Effect of molybdenum and titanium oxides on mechanical and thermal properties of cordierit

Journal of Non-Crystalline Solids 434 (2016) 13–22

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids journal homepage: www.elsevier.com/locate/jnoncrysol

Effect of molybdenum and titanium oxides on mechanical and thermal properties of cordierite–enstatite glass-ceramics Kei Maeda a,⁎, Yoichi Sera a, Atsuo Yasumori b a b

Asahi Glass Co., Ltd., Research Center, 1150 Hazawa-cho, Kanagawa-ku, Yokohama 221-8755, Japan Tokyo University of Science, Department of Materials Science and Technology, Japan

a r t i c l e

i n f o

Article history: Received 7 October 2015 Received in revised form 10 November 2015 Accepted 12 December 2015 Available online 22 December 2015 Keywords: Glass-ceramics Cordierite Enstatite Molybdenum

a b s t r a c t By adding either MoO3 or TiO2 as nucleating agents, glass at the eutectic point of cordierite (2MgO·2Al2O3·5SiO2) and enstatite (MgSiO3) in the MgO–Al2O3–SiO2 ternary system was crystallized. Although bulk crystallization required adding 10 wt% TiO2, approximately 0.1–0.5 wt% of MoO3 induced bulk nucleation when the parent glass was melted under reducing conditions. For both TiO2- and MoO3-nucleated glasses, the glass-ceramics containing both cordierite–enstatite as main crystalline phases were obtained by heat-treatment at 1200 °C. The fracture toughness of the glass-ceramics nucleated by TiO2 was 1.5 times higher than that of sintered cordierite ceramics. In addition to two main crystalline phases, magnesium titanium oxide and rutile also precipitated in TiO2-nucleated glass. The additional effects of TiO2 on the properties of the glass-ceramics were identified by comparing the TiO2- nucleated glass-ceramics and MoO3- nucleated one. The addition of TiO2 slightly increased the Young's modulus of the glass-ceramics. On the other hand, since TiO2 suppressed the precipitation of enstatite, the thermal deformation of TiO2-nucleated glass-ceramic at 1100 °C was larger than that of MoO3nucleated one. Therefore, MoO3 is the preferable nucleating agent of the glass-ceramics for application which requires high thermal endurance. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Glass-ceramics have found many industrial applications, including as cooking ware [1,2], cooktops [1,2], heat-resistant windows for stoves or furnaces [1,2], building walls [2,3], and magnetic disk substrates [4]. Recently, new applications of glass-ceramics have also been proposed such as for the components of portable electronic devices [5]. The glass-ceramic Li2 O–Al2O 3 –SiO 2 (LAS) system, which precipitates from β-quartz solid solutions or β-spodumene (Li2O·Al2O3·6SiO2), is a commercially important glass-ceramic material because of its low thermal expansion and transparency for some specific uses [1,2]. Glass-ceramics in the MgO–Al2O3–SiO2 (MAS) system have also been studied and applied to industrial uses. Cordierite (2MgO·2Al2O3·5SiO2) is one of the most important constituents of MAS glass-ceramics because it has low thermal expansion and is highly refractory. It also has superior electrical properties, such as a low dielectric loss. These superior characteristics of cordierite have attracted significant research efforts in cordierite-containing glass-ceramics. The first commercially successful product involving cordierite glass-ceramics was “Pyroceram” (Corning 9606), which was designed in the 1960s for missile radomes [1,2,6]. In the 1980s, glass-ceramic substrates were studied for use in electrical applications as substitutes for alumina substrates [7]. In recent years, ⁎ Corresponding author. E-mail address: [email protected] (K. Maeda).

http://dx.doi.org/10.1016/j.jnoncrysol.2015.12.001 0022-3093/© 2015 Elsevier B.V. All rights reserved.

