

Fig. 3.

TTT curves of the modified slags (a) slag series A and (b) slag series B.

Fig. 4.

Crystallization behaviors of samples (a) A2 at 1200°C and (2) B1 at 1160°C.

Fig. 5.

Break temperature and the temperature region that crystalline phases were formed.

From Fig. 3, another apparent variation trend of TTT curves was observed. For slag series A, the incubation time became shorter with the increase of B₂O₃ content, which indicated that the formation of rod shape crystal was improved by B₂O₃ addition. This suggested that the added B₂O₃ may change the structure of the slag melts in a direction of enhancement of the precipitation of the rod shape crystal. As for slag series B, the incubation time slightly became longer with increasing B₂O₃ content from sample B1 to B2, which suggested that the precipitation of the dendrite shape crystal was suppressed by B₂O₃ addition to some extent. However, the incubation time of sample B3 was shorter than samples B1 and B2, which was probably due to the enhancement effect of B₂O₃ on the crystallization of the rod shape crystal.

3.2. Effect of B₂O₃ on Crystalline Phases

The structures of crystals were first observed using SHTT images with varying basicity and B₂O₃ contents and then the samples with different structures were quenched from different temperatures and identified by SEM and XRD techniques. Figure 6 showed the XRD results of the samples quenched from different temperatures.

Fig. 6.

XRD patterns of the quenched samples.

For slag series A, the rod shape crystal was the primary crystalline phase and the XRD results indicated that the rod shape crystal is rutile, as shown in Fig. 6(a). While for sample A1 quenched from 1040°C, the crystalline phases were composed of rutile and gehlenite, as shown in Fig. 6(b), which suggested that the cloud shape crystal coexisting with rutile is gehlenite. For slag series B, the primary crystalline phase for sample B1 and B2 showed the dendrite shape, whereas the rod shape crystals were formed in sample B3. Figure 6(c) illustrates the XRD result of sample B1 quenched from 1160°C, and the dendrite shape crystals were confirmed to be perovskite. While for sample B3, the XRD result (Fig. 6(d)) indicated that the rod shape crystal was rutile.

To study the microstructure of the observed rod shape crystals, SEM tests were performed. Figure 7(a) showed the BSE images of sample A1 quenched from 1220°C and it can be clearly observed that the primary crystalline phase was typical rod shape structure. The EDS results indicated that chemical composition of the rod shape crystals was TiO₂. From the BSE images of the rod shape rutile, another interesting phenomenon was observed (Fig. 7(b)). During the preparation of the sample for SEM tests, some of the rod shape rutile was stripped from the glassy slags. This may suggest that the rutile shows a good separation property from the slags. This property could be in favor of the gravity separation or flotation separation in the following procedures. The microstructure of the primary crystalline phase in sample B1 were also examined by SEM tests and typical dendrite shape structures were observed, as clearly shown in Fig. 8. The EDS results also identified the formation of perovskite. These SEM results are consistent with the XRD analysis (Fig. 6(c)).

Fig. 7.

BSE images of sample A1 quenched at 1220°C.

Fig. 8.

BSE images of sample B1 quenched at 1160°C.

From the aforementioned analysis, it can be seen that the B₂O₃ addition promoted rutile as the primary crystalline phases for slag series B other than perovskite. To clarify the thermodynamics of the phase transformation in the slags, Factsage calculation was carried out in this study.¹⁸⁾ As shown in Fig. 9(a), the primary crystalline phase for sample B1 is perovskite as the temperature decreases, and several phases coexisted with the further decrease of temperature, which well agreed with the experimental results. As the B₂O₃ was added into the slag melt, the primary crystallization region is also perovskite as seen in Fig. (b). This is not consistent with the experiment result (Fig. 6(d)). The reason caused may be due to the lack of thermodynamics data of titanium.

Fig. 9.

Isopleth phase diagrams for sample (a) B1and (b) B3.

According to the theory of the structure of slag melts, B₂O₃ is a typical acidic oxide with the ion-oxygen (Coulombic force between the cation and oxygen anion) of 2.34,¹⁹⁾ which could significantly increase the acidity of the slag melts and restrict the linkage between Ca²⁺ and TiO₃^2– to form perovskite. The precipitation of rutile was therefore promoted. In addition, according to previous studies, the addition of B₂O₃ could reduce the viscosity of slag melt to some extent because of its low melting point,^20,21) which was also beneficial to crystalline formation. In this section, the variation of crystalline phases was mainly analyzed, which was the result of the co-modification of SiO₂ and B₂O₃.

