Journal Crystallization of 60Sio2–20Mgo–10Al2o3–10Bao Glass Ceramics

J. Am. Ceram. Soc., 88 [8] 2249–2254 (2005) DOI: 10.1111/j.1551-2916.2005.00337.x Journal Crystallization of 60SiO2–20MgO–10Al2O3–10BaO Glass Ceramics

E. Manor Department of Advanced Materials Engineering, Jerusalem College of Engineering, Jerusalem, Israel

R. Z. Shneckw Department of Materials Engineering, Ben-Gurion University of the Negev, Beer Sheva, Israel

Three distinctly different microstructures of silica (as quartz and of glass ceramics, that of uniformly dispersed small crystals, crystobalite), alumina, enstatite, and celsian, were found to de- is obtained by the intentional introduction of nucleating 1,2,4,12,13 velop in a 60SiO2–20MgO–10Al2O3–10BaO glass ceramic. At agents, or by choosing compositions prone to liquid- 10101C, growth of wormy fibrillar crystals was observed, indi- phase separation at certain temperatures. The small dispersed cating that crystal growth was diffusion controlled. At the droplets of the second liquid phase enhance the nucleation of intermediate temperature of 10801C, a coarse cellular microcrystals of similar composition.3,14 Otherwise, a great variety of structure developed with multiple spherical particles nucleated morphologies are observed. Diffusion-controlled growth is en- on their surfaces and in the surrounding glass. At 12001C, the countered when the crystallization takes place at low tempera- glass crystallizes in a denderitic morphology but the dendrites tures. In these cases, fine fibrillar crystals typically grow in were actually fragmented into multiple cube-shaped enstatite spherulitic,15 rosette4 or cellular5 morphologies. At higher tem- crystals, indicating a transition to interface-controlled growth. peratures, growth becomes interface controlled. A high rate of The crystals coarsen with time but maintain their order along growth takes place parallel to preferred crystal planes, forming the dendrite skeletons. complex denderitic morphologies.4,12,13 At still higher temperatures, large faceted crystals grow.4,16 The formation of several phases may take place in a sequential order10,11 or in a co-pre- cipitation process, forming interpenetrating structures.5 The in- I. Introduction vestigation of the transitions from one morphology to the other XIDE melts in the silica-rich corner of the SiO2–MgO– may promote understanding of the mechanisms of the pattern OAl2O3–BaO system tend to form glasses and certain com- formation. positions crystallize relatively easily upon subsequent heat treat- In the present paper, the development of distinctly different ment. Many well-known glass ceramic compositions, including microstructures of essentially the same three phases in a 60SiO2– commercial ones, are in fact subsystems of the SiO2–MgO– 20MgO–10Al2O3–10BaO glass is explored. The transition from Al2O3–BaO system. These materials are based on the beneficial one morphology to the other is brought about by changing the properties of two or three oxide crystalline phases: aluminum crystallization temperature in the absence of nucleating agents. silicate (mullite), magnesium silicates (enstatite and forsterite), aluminum–magnesium silicate (cordierite) barium silicates, and barium–aluminum silicate (celsian).1–8 Several four-component glass ceramics have been prepared.9–11 Chaim and Heuer10 crys- II. Experimental Procedure tallized a 52SiO2–13MgO–32Al2O3–2BaO glass and observed SiO2,MgO,BaO,Al2O3 powders of 99.9% purity were obtained the primary growth of denderitic m-cordierite that were con- from Merck Co (Darmstadt, Germany). The raw materials were sumed by secondary nucleation of denderitic b-cordierite. The thoroughly dry mixed and melted in 5 cm3 aluminosilicate cru- barium remained in the residual glass. Winter et al.11 have crys- cibles at 15501C in air for 2 h. After melting, the crucibles were tallized a 68SiO2–8MgO–22Al2O3–1.6BaO glass (and other cooled in still air. The resulting glass in one crucible was milled compositions). They found that the first phase that forms is a to powder, and samples of 10 mg were heated in an STD 2960 b-quartzss; then, small lamellae of Ba-osumilite grow in the re- TA Instrumentst thermo gravimetric-differential thermal anal- sidual glass, which provide nucleation sites for plate-shaped ysis (TG-DTA, New Castle, DE) apparatus at different rates. magnesia–alumina–silica osumilite. The MAS-osumilite subse- Three reaction temperatures were identified by the DTA anal- quently became the major phase. Referring to observations of ysis: 10101,10801,and12001C. The melted glass samples in the Chaim and Heuer,10 Winter et al.11 anticipate that only in a crucibles were subjected to recrystallization treatments in air for narrow range of BaO concentration will the Ba-osumilite and 1 2 h to 48 h. Specimens were cut by a diamond wheel, mounted, MAS-osumilite form. ground by 320, 400, and 600 mesh silicon carbide papers, and The occurrence of a sequence of phase changes during the polished by 12.5, 5, and 0.05 mm alumina slurry. Then they were crystallization of glasses is generally understood. A small dif- etched in 5% HF solution for 2–4 min and investigated by a Jeol ference in the free energy between silicate phases of similar com- 35C scanning electron microscope (Tokyo, Japan, SEM). The position may result in the formation of kinetically preferred whole contents of other crucibles were milled to powder for X- intermediate metastable phases, before the thermodynamic sta- ray diffraction analysis, performed with a Cu tube in a Philips ble phase appears.11 The interpretation of the development of apparatus (Almelo, The Netherlands). microstructure in glass ceramics is more difficult since it involves many competing kinetic processes. The desirable microstructure III. Results R. E. Loehman—contributing editor (1) DTA The heating DTA curves of the raw glass is shown in Fig. 1(a). Manuscript No. 10788. Received January 9, 2004; approved February 10, 2005. Two exothermic deflections in the curve can be distinguished. wAuthor to whom correspondence should be addressed. e-mail: [email protected] The first begins at 10101C and ends at 10801C; the second begins 2249 2250 Journal of the American Ceramic Society—Manor and Shneck Vol. 88, No. 8