cordierite glass-ceramics doped with NiO were investigated for applications involving infrared radiation [8]. Most recently, cordierite–enstatite families of glass-ceramics were proposed for use as substrates for GaN in light-emitting diodes, since those have good thermal stability (to 1100 °C), as well as good thermal expansion matching with GaN [9]. With a fracture toughness of 5 MPa m1/2, enstatite (MgSiO3) is also as one of the toughest glass-ceramics known [10]. Although some of the MAS glass-ceramics mentioned above are made by sintering glass powders [7], bulk-crystallized glass-ceramics still attract significant interest because their parent glass is easy to form by pressing or drawing techniques. For both cordierite and enstatite, TiO2 and/or ZrO2 are often used as nucleating agents for the bulk crystallization of MAS glass-ceramics. The process of nucleation via cordierite or enstatite has been studied extensively [11–13]. Recently, a new approach involving the X-ray absorption near edge structure technique has been used to investigate the nucleation kinetics of MAS glass doped with TiO2 [14] and ZrO2 [15]. Such sustained research into MAS glass-ceramics testifies to the significant long-term scientific and industrial interests that these materials have attracted. One of the issues for glass-ceramics involves conventional nucleating agents, such as TiO2 and ZrO2, is that those are required several wt% [1,6,9–12] for sufficient nucleation of the glass. These agents not only provide nucleating sites for the main crystalline phases but also precipitate themselves in crystalline form in the final material. These

14

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

Fig. 2. Heat treatment for crystallization of glasses. Fig. 1. Phase diagram of MAS [18] showing the glass composition of the present study.

2. Experiment agents are thus expected to affect the microstructure of the resulting glass-ceramics, and therefore, they are likewise expected to have nonnegligible effects on the physical properties of the final product. Unfortunately, identifying these effects has proven to be difficult because the bulk crystalline material in MAS systems is hard to obtain without using such nucleating agents. In a recent study that addressed this question, we found that 0.5 wt% of MoO3 is required to obtain bulk crystallization in MAS glass when the parent glasses were melted in reducing condition [16]. In the present work, to better understand the phase transition and other physical properties of glass-ceramics, we further exploit this phenomenon by comparing MoO3-nucleated glass-ceramics with glassceramics nucleated by conventional nucleating agents. For some applications of the glass-ceramics, the thermal endurance of the material plays an important role [9,17]. Since less residual glassy phase (high crystallinity) is preferable for this purpose, the parent glass composition should be close to stoichiometric composition of the crystalline phase, i.e., cordierite in MAS glass (Fig. 1). However, in that cases the glass composition shows high liquidus temperature that is undesirable for good glass formability. For the present work, we use the glass composition of 55SiO2– 20Al2O3–25MgO (wt%), which is close to the eutectic point between cordierite and enstatite (i.e., 57.3 wt% cordierite + 42.7 wt% enstatite; see Fig. 1 [18]) to ensure low liquidus temperature. In addition, if both cordierite and enstatite precipitate from the glass, the glass can, theoretically, become 100% crystalline. The glasses in this study were nucleated by either TiO2 or MoO3 and crystallized. The creep behavior was evaluated at high temperature for both TiO 2- and MoO3-nucleated glass-ceramics. Density, thermal expansion coefficient, Young's modulus, and fracture toughness of these glassceramics were also measured. The crystallization process and microstructures of the glass-ceramics were examined to understand the effect of these nucleating agents.

2.1. Preparation of parent glasses The samples of parent glass were prepared by the conventional laboratory-scale melting method. The nucleating agents comprised 5, 7.5, and 10 wt% TiO2 or 0.1, 0.2, and 0.5 wt% MoO3. Glass melts (300 g) were obtained from reagent-grade SiO2, Al2O3, MgO, TiO2, or MoO3 as starting materials. For glass containing MoO3, 0.1 wt% of carbon powder was mixed into the glass batches, which were then melted under a reducing atmosphere (oxygen concentration below 1%) by introducing a town gas burner flame into the electric furnace. In the same furnace, TiO2-containing glasses were melted in air. The glass batches were melted at 1550 °C for approximately 2 h in a platinum crucible. The glass compositions and preparation conditions are summarized in Table 1. The glass melts were homogenized by stirring with a platinum stirrer during melting, following which they were poured onto a carbon plate to form slab samples. These samples were annealed at 800 °C for 30 min in another furnace, followed by slow cooling (i.e., 1 °C/min) to room temperature. 2.2. Crystallization process The glass transition temperature Tg and crystallization temperature Tc of the glasses were measured by differential thermal analysis (DTA) with a Bruker TG-DTA 200SA at a heating rate of 10 °C/min. Using glass powders with particle sizes of 212–415 μm. Considering the glass transition temperature and crystallization temperature of the glasses, they were crystallized using the heat treatment depicted in Fig. 2. To acquire detailed data on the early stage of crystallization, the heat treatment of some samples was stopped at 1000 or 1100 °C. These glassceramic samples were cut and polished as required for characterization.

