Crystallization Kinetics and Structure of CaF₂–CaO–Al₂O₃–MgO–TiO₂ Slag for Electroslag Remelting

Abstract

The crystallization kinetics and structure of CaF₂–CaO–Al₂O₃–MgO–TiO₂ slag for electroslag remelting (ESR) were investigated by differential scanning calorimetry and Raman spectroscopy, respectively. The results show that increasing TiO₂ content from 4.2 mass% to 16.8 mass% in the slag lowers the crystallization rate of the slag. The crystallization of the primary crystalline phase (11CaO·7Al₂O₃·CaF₂) in the slags with 4.2–12.6 mass% TiO₂, and primary crystalline phase (11CaO·7Al₂O₃·CaF₂ and CaTiO₃) in slag with 16.8 mass% TiO₂ originates from constant nucleation rate, interface reaction controlled and one-dimensional growth, irrespective of the TiO₂ contents of the slag. Raman spectroscopy study indicates that TiO₂ plays a network-modifier role in relatively more complex Al–O–Al band and Q⁴ units by forming Q² units and less complex Ti₂O₆^4- chain unit, resulting the decreasing of the polymerization degree of the slag. The variation in slag structure is in agreement with the analysis of crystallization kinetics.

1. Introduction

Slag plays an important role in electroslag remelting (ESR) process. The main function of ESR-type slag includes not only that the liquid slag pool generates Joule heat for melting electrode and refines the impurities, but also that the slag film in the mold control horizontal heat transfer between initial solidifying steel shell and mold. Appropriate horizontal heat transfer in the mold can provide a sound condition for the good surface quality of the billet, which are largely dependent on the crystallization behaviors of slag, such as the morphology and extent of crystalline phase,^1,2,3) which is determined by the nucleation and growth of the crystals.⁴⁾ ESR-type TiO₂-bearing slag generally are applied for the production of Ti-containing steel and alloy by ESR, but poor surface quality of the remelted ingot is often observed during ESR production,⁵⁾ resulting from the inappropriate horizontal heat transfer in the mold. An analysis of the crystallization kinetics of slag can provide a guidance for designing the slag and optimizing the crystallization characteristics (such as crystallization temperature, crystallinity and the morphology of crystalline phase).⁶⁾

Several studies have been conducted to investigate the crystallization kinetics of TiO₂-bearing metallurgical slag, such as mold fluxes and steelmaking slag.^7,8,9,10) Wang et al.⁷⁾ found that TiO₂ substitution of SiO₂ in CaO-SiO₂-based mold fluxes hindered the cuspidine crystallization by increasing the activation energy for crystal growth. The activation energy for crystal growth in CaO-SiO₂-TiO₂-10%B₂O₃ (CaO/SiO₂=1) glassy slag increased first and then decreased with increasing TiO₂ content in slag, indicating that crystallization ability of the glassy slag decreased initially and then increased with the increase of TiO₂ content.⁸⁾ Klug et al.⁹⁾ reported that varying Na₂O content provided a possibility to control the crystallization kinetics of CaO–SiO₂–TiO₂ slag. Wang et al.¹⁰⁾ demonstrated that the activation energy for the primary crystal growth increased with increasing TiO₂ content from 0 to 9.67 mass% in P-bearing steelmaking slag. However, there is a lack of study on the crystallization kinetics of high-fluoride TiO₂-bearing slag.

The crystallization characteristics of slag are closely related to the structure of the slag. The structure of TiO₂-bearing metallurgical slag has been investigated by many researchers,^11,12,13,14) including the authors’ previous study about the structure of low-fluoride ESR-type TiO₂-bearing slag.¹⁵⁾ However, there is no investigation on the effect of TiO₂ on the high-fluoride TiO₂-bearing slag, especially the slag with high TiO₂ content.

