Recycling of Blast Furnace Slag and Fluorite Tailings into Diopside-Based Glass-Ceramics with Various Nucleating Agents’ Addition

Diopside-based glass-ceramics are successfully produced by recycling blast furnace slag and fluorite tailing with the addition of supplementary elements such as TiO2, Fe2O3 and Cr2O3 as nucleation agents, using a conventional melting method. The effects of various nucleating agents on the phase components and structure of the prepared glass-ceramics were evaluated by a differential scanning calorimeter, X-ray diffraction and scanning electron microscope–energy disperse spectrometer methods to determine the optimal dosage of nucleating agents. The results show that, in the preparation of diopside-based glass-ceramics, the suitable percentages of blast furnace and fluorite tailing are 55% and 45%, and the recommended composite nucleating agents consist of 1.5% Cr2O3, 2% TiO2 and 3% Fe2O3. Heat treatment was conducted at a nucleation temperature of 720 °C and a crystallization temperature of 920 °C, and the nucleation and crystallization durations were 1.0 h and 1.5 h, respectively. Under the abovementioned parameters, the obtained diopside-based glass-ceramics displayed a Vickers hardness of 7.12 GPa, density of 2.95 g·cm−3, water absorption of 0.02%, acid resistance of 0.23% and alkali resistance of 0.02%.


Introduction
Blast furnace slags are the major by-products in the ironmaking and steelmaking process, and its output is approximately 0.3 tons per ton of pig iron. In China, approximately 200 billion tons of blast furnace slag are generated per year, 40% of which are subjected to landfilling without further utilization [1][2][3]. Fluorite ore is widely distributed in South Africa, Mexico, and China, and its total reserves exceed 240 million tons [4]. Fluorite tailing is a typical industrial solid waste generated from the flotation process of fluorite ore, and its improper disposal not only occupies a large area of land, but also results in a serious contaminative source. Therefore, the recycle and comprehensive utilization of blast furnace slags and fluorite tailings are significant in order to solve the considerable waste of resources and environmental pollution.
Due to the major components in metallurgy slag (SiO 2 , CaO, MgO and Al 2 O 3 ) matching well with the requirements of glass-ceramics, recycling metallurgy slag, especially blast furnace slag, to prepare glass-ceramics through a melting or sintering method has been consistently explored since the 1960s [5][6][7][8][9]. However, considering the basicity of blast furnace slag, the addition of SiO 2 is necessary to guarantee the formation of glass phases during the high-temperature treatment. Previous literature [10][11][12] has reported that quartz sand, clay, kaolin and pure silica could be selected as proper supplements due to the relatively higher content of SiO 2 during the blast furnace slag glass-ceramics preparation. Besides, some researchers put forward that glass-ceramics could be prepared by blast furnace slag with the addition of other industrial solid waste. Yang et al. [13][14][15][16] investigated the preparation of glass ceramics from blast furnace slag and multi-solid wastes including fly ash, waste glass, red mud, potash feldspar, etc. These research results would be of great interest regarding the potential industrial application of the utilizing industrial solid wastes. With the addition of multi-solid wastes in glass-ceramics preparation, the addition of nucleating agents is also necessary to promote crystallization and improve the physical and chemical properties of the products. Wang et al. [17][18][19][20][21][22] investigated the effect of a single nucleating agent as well as composite nucleating agents on glass-ceramics and demonstrated that the addition of the proper nucleating agent could effectively improve the physical and chemical properties of glass-ceramics. Although the preparation of glass-ceramics from solid wastes indicates a promising future, the fluctuations in chemical composition and the large amounts of impurity in elements seriously limit the usage percentage of solid wastes in glass-ceramics production plants in China.
The diopside-based glass-ceramic in the CaO-MgO-Al 2 O 3 -SiO 2 (CMAS) system is one of the most widely applied glass-ceramics due to its high strength, high corrosion and high wear resistance [23][24][25]. The main components of blast furnace slag are CaO, MgO, Al 2 O 3 and SiO 2 , which favor diopside formation by adjusting the CaO/SiO 2 ratio in the glass network [26]. The main components of fluorite tailing are SiO 2 and CaF 2 , which have been proven to be effective in crystallization by reducing the viscosity of the crystallizing glass [27,28]. Therefore, mixtures of blast furnace slag and fluorite tailing are suitable for diopside-based glass-ceramics. Our previous research demonstrated that fluorapatite-based glass-ceramics could be obtained from blast furnace slag and fluorite tailing, which exhibit excellent bioactive and mechanical properties [29].