Table 1 Glass compositions and melting conditions used in this study. Sample

A

Glass composition

wt%

mol%

wt%

mol%

wt%

mol%

wt%

mol%

wt%

mo1%

wt%

mol%

Si02 Al2O3 MgO TiO2 MoO3 Total Melting temperature (°C) Melting Atmosphere

52.4 19.0 23.8 4.8

51.0 10.9 34,6 3.5

511 18.6 23.3 7.0

50.2 10.7 34.0 5.1

50.0 18.2 22.7 9.1

49.3 10.6 33.4 6.7

54.9 20.0 25.0

52.9 11.3 35.8

54.9 20.0 24.9

52.9 113 35.8

54.7 19.9 24.9

52.9 11.3 35.7

100 1550 Air

100

100 1550 Air

100

100 1550 Air

100

⁎ Oxygen concentration b 1%.

B

C

D

E

0.1 0.04 100 100.04 1550 Gas burner⁎

F

0.2 0.08 100 100.08 1550 Gas burner⁎

0.5 0.2 100 100.2 1550 Gas burner⁎

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

15

The crystalline phases that precipitated from these glasses were identified by X-ray powder diffraction (XRD; Shimadzu Lab-X XRD-6100) with Cu–Kα radiation. The samples were treated carefully to separate the crystalline phases at the surface from those inside the glasses. For some samples, the intensity of the XRD peaks was measured quantitatively to estimate the amount of material in a given crystalline phase. For these measurements, we used Mg2Si as an internal standard because its XRD peak (JCPDS 35-0773) does not overlap with any peaks from the crystals investigated in this study. Mg2Si was obtained from the Iida group at the Tokyo University of Science. The microstructure of the crystalline samples was observed using an optical microscope (Keyence VHX-500) or scanning electron microscope (SEM). SEM observations were made with two types of SEMs: a Hitachi HighTechnologies TM-3000 (SEM-I) and a Hitachi High-Technologies S4800 (SEM-II). Energy dispersive X-ray spectrometry (EDX) analysis was done using a Thermo Scientific Noran System6 (EDX-I, along with Hitachi High-Technologies SU-6600) or Bruker Xflash 5060FQ (EDX-II, along with High-Technologies SU-8220). For SEM and EDX analysis, the samples were polished but not etched, and coated with carbon to avoid charge buildup. 2.3. Properties of glass-ceramics The density of the glass-ceramics was determined by Archimedes' method. The Young's modulus was measured by pulse-eco with an ultrasonic thickness gauge (Olympus 38DL PLUS). The fracture toughness was measured by the single-edge-V-notched-beam (SEVNB) method [19]. We prepared a minimum of six 3 mm × 4 mm × 50 mm bars. V notches were introduced at the center of each bar using a razor blade with diamond paste (i.e., 1-μm particle size). Fig. 3 shows the typical features of a V notch. Three-point bending test was done with a Shimadzu micro-autograph (MST-I) with a 30-mm span and crosshead speed of 0.5 mm/min. The fracture toughness KIC was calculated using [19]

Fig. 4. DTA curves of glass samples. The glass-transition temperature Tg and crystallization temperature Tc are shown for each curve.

The creep behavior was evaluated by beam-bending method. The bars of 3 mm × 4 mm × 30 mm in size were supported by a 20-mm span, and the 2.5 kg load was applied in the center of the bars. These were kept at 1100 °C, and the deflections at the center of the bars were monitored and recorded.

n h io 1=2 K IC ¼ 3FS= 2ð1−α Þ3=2 Bw2 d Y;

3. Results

where F is fracture load, S is span, B is sample width (3 mm), w is sample height (4 mm), d is the notch depth, and α = d/W = 0.4 to 0.45 for all samples. The stress intensity shape factor Y is calculated using [19].

Fig. 4 displays the DTA curves obtained for the glasses. For all MoO3nucleated glasses, Tg is approximately 777 °C; for TiO2-nucleated glasses, Tg was lesser. The crystallization was observed in between 900 and 1100 °C for all samples.

  Y ¼ 1:9887–1:326α− 3:49–0:68α þ 1:35α 2  α  ð1−α Þð1 þ α Þ−2 :

3.1. Differential thermal analysis of glasses

3.2. Crystallization of TiO2-nucleated glasses For comparison, sintered cordierite (provided by AGC Ceramics Co., Ltd.; Lotec-MD) was also examined. All measurements mentioned above were carried out at room temperature. The thermal expansion coefficient was measured by dilatometer (Rigaku ThermoPlus TMA 8310) from room temperature to 1000 °C.