Based on the present experimental results, a possible industrial process could be proposed, in which slags A1 and B3 were the candidate samples for heat treatment. First the slag melts were co-modified by additives and second the temperature of the slag melts were designed in advance and reasonably controlled to ensure the full growth of the selected crystal. Then phase separation process was performed to obtain crystals and glassy phases.

3.3. Kinetics of Crystalline Precipitation

3.3.1. Analysis of Crystallization Kinetics

Generally, Johnson-Mehl-Avrami (JMA) equation has been recommended as a useful expression to study the kinetics of crystalline precipitation in the melts,^22,23) as presented in Eq. (1).   

$$X=1-exp\left[-k{\left(t-\tau \right)}^n\right]$$(1)

Where X is the volume fraction crystalline phase, t is the holding time, τ is the incubation time, k is the effective rate constant including nucleation and growth, and n is the Avrami exponent associated with the nucleation and growth mechanism. In Eq. (1), the volume fraction of crystalline phase, X at a certain holding time and a certain holding temperature could be approximately calculated as follows:²⁴⁾   

$$X=\frac{A_C}{A_T}$$(2)

Where A_C is the area of crystalline phase and A_T is the total area of slag melts. A_C and A_T were calculated by a software Imagetools 3.0 (The University of Texas Health Science Center at San Antonio, Texas, USA). Through rearranging Eq. (1), Eq. (3) can be deduced and by plotting $ln\left[ln\left(\frac{1}{1-X}\right)\right]$ versus. ln(t – τ), the values of k and n can be obtained and therefore the kinetics of crystalline phase are analyzed.   

$$ln\left[ln\left(\frac{1}{1-X}\right)\right]=lnk+nln\left(t-\tau \right)$$(3)

Figure 10 showed the relationship between that $ln\left[ln\left(\frac{1}{1-X}\right)\right]$ and ln(t – τ), which showed a good linear relation. The k and n values of the present system were therefore calculated through Eq. (3) and are listed in Table 2.

Fig. 10.

Crystalline volume fraction evolution for sample (a) A1 and (b) B1.

Table 2. Calculated values of n and lnk for samples.

Sample	Temperature (°C)	n	Mean value	lnk
A1	1240	1.006	n=1.151	–7.644
	1220	1.155		–7.433
	1200	1.291		–6.549
A2	1240	1.108	n=1.165	–7.347
	1220	1.180		–8.321
	1200	1.207		–7.478
A3	1260	1.130	n=1.257	–8.588
	1240	1.294		–7.756
	1220	1.348		–6.476
B1	1200	2.163	n=2.349	–13.010
	1180	2.535		–14.889
B2	1160	2.789	n=2.789	–12.607
B3	1200	1.109	n=1.328	–8.485
	1180	1.332		–8.327
	1160	1.543		–7.285

As aforementioned, the Avrami exponent n was related to the nucleation and growth mechanism of the crystalline phases. The relationship between Avrami exponent n and crystallization mechanism has been explored by several studies^25,26,27) and can be expressed as follows:   

$$n=a+bc$$(4)

Where n is the Avrami exponent, a is the nucleation constant, b is the dimension of crystalline growth, and c is the growth constant. The relationship between these constants and the nucleation and growth mechanism of crystalline precipitation is illustrated in Table 3 in detail. According to the results in Table 3 and the present experiments, the crystalline precipitation kinetics of rutile and perovskite can be identified.

Table 3. Relationship between crystalline mechanism and crystalline constants.

Constant	Value	Mechanism
a	0	Nucleation rate is constant
	0–1	Nucleation rate decreased with holding time
	1	Nucleation rate increased with holding time
b	1	1-D growth
	2	2-D growth
	3	3-D growth
c	1	Interface-controlled growth
	0.5	Diffusion-controlled growth