In this study, blast furnace slag and fluorite tailing are employed to prepare diopside glass-ceramics with the addition of different nucleating agents. Firstly, glass-ceramics without the addition of a nucleation agent were investigated to determine the appropriate blast furnace slag and fluorite tailing percentages. Next, the effects of a single nucleation agent, namely TiO 2 , Fe 2 O 3 and Cr 2 O 3 , on glass-ceramics were investigated to obtain the suitable dosage of the nucleation agent. Then, the addition of composite nucleating agents was explored by conducting orthogonal experiments to optimize the properties of diopsidebased glass-ceramics. This research aims to elucidate a low-cost preparation method for diopside-based glass-ceramics from typical solid wastes and provide theoretical support for the efficient utilization of blast furnace slag and fluorite tailing.

Raw Materials and Experimental Design
The raw materials used in this study include blast furnace slags and fluorite tailings, which are taken from an iron and steel company and ore treatment plant in China. In this article, the selected nucleating agents are TiO 2 , Fe 2 O 3 and Cr 2 O 3 . Considering the purity of the added nucleating agents, the reagent-grade TiO 2 (≥99.0%, average particle size ranging 0.1~0.3 µm), Fe 2 O 3 (≥96.0%, particle size less than 5 µm) and Cr 2 O 3 (AR) powders were purchased from the Aladdin company and used in the preparation of glass-ceramic samples via the melting method. The chemical compositions of the blast furnace slags and fluorite tailings are listed in Table 1. The XRD patterns of the blast furnace slag and fluorite tailing are shown in Figure 1. As typical water-quenched slag, there are no obvious diffraction peaks in the XRD patterns of blast furnace slag due to its amorphous characteristics. In fluorite tailing, the major phases are SiO 2 , CaF 2 and K 2 O·Al 2 O 3 ·6SiO 2 . In the CaO-MgO-Al 2 O 3 -SiO 2 (CMAS) system, the diopside-based glass-ceramic exhibits high strength, excellent corrosion and wear resistance. Considering the major components in the blast furnace slag and fluorite tailing, the molar ratios of CaO, SiO 2 , MgO and Al 2 O 3 were controlled in area 11 to guarantee the precipitation of diopside crystals, as shown in Figure 2. The detailed experimental scheme, according to the main crystalline phase designed in Figure 2, is listed in Table 2. The blast furnace slag percentage ranges from 45% to 65%. Correspondingly, the fluorite tailing percentage ranges from 55% to 35%. The nucleating agents selected in this study are TiO 2 , Fe 2 O 3 and Cr 2 O 3 , and the addition percentages are listed in Table 2.

Preparation Process of Glass-Ceramics
The conventional melting method was used to prepare the glass-ceramics from blast furnace slag and fluorite tailing. The detailed preparation process is shown in Figure 3. Firstly, the blast furnace slag, fluorite tailing and nucleating agent powders were mixed homogeneously through ball-milling. During the ball-milling process, the ball-milling duration was 30 min and the ball-milling speed was 400 r/min. The ball-to-material mass ratio was 3.0. Next, the mixtures were charged into a platinum crucible, heated in a muffle furnace to 200 • C with a heating rate of 5 • C/min and kept constant at this temperature for 1 h to remove the moisture. Then, the mixtures were heated to 1500 • C at a heating rate of 5 • C/min and left at this temperature for 2 h until the material completely melted and all bubbles disappeared. After that, a small part of the molten material was quenched with water, and the nucleation and crystallization temperatures were determined by a differential scanning calorimeter (DSC). The remaining samples were poured into a preheating mold, and then annealed at 600 • C. Then, the heat treatment was conducted, wherein the nucleation and crystallization durations were 1.0 h and 1.5 h, respectively. After heat treatment, the glass-ceramic samples were cooled down to room temperature naturally in the furnace to release internal stress.

Characterization Methods and Performance Testing
The heat treatment temperatures significantly affect the nucleation and crystallization of glass-ceramics. In this study, the DSC technique was used to obtain the nucleation and crystallization temperatures with a heating rate of 10 • C/min in a nitrogen atmosphere (NETZSCH STA 449F5, NETZSCH, Waldkraiburg, Germany).