Glass A (TiO2 = 5 wt%) and B(TiO2 = 7.5 wt%) strongly deformed by heat treatment up to 1000 °C, as shown in Fig. 5. When glass A was heated up to 1000 °C, crystallization occurred at the glass surface. The XRD pattern from the crystals precipitated at the surface of the glass A is displayed in Fig. 6, which also shows the XRD pattern of cordierite

Fig. 3. Notch introduced in center of sample for SEVNB method.

Fig. 5. Glasses A and B heated up to 1000 °C as per Fig. 2.

16

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

Fig. 6. XRD patterns of crystalline phase from glass A and cordierite obtained from AGC Ceramics Co.

obtained from AGC Ceramics Co., Ltd. (Lotec-M). These data confirm that the main crystalline phase observed at the surface of glass A is cordierite (JCPDS 12-0303). The large deformation is likely induced by the different thermal expansions of the surface cordierite and internal glass. Fig. 7 displays the internal microstructure of glass B heated up to 1000 °C. Although it showed bulk crystallization, the crystals were coarse due to poor nucleation. Conversely, glass C, which was doped with 10 wt% TiO2, deformed little when heated up to 1000 °C, but developed a homogeneous internal microstructure with internal crystal grains less than 10 μm in size, as shown in Fig. 8. Therefore, we further investigated glass C since it clearly showed bulk crystallization, Fig. 9 shows XRD patterns from the crystals precipitated in glass C, which was heat treated at 1000–1200 °C according to the heattreatment schedule shown in Fig. 2. The crystalline phases of enstatite (JCPDS 19-0768) and magnesium aluminum titanium oxide (MAT, MgAl2Ti3O10: JCPDS 05-0450) were detected in the glasses heat treated at 1000 and 1100 °C, whereas cordierite (JCPDS 12-0303), magnesium titanium oxide (MT, MgTi2O5: JCPDS 35-0796), and rutile (JCPDS 211276) were detected in addition to enstatite in the glasses heat treated at 1200 °C. Fig. 10 shows the microstructure of glass C heated up to 1100 °C, as imaged by a high-resolution SEM (SEM-II). Two crystals (plane dendrite and needle-like) are apparent in the SEM image. EDX mapping done with EDX-I (Fig. 11) reveals that the dendrite crystals are Mg rich and

Al poor, which allows us to identify them as enstatite, consistent with the XRD pattern shown in Fig. 9. However, because the Ti concentration is high in the needle-like crystals, we identify these to be MAT crystals, which is also consistent with the XRD results, as shown in Fig. 9. Fig. 12 shows the SEM-II images of the microstructure of glass C heat treated at 1200 °C shown in Fig. 2. The crystalline phases that develop at this temperature are rather complicated compared with those that develop at 1100 °C, as seen in the XRD results of Fig. 9, so we used EDXII with a higher resolution to analyze these materials (see Fig. 13). Unlike the glass heated up to 1100 °C, no dendritic enstatite (i.e., Mg-rich area) appears in the glass heat treated at 1200 °C (Fig. 12). The analysis by EDX-II of the darkest area in the SEM-II image (Fig. 12) reveals it to be Al rich (Fig. 13). Thus, these results combined with the XRD results identify this phase as cordierite crystal. Two different morphologies appear in the bright region in the SEM-II image (Fig. 12): slender grains about 1 μm long, and small spherical particles less than 1 μm in diameter. Although analysis by EDX-II indicates that both morphologies are Ti-rich phases (Fig. 13), each morphology contains different concentrations of Mg. Because the slender grains contain Mg, we identify these to be MT crystals, again consistent with the XRD results. However, we identify the small particles as rutile, because the Mg concentration is very low in these particles. Analysis by EDX-II also reveals the presence of some silica-rich phases. However, because no corresponding XRD pattern was produced, we consider this silica-rich phase to be a residual glassy phase.

Fig. 7. Optical micrograph of the early stage of crystallization (left) and SEM-I image (right) of glass B heated up to 1000 °C as per Fig. 2.

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

17

3.4. Evaluation of crystallinity

Fig. 8. SEM-I image of early stage of crystallization (glass C, heated up to 1000 °C as per Fig. 2).