In the previous article,¹⁶⁾ the effort has been made to investigate the role of TiO₂ on the crystallization characteristics of ESR-type TiO₂-bearing slag and its influence on the surface quality of as-cast ingots based on ESR pilot trials. However, the information is still indistinct with respect to the crystallization kinetics and structure of the ESR-type slag with various TiO₂ contents. The aim of the current study is to elucidate the crystallization kinetics of CaF₂–CaO–Al₂O₃–MgO–TiO₂ slag for ESR based on differential scanning calorimetry (DSC) data. The Raman spectroscopy analysis was performed to study the slag structure, which provides the curves for clarifying the crystallization kinetics of the slag.

2. Experimental

2.1. Sample Preparation

Reagent-grade CaCO₃, CaF₂, Al₂O₃, MgO and TiO₂ were used to produce slag samples. CaCO₃ powders were calcined at 1323 K for 10 hours in muffle furnace to produce CaO. The thoroughly mixed powders were melted in a platinum crucible at 1773 K for 5 minutes to homogenize their chemical composition, and then quenched in iced water and crushed. The quenched slags were examined by XRD to ensure glassy phase, as shown in Fig. 1. The fluorine content in the slag was determined by ion-selective electrode method. The contents of other components were analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES). The chemical compositions of the studied slag are shown in Table 1.

Fig. 1.

XRD patterns of as-quenched slag samples.

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

Sample No.	Chemical composition
	CaF2	CaO	Al2O3	MgO	TiO2
R1	26.3	35.6	31.4	2.5	4.2
R2	25.0	34.7	29.5	2.5	8.4
R3	26.2	31.3	27.5	2.4	12.6
R4	25.8	29.6	25.3	2.5	16.8

2.2. DSC Measurement

The DSC (Netzsch STA449F3; Netzsch Instrument Inc., Germany) experiments were performed under Ar gas atmosphere at the flow rate of 60 mL/min. For each DSC experiment, approximately 50 mg of slag was heated at the rate of 30 K/min from room temperature up to 1773 K in a platinum crucible and held for 1 minute, then cooled at various cooling rates (10 K/min, 15 K/min, 20 K/min and 25 K/min) to 623 K.

2.3. Raman Spectroscopy Measurement

To determine the various functional group structures of slag, the glassy slag was analyzed by Raman spectroscopy (LabRAM HR Evolution, HORIBA, France). The measurements were performed at room temperature using an excitation wavelength of 532 nm. Finally, the Raman spectra were fitted by assuming Gaussian line shapes for the peaks of different structural units.

3. Results and Discussion

The DSC curves at various cooling rates of the studied slag and the analysis of crystalline phase in the slag have been presented in the authors’ previous study.¹⁶⁾ It has been demonstrated that the crystallization tendency of the slag decreases with increasing TiO₂ content in slag. The primary crystalline phases are faceted 11CaO·7Al₂O₃·CaF₂ in the slag with 4.2 mass%-12.6 mass% TiO₂, and faceted 11CaO·7Al₂O₃·CaF₂ and blocky CaTiO₃ in the slag with 16.8 mass% TiO₂. Because that the primary crystalline phase in the slag is the dominant crystalline phase and its precipitation temperature represents the crystallization temperature of the slag, only the first exothermic peaks on DSC curves were employed to analyze the crystallization kinetics. For slag R3, only the DSC curves at the cooling rates of 10 K/min, 15 K/min and 20 K/min were used because the first exothermic peak overlaps partially with the second exothermic peak on DSC curve at the cooling rate of 25 K/min.

3.1. Crystallization Rate Analysis

The relative crystallinity is a basic parameter for analyzing the non-isothermal crystallization kinetics. The relative crystallinity at any crystallization temperature T_i is defined as the following equation:^17,18)   

$$C\left(T\right)=\frac{{\int }_{T_O}^{T_i}\left(\mathrm{d}Hc/\mathrm{d}T\right)\mathrm{d}T}{{\int }_{T_O}^{T_F}\left(\mathrm{d}Hc/\mathrm{d}T\right)\mathrm{d}T}$$(1)

where dHc/dT is the heat flow rate, T_i is the elapsed temperature during crystallization, T_F and T_O is the end and onset crystallization temperature, respectively. The relative crystallinity can be obtained from the DSC thermal data.