Fig. 1. (a) DTA proﬁles of 10mg glass powder in the particle size range of 50–150 mm heated at a rate of (I) 10, (II) 20, and (III) 301C/min and a fourth sample (IV) of particle size 200–600 mm heated at a rate 101C/min. (b) Ozawa plot used to determine the activation energies. (c) Kissinger plot. at 12001C. These temperatures were chosen for crystallization particle surfaces. However, the two reactions differ by their experiments. thermal behavior: while the peak temperature Tp of the ﬁrst The two reactions are considerably affected by the particle reaction varies by 39 K in the range of the applied heating rates, size, indicating that the crystallization mainly initiates at the Tp of the second reaction did not change within the experimental

Fig. 2. SEM micrographs showing ﬁbrillar microstructure of the specimens crystallized at 10101C. (a) 8 h, (b) 16 h, and (c, d) 32 h. August 2005 Crystallization of Glass Ceramics 2251

Fig. 3. SEM micrographs showing cellular microstructure of the specimens crystallized at 10801Cfor8h. error and its height decreased with increasing heating rate. The Crystallization at 12001C took place uniformly in the whole modified Ozawa equation17 was applied to the first reaction to volume of the samples and was completed after half hour as deduce Ec, the activation energy: assessed visually. Well-developed dendrites are revealed by electron microscopy (Fig. 4). Most of the dendrites are elongated; ln a ¼mEc=nRT P þ constant (1) the others are short but very branched. Higher magnifications reveal that the dendrites are fragment- a is the heating rate, n is the Avrami constant, and m is the ed into a large number of small crystals that are arranged along crystal growth dimensionality. For surface crystallization m 5 1 denderite skeletons. The small crystals are ordered along and n is also taken as unity. Fig. 1(b) shows that the activation straight lines in the two-dimensional sections. This order hints energy is 24079 kJ/mole. Applying the Kissinger equation:17 at the fact that the three-dimensional skeleton of the dendrites, followed by the small crystals, is retained up to advanced stages n 2 of the crystallization. Apparently numerous nucleation events ln ða =Tp Þ¼mEc=RTp þ constant (2) takes place in front of the growing denderitic branches, and yields 26079 kJ/mole. The activation energy of the reaction these events forms the denderitic skeleton The small cube- taking place at 12001C is very small. shaped crystals are ordered along lines but their orientations are not regular. This observation indicates that the nucleation events associated with a single denderitic skeleton are random (2) Appearance and Microstructure and not directly connected to ‘‘their’’ single dendrite. The Crystallization at 10101C was slow. It began with isolated spher- synergistic/cooperative nucleation is probably due to composit- ical crystallization centers, nucleated on the crucible walls and ional changes in front of the dendrites. on the free surfaces of cracks in the glass. The crystallization After 32 h at 12001C, Ostwald ripening of the small crystals centers grew and coalesced after 16 h. Microstructural exami- resulted in a reduction of the density of the small ones and nation revealed radial fibrillar growth of crystals in spherulites growth of the coarser crystals. They form groups of crystals with (Fig. 2). Coarsening is observed after 32 h; the crystals are nee- common orientations as depicted in Figs. 4(e) and (f). dle-shaped with undulated interface. At 10801C, the crystallization is cellular (Fig. 3). The cell boundaries appear as aggregates of small spherical particles. (3) Phase Composition Apparently, numerous crystals were nucleated on the bounda- X-ray diffraction of the crystallized glass revealed that four ries, grew into small spherical crystals, and tended to coalesce as compounds are formed: SiO2,(as quartz and cristobalite) Al2O3, crystallization proceeded. MgSiO3 (enstatite), and BaAl2Si2O8 (celsian) (Figs. 5–7). The 2252 Journal of the American Ceramic Society—Manor and Shneck Vol. 88, No. 8