3.3. Crystallization of MoO3-nucleated glass Fig. 14 shows internal microstructures of the MoO3 containing glasses D, E, and F heated up to 1000 °C according to the heat treatment schedule shown in Fig. 2. All exhibited bulk crystallization upon heat treatment. Since we observed the finest microstructure occurs for glass F, we investigated the crystalline phase transition of glass F at the different heat treatment temperatures shown in Fig. 2. XRD patterns of the crystalline phase precipitated in glass F are displayed in Fig. 15. For the glasses heated up to 1000 and 1100 °C, only enstatite was detected, whereas cordierite was also detected as a secondary phase upon heat treatment at 1200 °C. The measurement results of glass F made with SEM-II, EDX-I, and EDX-II are shown in Figs. 16–19 and revealed that Mg-rich crystals with a dendrite morphology precipitated at 1100 °C (Figs. 16 and 17), which is similar to what occurred in TiO2-nucleated glass (see Fig. 10). From the XRD results shown in Fig. 15, we identify these dendrite crystals as enstatite. At the center of the enstatite, small particles less than 20 nm in size are apparent. However, these particles are too small to determine their constituents by EDX. Upon heat treatment at 1200 °C, the dendrite crystals (enstatite) decompose (Fig. 18), and Al is enriched in the Mg-poor area (Fig. 19). We identify the Al-rich area as cordierite, consistent with the XRD pattern shown in Fig. 15. The small, bright particles (several tens of nm in size) are composed of molybdenum, and their oxygen concentrations are extremely low (Fig. 19). As for the TiO2-nucleated glass C, silica-rich phases are observed in glass F (Fig. 19).

To compare the amount of the dominant crystalline phases (cordierite and enstatite) that precipitate in both TiO2-nucleated glass C and MoO 3 -nucleated glass F, we analyzed the two quantitatively using XRD. The step scan profiles of the diffraction are displayed in Fig. 20. The peak areas at 2θ = 29.4° (cordierite) and 2θ = 31.2° (enstatite) were normalized by the peak area at 2θ = 24.2° (Mg2Si) and compared in Fig. 21. Although the amount of cordierite precipitated in glass C and glass F were almost identical, the amount of enstatite for glass C was less than that of glass F. Also for both glass C and glass F, the amount of enstatite decreased when the glasses were heat treated at 1200 °C. This result is consistent with the observation by SEM-II that the dendrite morphology of enstatite disappears after this heat treatment. Since a pure cordierite crystal was obtained (see Fig. 6), we created a calibration curve to estimate how much cordierite precipitates in glasses C and F. The three quantities of cordierite (25, 50, and 75 wt%) were mixed with the powders of the parent glass (glass F), and 10 wt% of Mg2Si to the total mixture was added. Their diffraction intensities were measured in the same way as described above. The obtained calibration curve is shown in Fig. 22. Because the normalized intensities of the XRD peaks of cordierite precipitated in glasses C and F are both 2.5 (Fig. 21), we estimated from Fig. 22 that the crystalline fraction of cordierite is 40 wt%. This result is lower than the theoretical value for the glass composition (57.3 wt%).

3.5. Properties of glass-ceramics The density, Young's modulus, and fracture toughness, and linier thermal expansion coefficient (25–1000 °C: CTE) of the glass-ceramics heat-treated at 1200 °C for 2 h are summarized in Table 2. The Young's modulus of glass-ceramic C was slightly higher than that of glassceramic F. The CTE of enstatite–cordierite containing glass-ceramics was higher than that of cordierite ceramics. This is reasonable because the CTE of the glass-ceramics precipitating enstatite as a main crystalline phase is as high as 8 × 10−6/°C [10]. Since these values are close to the required ones for GaN substrate [9], these glass-ceramics can be good candidates for this application. Besides, the higher fracture toughness of the glass-ceramic C than commercial cordierite ceramics ensures the availability of this material. Unfortunately, the fracture toughness of the glass-ceramic F was not obtained because those samples included

Fig. 9. XRD patterns of crystalline phases precipitated in glass C for various heat treatments as per Fig. 2.

18

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

Fig. 10. SEM-II image of glass C heated up to 1100 °C as per Fig. 2.