Figure 2 presents the relative crystallinity as a function of crystallization time of the slag at various cooling rates. It is shown in Fig. 2 that all the curves exhibit sigmoidal shape. The time required for finishing crystallization process increases with decreasing the cooling rate, suggesting that crystallization proceeds at a lower rate with reducing the cooling rate. A critical parameter that can be estimated from the curves is the crystallization half time (t_1/2), which is defined as the crystallization time corresponding to the relative crystallinity of 0.5.¹⁹⁾ It is clear that the t_1/2 values decrease with increasing cooling rate, indicating that the crystallization rate of the slag increases with increasing the cooling rate.

Fig. 2.

Relative crystallinity vs crystallization time of the slag at various cooling rates. (Online version in color.)

Crystallization consists of both nucleation and crystal growth processes. According to the schematic diagram of nucleation rate and crystal growth rate as a function of temperature,²⁰⁾ the crystallization rate is not a constant during cooling process. To compare the crystallization rate of different slags, the crystallization rate parameter (CRP), which is not dependent on the cooling rate, is determined from the slope of the plots of the reciprocal of crystallization half time (1/t_1/2) vs the cooling rate,^21,22) as shown in Fig. 3. A smaller slope of fitting line means a slower crystallization rate. The CRP values for slag R1, R2, R3 and R4 are 0.0433, 0.0408, 0.0324 and 0.0389, respectively. The calculated CRP values exhibit a decreasing tendency, except CRP value for slag R3 which may be because that the deviation is introduced by fitting only three data points. The decreasing CRP values suggested that increasing TiO₂ addition lowered the crystallization rate of the slag due to the restrained nucleation and growth of the crystalline phase.

Fig. 3.

Plots of reciprocal crystallization half time (1/t_1/2) vs the cooling rate of the slag. (Online version in color.)

3.2. Crystallization Kinetics Analysis

For non-isothermal crystallization kinetic analysis, various models have been developed to reveal the crystallization kinetic of metallurgical slags and polymers.^{24,25,26,27,28,29,30)} The most common description model for non-isothermal crystallization was Mo model,³¹⁾ which combined the Avrami equation^24,25,26) and the Ozawa equation.^27,28) The Avrami equation and Ozawa equation are as follows, respectively:   

$$1-x\left(t\right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(-Z{t}^n\right)$$(2)

  

$$1-C\left(T\right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(-K/{\mathrm{\Phi }}^m\right)$$(3)

where x(t) is the relative crystallinity at crystallization time t; n is the Avrami exponent; Z is the effective crystallization rate constant; C(T) is the relative crystallinity at crystallization temperature T; K is a function related to the crystallization rate; m is the Ozawa exponent; Φ is the heating/cooling rate.

Considering the limitation of the Avrami and Ozawa models for descripting non-isothermal crystallization kinetics,^32,33,34) the Mo model was derived to describe non-isothermal crystallization by combining the Avrami equation and the Ozawa equation, and the Mo equation is as follow:   

$$\mathrm{l}\mathrm{n}\mathrm{\Phi }=\mathrm{l}\mathrm{n}F\left(T\right)-a\mathrm{l}\mathrm{n}t$$(4)

where F(T) refers to the heating/cooling rate Φ that must be selected when reaching unit relative crystallinity in unit crystallization time t, a is the Mo exponent, which is the ratio of the Avrami exponent n to the Ozawa exponent m.

The Ozawa exponent m, which has an identical physical meaning to the Avrami exponent n for isothermal crystallization,³²⁾ is used to determine the crystallization mechanism of the crystal in Mo model, and the corresponding relationship between the value of n and the crystallization mechanism of the crystal are summarized in Table 2. The Ozawa exponent m can be obtained from the Avrami exponent n based on Eq. (2) and Mo exponent a based on Mo Eq. (4), as well as the relationship m = n/a.