The crystalline phases were identified using X-ray diffraction (XRD, Ultima IV, Rigaku, Japan, Cu Kα, 10-90 • ) with a scanning speed of 2 • /min. The fractured surfaces of the glassceramic samples were chemically etched for 3-5 s in a 4 vol % HF solution to observe the microstructures using scanning electron microscopy (SEM, SU5000, Hitachi, Japan). Meanwhile, the elemental compositions were determined using energy-dispersive spectroscopy (EDS, X-MAX 20, Oxford, UK).
The properties of the obtained glass-ceramics include Vickers hardness, density, water absorption, and acid and alkali resistance. The Vickers hardness of the glass-ceramics was measured using an automatic turret microhardness tester (HV-1000IS, Shanghai Institute of Optics and Fine Mechanics). Firstly, the glass-ceramics sample was ground and polished. Then, Vickers diamond indenters were used to press the test sample. The load was 500 g and the loading time was 10 s. After the test force was removed, the diagonal length of the indentation was measured. The Vickers hardness was calculated according to Equation (1).
where HV is the Vickers hardness (MPa), P is the loading (N), α is the angle between two opposite faces of Vickers diamond indenters and d is the average length of the diagonal indentation (mm). The water absorption of the glass-ceramics sample was tested by the following methods. Firstly, the sample was dried to a constant weight. Then, the samples were soaked in distilled water at room temperature for 48 h and taken out. The moisture on the surface of the samples was wiped off with a wrung-out wet towel, and the mass was immediately weighed. The water absorption rate was calculated according to Equation (2).
where Wa is the water absorption (%), m 0 is the mass of the sample before immersion (g) and m 1 is the mass of the sample after immersion (g).
In the testing process for the chemical-resistant properties, the sample was dried to a constant weight. Then, the sample was immersed in the prepared sulfuric acid solution (1 vol.% chemical pure concentrated sulfuric acid) or sodium hydroxide solution (1.0 wt% chemical pure sodium hydroxide), wherein the liquid level was 30 mm higher than the sample, and the container mouth was sealed. It was soaked for 650 h, then taken out and washed with deionization water until the pH value was neutral. After drying, it was weighed. The mass loss rate was calculated according to Equation (3) to evaluate the acid or alkali corrosion resistance.
where K is the mass loss rate (%), w 0 is the mass of the sample before immersion (g) and w 1 is the mass of the sample after immersion (g). Figure 4 shows the DSC curves of the parent glass with different percentages of blast furnace slag and fluorite tailing. In the DSC testing, the parent glass samples A1-A5 were heated with a heating rate of 10 K/min under an N2 protective atmosphere. Generally, Tg is defined as the glass transition temperature, which is considered the nucleation starting temperature. Due to the indistinct exothermic peak in the DSC curves, the differential of the DSC curves was introduced to determine the location of the indistinct exothermic peak at Tg. Exothermic peaks associated with glass transition temperatures (Tg) were observed between 676 • C and 700 • C. Sample A5 exhibits the highest Tg, which may be attributed to the lowest fluorite tailing percentage among the five batches. The exothermic peaks of crystallization (Tc) for sample A1 appeared around 886 • C. For samples A2 to A5, there were two different exothermic peaks of crystallization that could be observed, implying a double crystallization process occurred and the crystallization was facilitated by fluorite tailing [29]. According to the literature [27], the nucleation and crystallization temperatures for heat treatment schedules should be set (20-100) • C higher than Tg and Tc. Considering the powder sample used in DSC testing performed at a lower crystallization boundary energy, the difference value was (10-40) • C. Therefore, the nucleation temperature selected was at 710 • C, and the double crystallization temperatures were set at 900 • C and 980 • C. Figure 5 presents the morphologies of samples A1 to A5 after heat treatment. The surface and cross-section appearances indicate that large amounts glass phases existed in the glass-ceramic samples.  