Fig. 11. EDX-I area mapping of glass C heated up to 1100 °C as per Fig. 2.

some cracks during crystallization so that the sample did not break from the V notches but broke from the cracks by the bending test. We next evaluated the creep behavior of the glass-ceramics containing both enstatite and cordierite after heat treatment at 1200 °C for 2 h. The results are illustrated in Fig. 23. As seen in Fig. 23, the deflection of glass-ceramic F (MoO3 nucleated) was less than that of glass-ceramic C (TiO2 nucleated). The result indicates that glass-ceramic C easily deforms at high temperature. 4. Discussion Bulk-crystallized glass-ceramics that precipitate enstatite and cordierite as their crystalline phases were obtained by heating the parent

glasses at 1200 °C for 2 h and using two different nucleating agents: TiO2 and MoO3. Magnesium aluminum titanium oxide (MAT), magnesium titanium oxide (MT), and rutile also appeared in the TiO2nucleated glass-ceramics. The introduction of TiO2 in MAS glass is reported to promote phase separation [11] and lead to the precipitation of nanocrystalline MAT [14] upon heat treatment at nucleation temperatures, and rutile at higher temperature [12]. In the present study, a similar reaction occurred, although the chemical composition of the parent glass was slightly different from that in previous studies. In this study, obtaining a glass-ceramic with a fine microstructure required adding 10 wt% TiO2 (7.2 mol%). Beall reported that adding 9 wt% (6.9 mol%) TiO2 is required for internal nucleation of Corning 9606 [20]. In a recent study on cordierite stoichiometric composition by Cormier

Fig. 12. SEM-II image of glass C heat treated at 1200 °C as per Fig. 2.

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

19

Fig. 13. EDX-II area mapping image of glass C heat treated at 1200 °C as per Fig. 2.

et al. [14], bulk crystallization was induced when the glass contained more than 8 mol% of TiO2. Thus, the similar behavior was confirmed in this study in terms of threshold in TiO2 content for an active nucleation. Cormier et al. [14] also reported that the glass transition temperature decreases according to the addition of TiO2 into the glass as we observed in this study. Based on these results, we infer that TiO2 stably dissolves

in MAS glasses with up to 7–8 mol%, and bulk nucleation occurs when the TiO2 concentration exceeds its solubility. In contrast, adding only 0.1 to 0.5 wt% MoO3 induced bulk nucleation. High-resolution SEM combined with EDX indicates the presence of small particles of molybdenum in glass-ceramic F heat treated at 1200 °C (see Fig. 19). Since oxygen concentrations are poor in these

Fig. 14. SEM-I image of early stage of crystallization in glasses containing MoO3 (heated up to 1000 °C as per Fig 2).

Fig. 15. XRD patterns of crystalline phases precipitated in glass F.

20

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

Fig. 16. SEM-II image of glass F heated up to 1100 °C as per Fig. 2.

particles, we consider that they are composed of metallic molybdenum. In the glass-ceramic heat treated at 1100 °C, similar but smaller particles appeared at the center of the enstatite dendritic crystals (see Fig. 16). These observations support our conclusions from a previous study [16] that metallic molybdenum particles with a wide size distribution provide heterogeneous nucleating sites. Because metallic substances do not dissolve well in silicate, a sufficient nucleating effect could arise even for a small concentration of metallic molybdenum particles. Because of the small quantity of nucleating agent, the crystalline phases in the MoO3-nucleated glass heat treated at 1200 °C are simply enstatite and cordierite; therefore, the phase assemblies corresponding to the MAS phase diagram are easy to identify. Our analysis of the XRD results estimates the amount of cordierite to be 40 wt%, although the parent glass can precipitate 57.3 wt% cordierite +42.7 wt% enstatite theoretically. If we assume that the parent glass precipitates out 40 wt% of stoichiometric cordierite, a simple calculation shows that the chemical composition of the parent glass becomes 34.4SiO2–6.1Al2O3–19.5MgO (wt%), which is corresponding to 33.2SiO2–3.4Al2O3–27.9MgO (mol%). Based on the analysis of the EDX results, the precipitated enstatite does not seem to be pure MgSiO3 but contains some amount of Al2O3 (Fig. 19). The Al2O3 likely enters into enstatite by the substitution Mg2 + + Si4 + → Al3 + + Al3 + [21]. If Al2O3 and MgO are entirely consumed as constituents of enstatite, 27.9SiO2–3.4Al2O3–27.9MgO (mol%) of enstatite precipitates, and 5.3 mol% SiO2 remains as residual glassy phase. This corresponds to 5 wt% of SiO2 in the original parent glass. Therefore, we can assume that the glass-ceramic comprises