Table 2. Value of n for various crystallization mechanism.¹⁷⁾

	Crystallization Mode
	Diffusion controlled	Interface reaction controlled
Constant nucleation rate
3-dimensional growth	2.5	4
2-dimensional growth	2	3
1-dimensional growth	1.5	2
Constant number of nuclei
3-dimensional growth	1.5	3
2-dimensional growth	1	2
1-dimensional growth	0.5	1
Surface nucleation	0.5	1

Figure 4 shows the plots of ln[−ln(1−x(t))] vs ln t at various cooling rates according to Eq. (2). The plots show basically linear relationship between ln[−ln(1−x(t))] and ln t. The fitting straight lines for each slag are nearly parallel for various cooling rates, which indicates that the cooling rate has unobvious influence on the Avrami exponent n for non-isothermal crystallization of the slag. The determined values of Avrami exponent n at various cooling rates are listed in Table 3.

Fig. 4.

Plots of ln[−ln(1−x(t))] vs ln t for non-isothermal crystallization of the slag at various cooling rates. (Online version in color.)

Table 3. Values of Avrami exponent n for non-isothermal crystallization determined by Avrami equation.

Sample No.	Avrami exponent n
	Φ=10 (K/min)	Φ=15 (K/min)	Φ=20 (K/min)	Φ=25 (K/min)	Average value
R1	3.01	3.38	3.41	3.15	3.23
R2	3.17	3.37	3.20	3.44	3.29
R3	3.49	3.16	3.21	–	3.28
R4	3.46	3.87	3.46	3.69	3.62

Based on Mo equation, the plots of ln Φ vs ln t at different relative crystallinity are shown in Fig. 5. It is clear that ln Φ exhibits good linear relationship with ln t, indicating that Mo model can be applied to identify the non-isothermal crystallization kinetics of the studied ESR-type TiO₂-bearing slag. The values of F(T) estimated from the intercept of the fitting line increase with the increasing of the relative crystallinity, suggesting that the higher cooling rate is needed for reaching unit relative crystallinity in unit crystallization time. The values of Mo exponent a determined from the slope of the fitting straight lines in Fig. 5 are listed in Table 4.

Fig. 5.

Plots of ln Φ vs ln t for non-isothermal crystallization of the slag at different relative crystallinity x. (Online version in color.)

Table 4. Values of Mo exponent a at different relative crystallinity x determined from the Mo equation.

Sample No.	Mo exponent a
	x=0.1	x=0.2	x=0.3	x=0.4	x=0.5	x=0.6	x=0.7	x=0.8	x=0.9	Average value
R1	1.37	1.50	1.52	1.46	1.36	1.35	1.29	1.29	1.40	1.40
R2	1.28	1.29	1.32	1.33	1.36	1.38	1.41	1.45	1.47	1.37
R3	1.54	1.49	1.48	1.47	1.47	1.48	1.48	1.46	1.39	1.47
R4	1.85	1.59	1.56	1.58	1.60	1.63	1.67	1.70	1.69	1.65

According to the Avrami exponent n in Table 3 and Mo exponent a in Table 4 as well as the relationship m = n/a, the determined Ozawa exponent m are 2.31, 2.40, 2.23 and 2.19 for slag R1, R2, R3 and R4, respectively. It suggests that the Ozawa exponent m for the four slag samples are approximately 2. Given that constant number nuclei did not occur in the non-isothermal crystallization process, a comparison of the determined values of Ozawa exponent m with theoretical values shown in Table 2 suggested that the crystallization of the primary crystalline phase in the slag might be two possible cases: (1) constant nucleation rate, diffusion controlled and two-dimensional growth; (2) constant nucleation rate, interface reaction controlled and one-dimensional growth. The SEM observations showed the primary crystalline phases in the slag were faceted morphology,¹⁶⁾ indicating that the crystallization of the crystalline phase is controlled by interfacial reaction.^35,36) It can be concluded that the crystallization of primary crystalline phase (11CaO·7Al₂O₃·CaF₂) in the slag with 4.2–12.6 mass% TiO₂ and primary crystalline phase (11CaO·7Al₂O₃·CaF₂ and CaTiO₃) in slag with 16.8 mass% TiO₂ originates from constant nucleation rate, interface reaction controlled and one-dimensional growth. It is similar to the result that the crystallization of glassy mold flux is controlled by interface reaction,³⁷⁾ but different from the findings that the crystallization of slag is controlled by element diffusion.^38,39,40)