Figure 6 shows the XRD patterns of samples A1 to A5 after heat treatment. Though the diopside phase (Ca(Mg,Al)(Si,Al) 2 O 6 ) peaks could be observed in the five samples, the diffraction peak intensity was insufficient, which indicates that diopside glass-ceramics preparation is difficult when only using blast furnace slag and fluorite tailing without the addition of other nucleating agent. In samples A4 and A5, small amounts of a wollastonite phase (CaSiO 3 ) appeared, which is mainly due to the higher blast furnace slag percentage decreasing the SiO 2 content and facilitating the precipitation of crystals with a lower Si/O ratio. In sample A3, the diopside diffraction peak intensity was higher relative to the other samples. Therefore, sample A3, containing a blast furnace slag percentage of 55% and a fluorite tailing percentage of 45%, was selected as the basic case for the subsequent research considering the effect of different nucleating agents.  Figure 7 shows the DSC curves of the parent glass with different TiO 2 content. In the DSC testing, the parent glass samples with different TiO 2 content were heated at a heating rate of 10 K/min under an N 2 protective atmosphere. With an increase in TiO 2 addition from 0 to 5%, the parent glass transition temperature Tg clearly decreased from 683 • C to 664 • C. This is because Ti 4+ , exhibiting strong field intensity, would exist as [TiO 6 ] or [TiO 4 ] in the glass and destroy the glass network structure, thereby decreasing the stability of the glass phase. Compared with the A3 sample, the crystallization temperature Tc decreased first and then increased. The shape of the crystallization exothermic peak in the Ti3 sample was quite sharp, which indicates the appropriate TiO 2 addition could improve crystallization ability. Based on the Tg and Tc obtained from the DSC curves, the nucleation and crystallization temperatures were set at 690 • C and 900 • C, respectively.  Figure 8 shows the XRD patterns, crystallinity and morphologies of the glass-ceramics with different TiO 2 content after heat treatment. It should be pointed out that the crystallinity of the glass-ceramics in the present study was analyzed by the multi-peak separation method through Jade software. With the addition of TiO 2 , the main crystalline phases in the glass-ceramics were anorthite CaAl 2 Si 2 O 8 and Ti-bearing diopside (Ca,Ti)(Mg,Al)(Si,Al) 2 O 6 . In sample Ti1, small amounts of anorthite and Ti-bearing diopside peaks appeared, and the crystallinity in this case was only 30.2%. When the TiO 2 content was increased to 3% and 5%, the intensity of CaAl 2 Si 2 O 8 and (Ca,Ti)(Mg,Al)(Si,Al) 2 O 6 peaks noticeably increased, and the corresponding crystallinity also increased to 45.9% and 50.2%. Therefore, TiO 2 addition could effectively facilitate crystallization. Besides, the glass-ceramics samples with TiO 2 addition exhibited structural integrity without cracks.  Figure 9 shows the SEM-EDS analysis of the glass-ceramics with different TiO 2 content. As shown in Figure 9a, large amounts of glass phases existed in sample Ti1 due to its low crystallinity. From the magnification image in Figure 9b, many crystal grains can be seen distributed in the glass phases. As shown in Figure 9b, large amounts of columnar crystals were precipitated when increasing the TiO 2 content to 3%, indicating one-dimensional volumetric crystallization occurred. As shown in the elemental distribution images of sample Ti3, the Ti distribution overlapped with Mg and Al, which is consistent with the precipitation of Ti-bearing diopside (Ca,Ti)(Mg,Al)(Si,Al) 2 O 6 from XRD analysis results. Therefore, it can be deduced that [TiO 4 ] and [TiO 6 ] may act as nucleation particles and promote crystallization. In sample Ti5, the size of columnar crystals increased noticeably with uneven distribution, as shown in Figure 9c. In view of the crystallinity and crystal distribution, the content of TiO 2 added should be less than 3%.  Figure 10 shows the DSC curves of the parent glass with different Fe 2 O 3 content in the range of 1 wt% to 5 wt%. Compared with the A3 sample, the nucleation temperature Tg and crystallization temperature Tc increased first and then decreased. Simultaneously, the steepness of the crystallization peak first increased and then decreased with Fe 2 O 3 addition. The influence of Fe 2 O 3 on the glass system could be divided into two different parts [30]. On the one hand, Fe 3+ entered the glass phase in the form of [FeO 4 ] 5− , which could enhance the viscosity of glass and suppress crystallization. On the other, with a further increased Fe 2 O 3 content, a fraction of Fe 3+ occupied the octahedral position in the form of [FeO 4 ] 5− as well as Fe 2+ , which would promote crystallization and decrease the nucleation temperature Tg and crystallization temperature Tc. Based on the Tg and Tc obtained from the DSC curves, the nucleation temperature and crystallization temperature were set at 720 • C and 910 • C, respectively.    