3.3.2. Kinetics of Rutile Precipitation

For slag series A and sample B3 that rod shape rutile could be formed in the melts, the length evolution of the rutile with time could be calculated, as shown in Fig. 11. As can be noted, the length of rod shape rutile presented a well linear relationship with holding time. According to the study of Jackon,²⁵⁾ this linear relationship indicated that the rate-controlling step of rutile growth should be interfacial reaction and therefore the growth constant c equaled 1. The experimental observations also showed that the rod shape rutile was first formed along the side of thermocouple, which clearly suggested the precipitation of rutile was heterogeneous nucleation and 1-D growth and therefore the dimension of crystalline growth b equaled 1. Then based on the mathematical relation of these constants, it can be concluded that the value of the nucleation constant a was in the range of 0–1, which suggested the nucleation rate of rutile in the slag melts decreased with holding time. Our previous study investigated the growth mechanism of rutile in a slag melt with the basicity of 0.5⁵⁾ and the one-dimensional growth of rutile was also observed. From Table 2, another variation trend of the value of n could be visible. As the temperature decreased for the samples with rutile formation, the value of n became larger and correspondingly the value of a became larger, which revealed that lower temperature improved the nucleation rate due to the higher undercooling degree. And meanwhile, the value of n increased with the addition of B₂O₃ for slag series A with rutile formation, which suggested that B₂O₃ promoted the nucleation of rutile in the slag melts.

Fig. 11.

Length evolution of rutile for sample A1 as a function of time.

3.3.3. Kinetics of Perovskite Precipitation

As for the kinetics of perovskite precipitation, similar analysis was carried out. The experimental images of SHTT and SEM presented that the precipitated perovskite showed dendrite shape or cubic shape structure, which suggested that perovskite was 3-D growth, i.e., the dimension of crystalline growth b equaled 3. According to the previous studies of Li et al.⁸⁾ and Lou et al.,²⁸⁾ the crystallization of perovskite was diffusion controlled, i.e., growth constant c equaled 0.5. And similarly based on the mathematical relation of these constants, it can be inferred that the value of the nucleation constant a was in the range of 0–1, which indicated the nucleation rate of perovskite in the slag melts also decreased with holding time. The foregoing analysis could be verified by the observation of the varying phenomena of nucleation with temperature to some extent in this study. It also can be observed that the value of n slightly decreased from sample B1 to sample B2, which could be resulted from the restriction effect of B₂O₃ on the crystallization of perovskite.

From the foregoing analysis, it can be seen that the added B₂O₃ could remarkably enhance the growth of rutile, which provided one potential of SiO₂ reduction. However, it should be pointed out that the further massive reduction of SiO₂ needs the co-modification by other reagent addition, such as ZrO₂, P₂O₅, et al. It is expected these additions can substantially decrease the addition of SiO₂ and enhance the crystallization of the selected crystalline phases.

4. Conclusions

The present study provided a fundamental analysis of the Ti-bearing blast furnace slags co-modified by SiO₂ and B₂O₃ aiming at extracting the titanium resource. Two series of slags were designed and prepared with different levels of B₂O₃ and different CaO/SiO₂ ratios of 0.5 and 0.7. A Single Hot Thermocouple Technique (SHTT) was used to carry out the isothermal experiments and Time Temperature Transformation (TTT) curves were therefore obtained. Moreover, the kinetics of the crystalline phase including rutile and perovskite, were identified. The main conclusions were summarized as below:

(1) The added B₂O₃ greatly influenced the crystalline phase precipitated in the slag melts. For slag series B, the crystalline phase changed into rod shape rutile from the primary phase, perovskite with the addition of B₂O₃.

(2) The precipitation of rutile was enhanced while that of perovskite was suppressed by B₂O₃ addition, because B₂O₃ was a typical acidic oxide, which resulted in the increasing acidity and the decreasing viscosity of the slag melts.

(3) Rutile precipitated in the slag melts showed 1-D growth while perovskite was 3-D growth. With the increase of holding time, the nucleation rate of the formed crystals became smaller.

Acknowledgement

The authors gratefully acknowledge financial support by the Common Development Fund of Beijing and the National Natural Science Foundation of China (51172003, and 51172001). Supports by the National High Technology Research and Development Program of China (863 Program, 2012AA06A114) and Key Projects in the National Science & Technology Pillar Program (2011BAB03B02 and 2013BAC14B07) are also acknowledged.