3.3. Effect of TiO₂ on the Structure Using Raman Spectroscopy

Figure 6 shows the room temperature Raman spectra of the glassy samples with varying TiO₂ contents. The Raman spectra curves at 500–600 cm⁻¹ and 600–1000 cm⁻¹ are belong to bridged oxygen region and non-bridged oxygen region, respectively. To further understand the change of structural units with TiO₂ content in slag, the Raman spectra whose all backgrounds had been subtracted was deconvolved by Gaussian fitting. The deconvolution results of Raman spectra are shown in Fig. 7, for which all Raman spectra are successfully fitted at the frequency range of 480 to 1000 cm^–1. It can be seen that there are four bands near at 550 cm⁻¹, 740 cm⁻¹, 800 cm⁻¹ and 840 cm^–1, respectively. The Raman band at around 550 cm⁻¹ is assigned to bridged oxygen region within Al–O–Al band.^41,42) The band at about 740 cm⁻¹ corresponds to the [AlO₄]⁵⁻ tetrahedral network units with two bridged oxygen and two non-bridged oxygen per aluminum and is referred to Q² units (superscript refers to the number of bridged oxygen).^41,43) The band at about 800 cm⁻¹ is assigned to the [AlO₄]⁵⁻ tetrahedral network units with four bridged oxygen per aluminum and is referred to Q⁴ units.⁴²⁾ The band at about 840 cm⁻¹ is due to the Ti–O stretch band of ${\text{Ti}}_{\text{2}}{\text{O}}_6^{4-}$ chain unit.⁴⁴⁾

Fig. 6.

Original Raman spectra of slags with different TiO₂ contents. (Online version in color.)

Fig. 7.

Deconvoluted Raman spectra of quenched slags with different TiO₂ contents. (Online version in color.)

The relative height of the peak on the curves corresponding to the bridged oxygen region and non-bridged oxygen region is an indication of the relative fraction of bridged oxygen and non-bridged oxygen in the slag melts. As shown in Fig. 7, the relative height of the Al–O–Al band decreases obviously with increasing TiO₂ content in the slag, indicating that the relatively more complex Al–O–Al band is depolymerized correspondingly. These results suggest that the relative fraction of bridged oxygen decreases and the relative fraction of non-bridged oxygen increases with the addition of TiO₂ in the slag, as well as TiO₂ acting as a network-modifier to decrease the degree of polymerization. The bands near at 740 cm⁻¹, 800 cm⁻¹ and 840 cm^–1 are all located at the non-bridged oxygen region. The area of the peak represents the molar fraction of different structural units in the non-bridged oxygen region. Figure 8 shows the dependence of the relative fraction of individual structural units on the TiO₂ content in the slag. It can be seen that the relative fraction of Q² units and ${\text{Ti}}_{\text{2}}{\text{O}}_6^{4-}$ chain increase from 0.31 to 0.35 and from 0.17 to 0.25, respectively, while the relative fraction of Q⁴ units decreases from 0.52 to 0.39. According to the above analysis, It is concluded that TiO₂ plays a network-modifier role in the relatively more complex Al–O–Al band and Q⁴ units by forming Q² units and less complex ${\text{Ti}}_{\text{2}}{\text{O}}_6^{4-}$ chain unit, resulting in the decreasing of the polymerization degree of CaF₂–CaO–Al₂O₃–MgO–TiO₂ ESR-type slag.

Fig. 8.

Relative fraction of structural units in slag melts with various TiO₂ contents. (Online version in color.)

The current investigation on the slag structure is conducive to clarify the crystallization kinetics of the slag. The aforementioned results demonstrate that the degree of polymerization of the slag decreases with increasing TiO₂ content in slag. In this case, the element diffusion from liquid to interface was accelerated, which was in favor of crystallization. However, the overall crystallization tendency of the studied slag decreases with the addition of TiO₂. Therefore, the crystallization of the primary crystalline phase in the slag were controlled by interfacial reaction instead of element diffusion. Otherwise, the crystallization tendency of the studied slag would increase with increasing TiO₂ content in the slag. The structural analysis is the evidence that the interfacial reaction is the controlling step of crystallization behavior of the studied CaF₂–CaO–Al₂O₃–MgO–TiO₂ ESR-type slag.