1 1 Fig. 4. SEM micrographs showing microstructure of the specimens crystallized at 1200 C: (a–c) different magniﬁcations of dendrites crystallized for 2 h; (d–f) Different magniﬁcations of large crystals after 32 h.

ﬁrst three phases are found after crystallization at 10101Cfor have similar shapes (Figs. 2 and 3). It was not possible to dis- over 16 h. Longer treatment allows for more enstatite and tinguish by EDS microanalysis between the different phases re- alumina to grow. At 10801C, significant amount of alumina vealed by the diffraction. EDS showed only that the crystals are ﬁrst forms but almost disappears after 16 h. Celsian continu- enriched with magnesium relative to the glassy phase. ously grows at this temperature and is practically absent at the other temperatures. At 12001C, again, only enstatite and quartz IV. Discussion are revealed (with traces of celsian). The microstructure illus- trated in Figs. 4(e–f) consists of clusters of large plate-shaped The formation of three different morphologies as a result of and cube-shaped crystals. EDS microanalysis reveals that all the the crystallization of silica, alumina, enstatite, and celsian by large crystals have the same composition conforming to ensta- annealing at three different temperatures was observed in the tite. The microcrystals grown in each of the other specimens present study. August 2005 Crystallization of Glass Ceramics 2253

Fig. 7. X-ray diffraction patterns of glass crystallized at 12001C: (a) 0.5 h and (b) 32 h.