Figure 12 shows the SEM-EDS analysis of the glass-ceramics with different Fe 2 O 3 contents. A double-layer structure of glass-ceramics was observed and analyzed, as shown in Figure 12a,b. In the outside area of the Fe1 sample, large amounts of diopside crystals appeared. Furthermore, in the inner-side area, the crystals gradually disappeared, and the glass phase still existed. As shown in Figure 12b, the interface between the crystal area and the glass phase area was obvious, indicating that surface crystallization occurred with Fe 2 O 3 addition. When the Fe 2 O 3 content was increased to 3%, both the crystal size and distribution were unform, indicating that 3% Fe 2 O 3 addition is appropriate for preparing diopside glass-ceramics from blast furnace slag and fluorite tailing. In the Fe5 sample, the growth of columnar crystals was facilitated, and the agglomeration of columnar crystals gradually appeared, as shown in Figure 12d. In the elemental distribution images, Fe was distributed homogenously in the crystal and glass phases, which is consistent with the previous literature [31].  Figure 13 shows the DSC curves of the parent glass with different Cr 2 O 3 content. When the Cr 2 O 3 addition was increased from 0 to 4%, the parent glass transition temperature Tg noticeably increased from 683 • C to 715 • C. The crystallization temperature Tc noticeably decreased because crystallization was effectively facilitated by the Cr 2 O 3 introduced. Further increasing the Cr 2 O 3 from 2% to 4%, the crystallization temperature decreased, and the exothermic peak strength was gradually weakened due to the fact that the extra Cr 2 O 3 addition could increase the viscosity of glass and block the atomic diffusion, which indicates that Cr 2 O 3 addition is excessive in the sample Cr3 and Cr4. Previous research demonstrated that the prevention of Cr 2 O 3 precipitating in the glass matrix as a form of chromium spinel crystallites during the cooling stage of glass is mainly responsible for the reduction of viscosity. In contrast, the extra Cr 2 O 3 could promote the precipitation of chromium spinel crystals and increase the viscosity, thereby restraining the bulk nucleation and subsequent crystallization processes [22,32]. Based on the Tg and Tc obtained from the DSC curves, the nucleation temperature and crystallization temperature were set at 720 • C and 910 • C, respectively.  Figure 14a shows the XRD patterns of the glass-ceramics with different Cr 2 O 3 content after heat treatment. In sample Cr1, no diffraction peaks appeared and the sample was still amorphous. When the Cr 2 O 3 content was increased to 2% and 3%, the diopside phase Ca(Mg,Al)(Si,Al) 2 O 6 noticeably precipitated. In sample Cr4, small amounts of NaFe(SiO 3 ) 2 and KFeSiO 4 phase also precipitated as well as the diopside phase. Noticeably, the intensity of the diffraction peaks was weakened compared with Cr2 and Cr3, which is because higher Cr 2 O 3 content significantly increased the viscosity of glass and suppressed the crystals' precipitation. As shown in Figure 14b, crystallinity perceptibly increased with an increase in the Cr 2 O 3 content from 1% to 2%, and then decreased slightly with the increase in Cr 2 O 3 content from 2% to 4%, reaching the maximum value of 53% with a Cr 2 O 3 content of 2%.  Figure 15 shows the SEM-EDS analysis of the glass-ceramics with different Cr 2 O 3 contents. In sample Cr1, there were no crystals observed, and large amounts of glass phases existed, which is consistent with the XRD analysis results. In sample Cr2, large amounts of granular crystals precipitated with a particle size of 1-2 µm, indicating threedimensional volumetric crystallization occurred. Significantly, the granular crystals' size and distribution were uniform, resulting in a compact structure of diopside glass-ceramic. When the Cr 2 O 3 content was increased to 3%, the viscosity and nucleation points increased, which facilitated the growth of granular crystals and also accelerated the agglomeration of granular crystals. In sample Cr4, the atomic diffusion was further limited by the higher viscosity of glass, and the chromium-rich phases noticeably gathered as shown in the elemental distribution images. As a result, the crystals' growth mode transformed from three-dimensional volumetric crystallization to one-dimensional volumetric crystallization and the columnar crystals gradually precipitated in sample Cr4. In view of the crystallinity and microstructure, the recommended Cr 2 O 3 addition content is 2%.