4. Conclusions

The crystallization kinetics and structure of the CaF₂–CaO–Al₂O₃–MgO–TiO₂ slag for ESR were investigated. The main conclusions were summarized as follows:

(1) The crystallization rate of the slag increases with increasing the cooling rate of the slag. Increasing TiO₂ content from 4.2 mass% to 16.8 mass% in the slag lowers the crystallization rate of the slag due to the restrained nucleation and growth of the crystalline phase.

(2) The crystallization of the primary crystalline phase (11CaO·7Al₂O₃·CaF₂) in the slags with 4.2–12.6 mass% TiO₂, and primary crystalline phase (11CaO·7Al₂O₃·CaF₂ and CaTiO₃) in slag with 16.8 mass% TiO₂ originates from constant nucleation rate, interface reaction controlled and one-dimensional growth, irrespective of TiO₂ contents in the slag.

(3) The polymerization degree of the slags decreases with increasing TiO₂ content. TiO₂ plays a network-modifier role in relatively more complex Al–O–Al band and Q⁴ units by forming Q² units and less complex ${\text{Ti}}_{\text{2}}{\text{O}}_6^{4-}$ chain unit.

Acknowledgments

The financial support by the National Natural Science Foundation of China (Grant Nos. 51874026 and 51774225), the Fundamental Research Funds for the Central Universities (Grant No. FRF-TP-18-004A3), and Guangdong YangFan Innovative & Entepreneurial Research Team Program (Grant No. 2016YT03C071).