40 wt% cordierite, 55 wt% enstatite, and 5 wt% SiO2 (residual glassy phase). The small thermal deformation of the glass-ceramic F shown in Fig. 23 supports this estimated high crystallinity. On the other hand, precipitation of enstatite is obviously less for TiO2- nucleated glass-ceramic C as compared with MoO3- nucleated one as shown in Fig. 21. Therefore, the residual glass in glass-ceramic C is likely more than that in glass-ceramic F. Moreover, TiO2 tends to dissolve in the glass and lower its glass transition temperature (i.e., lower the viscosity) as discussed above. The fracture toughness of the glass-ceramics C is rather high compared with that of pure cordierite ceramics, as shown in Table 2. This result seems reasonable because the enstatite glass-ceramic has a high fracture toughness (i.e., 5 MPa m1/2), as described in Section 1. The Young's modulus of the TiO2- nucleated glass-ceramic C is slightly higher than that of MoO3- nucleated one, because rutile has a high Young's modulus due to oxygen close-packed structure [22,23], and also TiO2 is known as a component which increases the Young's modulus of glass. [24]. Thus, by studying glass-ceramics with MoO3 as nucleating agent, we identified not only the intrinsic phase assemblies and properties of enstatite–cordierite glass-ceramics, but also the additional effect of TiO2. Addition of TiO2 is advantageous to increase Young's modulus of the glass-ceramics, however, it suppresses the precipitation of enstatite, which results in the low thermal endurance. Therefore, MoO3 is preferable nucleating agent of the glass-ceramics for application which requires high thermal endurance. Based on the results obtained in this study, the

Fig. 17. EDX-I area map of glass F heated up to 1100 °C as per Fig. 2.

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

21

Fig. 18. SEM-II image of glass F heat treated as per Fig. 2 at 1200 °C.

Fig. 19. EDX-II area map of glass F heat treated as per Fig. 2 at 1200 °C.

glass-ceramics with further improved properties are expected to be developed in future. 5. Conclusion We obtained bulk-crystallized cordierite–enstatite eutectic glass by adding only 0.1–0.5 wt% of MoO3 as nucleating agent and melted the

Fig. 20. Step-scan profiles of glasses C and F, both heat treated at 1200 °C as per Fig. 2. As an internal standard, 10 wt% of Mg2Si was added to each glass.

glass under reducing condition, although 10 wt% of TiO2 was required for sufficient nucleation. During crystallization, enstatite precipitated at 1000–1100 °C; later, some enstatite decomposed and cordierite precipitated upon further heat treatment at 1200 °C, for both MoO3- and TiO2-nucleated glasses. In addition to the two main crystalline phases (i.e., cordierite and enstatite), magnesium titanium oxide and rutile also precipitated in the TiO2-nucleated glass. Owing to small concentration of MoO3 as a nucleating agent, the additional effects of TiO2 were revealed. The introduction of TiO2 slightly increased the Young's modulus of the glass-ceramics. Since introduction of TiO2 suppressed the

Fig. 21. Intensity of diffraction peak of enstatite and cordierite normalized to that of Mg2Si.

22

K. Maeda et al. / Journal of Non-Crystalline Solids 434 (2016) 13–22

high-resolution SEM–EDX analysis. We would also like to thank Professor Iida at Tokyo University of Science for providing the Mg2Si for the XRD standard and Mr. Kouichi Kanda at AGC Ceramic Co., Ltd. for providing the cordierite ceramic as a reference for the mechanical properties. References

Fig. 22. Calibration curve for cordierite. Vertical axis shows normalized XRD peak intensity and horizontal axis shows the cordierite content in glass F.

Table 2 Physical properties of glass-ceramics C and F and sintered cordierite ceramic. “MT” is magnesium titanium oxide.

Density (g cmd) Thermal expansion Coefficient (K−1) Young's modulus (GPa) Kic by SEVNB method (MPa m1/2) Crystalline phases

Glass-ceramic C

Glass-ceramic F

Cordierite ceramic

2.84 6.2 × 10−6

2.77 6.1 × 10−6

2.55 1.8 × 10−6

149 ± 2 2.30 ± 0.12

142 ± 2 –

138 ± 2 1.40 ± 0.08

Cordierite, enstatite, MT, rutile

Cordierite, enstatite

Cordierite

precipitation of enstatite and lowered the glass transition temperature of the glass, it reduced thermal endurance of the glass-ceramics due to large amount of residual glassy phase and/or its low viscosity.

Acknowledgments The authors are grateful to Mr. Atsushi Miyaki at Hitachi High Technologies and Mr. Iwao Yamazaki at Bruker AXS for their help with the

Fig. 23. The deflection at 1100 °C of glass-ceramics. The specimens are 3 mm × 4 mm × 30 mm bars. These were supported with 20 mm-span and 2.5 kg load was applied at the center of the bars.