Optimization of Composite Nucleating Agent Addition
Based on the effect of a single nucleating agent on the diopside glass-ceramic, the addition of composite nucleating agents was investigated by conducting orthogonal experiments. The detailed selected levels of the composite nucleating agents are listed in Table 3. Figure 16 shows the DSC curves of the parent glass with the addition of different composite nucleating agents. The parent glass transition temperature ranged from 659 • C to 678 • C, and the crystallization peaks were located in the temperature range of 868 • C to 888 • C. In case 2 and case 3, the crystallization temperatures were relatively lower, and the shape of the crystallization peaks was quite sharp, which indicates the two cases have a better potential crystallization ability. Based on the DSC analysis results, the nucleation temperature and crystallization temperature were set at 720 • C and 920 • C, respectively.   Figure 17a shows the XRD patterns and crystallinity of the glass-ceramics with composite nucleating agents. In the nine batch samples, the diopside phase precipitated as the main and single crystalline phase, which indicates the addition of composite nucleating agents would not alter the crystal species. The crystallinity of the nine batch samples, shown in Figure 17b, ranged from 60% to 65%. In case 3, the crystallinity reached 64% and 65%, respectively.  Figure 18 presents the SEM images of glass-ceramics with the addition of different composite nucleating agents. In case 1, case 6 and case 8, the bulky crystals appeared with different sizes, which tends to produce a porous structure of the obtained glass-ceramics. In case 3, case 4, case 5, case 7 and case 9, large amounts of granular crystals precipitated and agglomerated, which results in an uneven crystals distribution. In case 2, the crystals' size decreased to 0.2~0.5 µm. Simultaneously, the granular distribution was more uniform, and the grain boundary cracks diminished gradually. In view of the structure, crystal size and crystal distribution homogeneity, the composite nucleating agents in case 2 are obviously better than the others. To evaluate the glass-ceramics prepared from blast furnace slag and fluorite tailing, properties including Vickers hardness, density, water absorption, acid resistance and alkali resistance were tested and are listed in Table 4. The best properties of the glass-ceramics were obtained in case 2, which is consistent with the abovementioned XRD and SEM analysis results. Therefore, in the preparation of diopside-based glass-ceramics, the appropriate percentages of blast furnace and fluorite tailing are 55% and 45%, and the recommended composite nucleating agents consist of 1.5% Cr 2 O 3 , 2%TiO 2 and 3% Fe 2 O 3 . Under this condition, the diopside glass-ceramics displayed a Vickers hardness of 7.12 GPa, density of 2.95 g·cm −3 , water absorption of 0.02%, acid resistance of 0.23% and alkali resistance of 0.02%.
Cimdins et al. [1] tested the physical and chemical properties of the glass-ceramics prepared from industrial waste, such as metallurgical slag, fly ash, etching refuse, peat, coal ash, glass waste and so on. The reported data include water uptake of 0.34-3.23 wt%, a final density of 2.93-3.05 g/cm 3 , and bending strength of 80-96 MPa. The material containing only waste had a durability (mass loss) of 3.02% in 0.1 N HCl. It can be seen that the physical and chemical properties of the glass-ceramics in our study are close to the data reported in the literature. Specifically, the acid and alkali resistance of the glass-ceramics prepared from blast furnace slag and fluorite tailings is better than that in the reported data.

Conclusions
The effects of nucleating agents on the diopside-based glass-ceramics prepared from blast furnace slag and fluorite tailing were investigated in this paper. The following conclusions could be drawn from this study.
(1) The crystallization characteristics and mechanisms correlate to the addition of various nucleating agents. With the addition of Fe 2 O 3 (ranging between 1 wt% and 5 wt%) or TiO 2 (ranging between 1 wt% and 5 wt%), the parent glass realized surface crystallization, while the addition of Cr 2 O 3 (ranging between 1 wt% and 4 wt%) promoted the transformation of the crystallization mode to three-dimensional volumetric crystallization. (2) The diopside-based glass-ceramics could be prepared from 55% blast furnace and 45% fluorite tailing with the addition of composite nucleating agents consisting of 1.5% Cr 2 O 3 , 2% TiO 2 and 3% Fe 2 O 3 . The heat treatment was conducted at a nucleation temperature of 720 • C and a crystallization temperature of 920 • C, and the nucleation and crystallization durations were 1.0 h and 1.5 h, respectively. Institutional Review Board Statement: Not applicable.