References

• 1)   L.  Zhang,  L. N.  Zhang,  M. Y.  Wang,  G. Q.  Li and  Z. T.  Sui: J. Non-Cryst. Solids., 353 (2007), 2214.
• 2)   L.  Zhang,  L. N.  Zhang,  M. Y.  Wang,  T. P.  Lou,  Z. T.  Sui and  J. S.  Jang: J. Non-Cryst. Solids., 352 (2006), 123.
• 3)   Y.  Peng: Titanium Ind. Prog., 22 (2005), 44.
• 4)   W. B.  Zhou,  C. M.  Ke,  Q.  Zhang,  K. N.  Zhang and  S. F.  Du: Multipurp. Util. Min. Resour., 3 (2007), 35.
• 5)   J.  Li,  X. D.  Wang and  Z. T.  Zhang: ISIJ Int., 51 (2011), 1396.
• 6)   L.  Zhang,  L. N.  Zhang,  M. Y.  Wang,  G. Q.  Li and  Z. T.  Sui: ISIJ Int., 46 (2006), 458.
• 7)   W.  Zhang,  L.  Zhang,  J. H.  Zhang and  N. X.  Feng: Ind. Eng. Chem. Res., 51 (2012), 12294.
• 8)   J.  Li,  Z. T.  Zhang,  L. L.  Liu,  W. L.  Wang and  X. D.  Wang: ISIJ Int., 53 (2013), 1696.
• 9)   J.  Li,  Z. T.  Zhang and  X. D.  Wang: Ironmaking Steelmaking, 39 (2012), 414.
• 10)   L.  Zhang,  J. H.  Zhang,  W.  Zhang and  G. Q.  Li: Ind. Eng. Chem. Res., 52 (2013), 4924.
• 11)   F.  Valighazvini,  F.  Rashch and  R. K.  Nekouei: Ind. Eng. Chem. Res., 52 (2013), 1723.
• 12)   L. S.  Zhao,  Y. H.  Liu,  L. N.  Wang,  H. X.  Zhao,  D. S.  Chen,  B. N.  Zhong,  J. C.  Wang and  T.  Qi: Ind. Eng. Chem. Res., 53 (2014), 70.
• 13)   Y.  Wang,  J.  Li,  L. N.  Wang,  T. Y.  Xue and  T.  Qi: Ind. Eng. Chem. Res., 49 (2010), 7693.
• 14)   D. M.  Chen,  D.  Yang,  Q.  Wang and  Z. Y.  Jiang: Ind. Eng. Chem. Res., 45 (2006), 4110.
• 15)   B.  Grzmil,  M.  Glen,  B.  Kic and  K.  Lubkowski: Ind. Eng. Chem. Res., 50 (2011), 6535.
• 16)   Y.  Kashiwaya,  C. E.  Cicutti and  A. W.  Cramb: ISIJ Int., 38 (1998), 357.
• 17)   Y.  Kashiwaya,  C. E.  Cicutti,  A. W.  Cramb and  K.  Ishii: ISIJ Int., 38 (1998), 348.
• 18)   C. W.  Bale,  P.  Chartrand,  S. A.  Degterov,  G.  Eriksson,  K.  Hack,  R. B.  Mahfoud,  J.  Melancon,  A. D.  Pelton and  S.  Peterson: Calphad, 26 (2002), 189.
• 19)   Y.  Waseda and  J. M.  Toguri: The Structure and Properties of Oxide Melts, World Scientific Publ., Singapore, (2002), 3.
• 20)   L. A.  McCulloch: J. Am. Chem. Soc., 59 (1937), 2650.
• 21)   S.  Sridhar,  K. C.  Mills,  O. D. C.  Afrange,  H. P.  Lörz and  R.  Carli: Ironmaking Steelmaking, 27 (2000), 238.
• 22)   M. C.  Weinberg,  D. P.  Birnie, III and  V. A.  Shneidman: J. Non-Cryst. Solids, 219 (1997), 89.
• 23)   D. A.  Porter and  K. E.  Eastering: Phase Transformations in Metals and Alloys, 2nd ed., CRC Press, New York, (2004), 287.
• 24)   L. J.  Zhou,  W. L.  Wang,  F. J.  Ma,  J.  Li,  J.  Wei,  H.  Matsuura and  F.  Tsukihashi: Metall. Mater. Trans. B, 43 (2012), 354.
• 25)   K. A.  Jackson: Liquid Metals and Solidification, ASM, Cleveland, OH, (1958), 174.
• 26)   D. R.  MacFarlane and  M.  Fragoulis: Phys. Chem. Glasses, 37 (1986), 228.
• 27)   J.  Huberty and  H. F.  Xu: J. Solid State Chem., 181 (2008), 508.
• 28)   T. P.  Lou,  Y. H.  Li,  L. S.  Li and  Z. T.  Sui: J. China Ceram. Soc., 28 (2000), 255.