References

• 1)   J. W.  Cho,  H.  Shibata,  T.  Emi and  M.  Suzuki: ISIJ Int., 38 (1998), 268.
• 2)   J. W.  Cho,  H.  Shibata,  T.  Emi and  M.  Suzuki: ISIJ Int., 38 (1998), 440.
• 3)   Y.  Kobayashi,  R.  Maehashi,  R.  Endo and  M.  Susa: ISIJ Int., 53 (2013), 1725.
• 4)   R. J.  Kirkpatrick: Am. Mineral., 60 (1975), 798.
• 5)   X. C.  Lu: Foundry, 51 (2002), 378.
• 6)   C. B.  Shi,  M. D.  Seo,  H.  Wang,  J. W.  Cho and  S. H.  Kim: Metall. Mater. Trans. B, 46 (2015), 345.
• 7)   Z.  Wang,  Q. F.  Shu and  K. C.  Chou: Metall. Mater. Trans. B, 44 (2013), 606.
• 8)   Z.  Wang,  Q. F.  Shu and  K. C.  Chou: ISIJ Int., 55 (2015), 709.
• 9)   J. L.  Klug,  R.  Hagemann,  N. C.  Heck,  A. C. F.  Vilela,  H. P.  Heller and  P. R.  Scheller: Steel Res. Int., 83 (2012), 1186.
• 10)   Z. J.  Wang,  Y. Q.  Sun,  S.  Sridrar,  M.  Zhang and  Z. T.  Zhang: Metall. Mater. Trans. B, 48 (2017), 527.
• 11)   K.  Zheng,  Z. T.  Zhang,  L. L.  Liu and  X. D.  Wang: Metall. Mater. Trans. B, 45 (2014), 1389.
• 12)   H.  Park,  J. Y.  Park,  G. H.  Kim and  I.  Sohn: Steel Res. Int., 83 (2012), 150.
• 13)   S. F.  Zhang,  X.  Zhang,  C. G.  Bai,  L. Y.  Wen and  X. W.  Lv: ISIJ Int., 53 (2013), 1131.
• 14)   J. L.  Li,  Q. F.  Shu and  K. C.  Chou: Can. Metall. Q., 54 (2015), 85.
• 15)   C. B.  Shi,  D. L.  Zheng,  S. H.  Shin,  J.  Li and  J. W.  Cho: Int. J. Miner. Metall. Mater., 24 (2017), 18.
• 16)   D. L.  Zheng,  J.  Li,  C. B.  Shi and  J. T.  Ju: Metall. Mater. Trans. B, 50 (2019), 1148.
• 17)   M. D.  Seo,  C. B.  Shi,  J. Y.  Baek,  J. W.  Cho and  S. H.  Kim: Metall. Mater. Trans. B, 46 (2015), 2374.
• 18)   G. Z.  Papageorgiou,  D. S.  Achilias,  D. N.  Bikiaris and  G. P.  Karayannidis: Thermochim. Acta, 427 (2005), 117.
• 19)   G. S.  Deshmukh,  D. R.  Peshwe,  S. U.  Pathak and  J. D.  Ekhe: Thermochim. Acta, 581 (2014), 41.
• 20)   W.  Höland and  G. H.  Beall: Glass-Ceramic Technology, 2nd ed., Wiley, Hoboken, NJ, (2012), 40.
• 21)   M.  Wu,  G. Z.  Yang,  M.  Wang,  W. Z.  Wang,  W. D.  Zhang,  J. C.  Feng and  T. X.  Liu: Mater. Chem. Phys., 109 (2008), 547.
• 22)   P.  Supaphol,  N.  Dangseeyun and  P.  Srimoaon: Polym. Test., 23 (2004), 175.
• 23)   J. W.  Huang,  H. C.  Hung,  K. S.  Tseng and  C. C.  Kang: J. Appl. Polym. Sci., 100 (2006), 1335.
• 24)   M.  Avrami: J. Chem. Phys., 7 (1939), 1103.
• 25)   M.  Avrami: J. Chem. Phys., 8 (1940), 212.
• 26)   M.  Avrami: J. Chem. Phys., 9 (1941), 177.
• 27)   T.  Ozawa: J. Therm. Anal., 2 (1970), 301.
• 28)   T.  Ozawa: Polymer, 12 (1971), 150.
• 29)   J.  Málek: Thermochim. Acta, 200 (1992), 257.
• 30)   I. W.  Donald: J. Mater. Sci., 30 (1995), 904.
• 31)   T. X.  Liu,  Z. S.  Mo and  H. F.  Zhang: J. Appl. Polym. Sci., 67 (1998), 815.
• 32)   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.
• 33)   M.  Joshi and  B. S.  Butola: Polymer, 45 (2004), 4953.
• 34)   D. W.  Henderson: J. Therm. Anal. Calorim., 15 (1979), 325.
• 35)   J.  Guo,  M. D.  Seo,  C. B.  Shi,  J. W.  Cho and  S. H.  Kim: Metall. Mater. Trans. B, 47 (2016), 2211.
• 36)   J. L.  Li,  Q. F.  Shu,  X. M.  Hou and  K. C.  Cho: ISIJ Int., 55 (2015), 830.
• 37)   L. J.  Zhou,  H.  Li,  W. L.  Wang and  J.  Chang: Metall. Mater. Trans. B, 49 (2018), 3019.
• 38)   S. L.  Xie and  W. L.  Wang: Steel Res. Int., 87 (2016), 376.
• 39)   C. B.  Shi,  J. W.  Cho,  D. L.  Zheng and  J.  Li: Int. J. Miner. Metall. Mater., 23 (2016), 627.
• 40)   J. L.  Li,  B. J.  Yan,  Q. F.  Shu and  K. C.  Chou: Metall. Mater. Trans. B, 46 (2015), 2458.
• 41)   P.  McMillan and  B.  Piriou: J. Non-Cryst. Solids, 55 (1983), 221.
• 42)   M.  Licheron,  V.  Montouillout,  F.  Millot and  D. R.  Neuville: J. Non-Cryst. Solids, 357 (2011), 2796.
• 43)   D. R.  Neuville,  G. S.  Henderson,  L.  Cormier and  D.  Massiot: Am. Mineral., 95 (2010), 1580.
• 44)   B. O.  Mysen,  F. J.  Ryerson and  D.  Virgo: Am. Mineral., 65 (1980), 1150.