[1] Chapter 7, pyroceram, in: Z. Strnad (Ed.), Glass-Ceramic Materials, Elsevier 1986, pp. 190–195. [2] M. Montazerian, S.P. Singh, E.D. Zanotto, An analysis of glass-ceramic research and commercialization, Am. Ceram. Soc. Bull. 94 (4) (2015) 30–35. [3] Chapter 7, Slagsitall, in: Z. Strnad (Ed.), Glass-Ceramic Materials, Elsevier 1986, pp. 230–234. [4] N. Goto, K. Yamguchi, Glass-ceramic for magnetic disks and method for manufacturing the same, US patent 5391522 A (1995). [5] G.H. Beall, et al., White, opaque, β-spodumene/rutile glass-ceramic articles and methods for making the same, US patent 2014/0057092 A1 (2014). [6] Grossman et al., Glass-ceramic compositions designed for radomes, US patent 4304603 (1981). [7] S.H. Knickrbocker, A.H. Kumar, L.W. Herron, Cordierite glass-ceramics for multilayer ceramic packaging, Am. Ceram. Soc. Bull. 72 (1993) 90–95. [8] S. Wang, K. Liang, Crystallization behavior and infrared radiation property of nickel– magnesium cordierite based glass–ceramics, J. Non-Cryst. Solids 354 (2008) 1552–1555. [9] G.H. Beall, J.G. Couillard, S. Marjanovic, Glass-ceramic substrates for semiconductor processing, US patent 2014/0001485 A1 (2014). [10] G.H. Beall, Chain silicate glass-ceramics, J. Non-Cryst. Solids 129 (1991) 163–173. [11] 3.4 The SiO2–Al2O3–MgO system, in: Z. Strnad (Ed.), Glass-ceramic materials, Elsevier 1986, pp. 97–101. [12] T.I. Barry, J.M. Cox, R. Morrell, Cordierite glass-ceramics — effect of TiO2 and ZrO2 content on phase sequence during heat treatment, J. Mater. Sci. 13 (1978) 594–610. [13] W. Hutton, J.S. Thorp, The vibrational spectra of MgO–Al2O3-SiO2 glasses containing TiO2, J. Mater. Sci. 20 (1985) 542–551. [14] L. Cormier, O. Dargaud, N. Menguy, Investigation of the role of nucleating agents in MgO–SiO2–Al2O3–SiO2–TiO2 glasses and glass-ceramics: a XANES study at the Ti Kand L2, 3-edges, Cryst. Growth Des. 11 (1) (2011) 311–319. [15] O. Dargaud, G. Calas, L. Cormier, L. Galoisy, C. Jousseaume, G. Querel, M. Newville, In Situ study of nucleation of zirconia in an MgO–Al2O3–SiO2 glass, J. Am. Ceram. Soc. 93 (2010) 342–344. [16] K. Maeda, A. Yasumori, Effect of molybdenum and tungsten oxides on nucleation and crystallization behaviors of MgO–Al2O3–SiO2 glasses, J. Non-Cryst. Solids 427 (2015) 152–159. [17] H. Charles, I.I.I. Drummond, Glass formation and crystallization in high-temperature glass-ceramics and Si3N4, J. Non-Cryst. Solids 123 (1990) 114–128. [18] E.M. Levin, C.R. Robbins, H.F. McMurdie (Eds.), Fig. 712, Phase Diagrams for Ceramists, vol. 1, The American Ceramic Society 1964, p. 246. [19] J. Kubler, Fracture toughness of ceramics using the SEVNB method; round robin, VAMAS Report, No. 37, ESIS Document D2-99, 1999. [20] G.H. Beall, Design and properties of glass-ceramics, Annu. Rev. Mater. Sci. 22 (1992) 91–119. [21] P. Anastasiou, F. Seifert, Solid solubility of Al2O3 in enstatite at high temperatures and 1–5 kb water pressure, Contrib. Mineral. Petrol. 34 (1972) 272–287. [22] W.D. Kingery, H.K. Bowen, D.R. Uhlmann, Oxide structures in: introduction to ceramics, fourth ed. John Wiley & Sons, 1975 62. [23] D.H. Chung, W.R. Buessem, The Voigt–Reuss–Hill (VRH) approximation and the elastic moduli of polycrystalline ZnO, TiO2 (rutile), and α–Al2O3, J. Appl. Phys. 39 (6) (1968) 2777–2782. [24] 3.5.1 Elastic properties, Glass Nature, Structure, and Properties, Springer-Verlag 1990, pp. 246–254.