Fig. 5. X-ray diffraction patterns of glass crystallized at 10101C: (a) 16 h and (b) 48 h. A, corundum (JCPDS 10-173); Cr, cristobalite (39-1425); Growth of wormy fibrillar crystals was observed at 10101C. celsian (14-551); E, enstatite (19-768); Q, quartz (33-1161). This morphology is typical of crystals grown at low temperatures. At these temperatures, the glass viscosity is high and the rate of diffusion is small. Fine morphology, allowing rapid solute redistribution, develops when diffusion is the rate-control- ling step in the transformation. At 12001C, the glass crystallizes in a denderitic morphology. The highly branched shape of the dendrites indicates an unstable manner of growth. Denderitic growth is considered to be mainly a diffusion-controlled process.18 Each perturbation on a dendrite surface increases the gradi- ent of solute concentrations ahead of it, thereby increasing the rate of solute diffusion and the constitutional supercooling in the glass, which causes the growth to be unstable. The size scale of the denderitic branches is determined by a compromise between the competing tendencies to increase diffusion rates and to reduce surface energy.18 The denderites observed in Fig. 4 clearly exhibit the anisotropy of the crystal growth: the denderitic branches grow along preferred (crystyallograpic) direc- tions and the fragments of the dendrites are faceted due to growth parallel to preferred crystallographic planes. This indicates a way of reducing the surface energy of the denderitic fine branches but probably also a transition to an interface-controlled crystallization due to the increasing importance of the surface reaction. The DTA analysis of this reaction at 12001C confirms that it has a different mechanism and a much smaller activation energy relative to the reactions at the other temperatures. Prolonged exposure to 12001C causes the dendritic structure to disintegrate into individual faceted crystals, preserving the dendritic skeletons. Two processes are probably taking place: (a) grain coarsening derived by the tendency to reduce the large in- terfacial energy of the large number of dendritic fragments1,2,4 and (b) transition to the thermodynamically more stable phase, enstatite. The absence of a barium-containing crystalline phase in silicate glasses containing as much as 10 wt% BaO was previously observed.6 Winter et al.11 suggested that only in a Fig. 6. X-ray diffraction patterns of glass crystallized at 10801C: (a) 1 h narrow range of BaO concentrations Ba-osumilite and MAS- and (b) 32 h. osumilite will form. It is interesting to note that the present 2254 Journal of the American Ceramic Society—Manor and Shneck Vol. 88, No. 8 observation is in agreement with these findings. Hence, silica Z. Literman, H. Benjamin, J. Kirshenbuam, E. Shagie, A. Linwald, G. Pelleg, and polymorphs may exist as metastable solid solutions as suggested D. Ephroni. by Chaim and Heuer.10 At the intermediate temperature of 10801C, a coarse cellular References microstructure developed with multiple spherical particles on 1 the cell surfaces and probably in the surrounding glass (Fig. 3). P. W. Mcmillan, Glass Ceramics, 2nd edition, Academic Press, London, 1979. 2A. I. Berezhnoi, Glass Ceramics and Photo-Sitalls, trans. by S. A. Mersol, Ple- The spherical shape of the particles indicates that the surface num Press, New York, 1970. reaction is less important than at 12001C, since the driving force 3W. D. Kingery, H. K. Bowen, and D. R. Uhlmann, Introduction to Ceramics, for crystallization is larger. We can suggest one of the following 2ndedition;368pp.JohnWiley,NewYork,1976. 4 explanations for the observed morphology: M. H. Lewis, J. Metcalf-Johansen, and P. S. Bell, ‘‘Crystallization Mechanisms in Glass–Ceramics,’’ J. Am. Ceram. Soc., 62, 278–88 (1979). (1) Liquid-phase separation takes place at a temperature of 5M. A. Mccoy and A. H. Heuer, ‘‘Microstructural Characterization and 10801C. The growing cells preferably incorporate the liquid Fracture Toughness of Cordierite–ZrO2 Glass Ceramics,’’ J. Am. Ceram. Soc., droplets of a closer composition. Liquid-phase separation is 71, 673–7 (1988). 4 6 typically found at intermediate temperatures. N. P. Bansal and M. J. Hyatt, ‘‘Crystallization Kinetics of BaO–Al2O3–SiO2 Glasses,’’ J. Mater. Res., 4, 1257–64 (1988). (2) The nucleation rate of bulges on the cell surfaces is 7W. E. Lee, M. Chen, and P. James, ‘‘Crystallization of Celsian Glass,’’ J. Am. large but their growth rate is small at the relatively low temper- Ceram. Soc., 78, 2180–6 (1995). atures. Hence, long cells are growing with multiple spherical 8D. Bahat, ‘‘Kinetic Study on the Hexacelsian–Celsian Phase Transformation,’’ branches. J. Mater. Sci., 5, 805–10 (1970). 9K. P. Gadkaree and K. Chyung, ‘‘Silicon Carbide Whisker Reinforced Glass and Glass Ceramic Composites,’’ Am. Ceram. Soc. Bull., 65, 370–6 (1986). 10R. Chaim and A. H. Heuer, ‘‘Crystallization in a Barium Containing V. Summary Magnesium Aluminusilicate Glass Ceramic,’’ J. Am. Ceram. Soc., 75, 1512–21 DTA analysis revealed two exothermic surface crystallization (1992). 11W. Winter, A. Berger, G. Muller, and W. Pannhorst, ‘‘Crystallization Mech- reactions with different activation energies. However, crystalli- anism of MAS-osumilite with Composition Mg2Al4Si11O30 from Glass,’’ J. Am. zation (up to 12001C, up to 48 h) of 60SiO2–20MgO–10Al2O3– Ceram. Soc., 76, 1837–43 (1993). 10BaO glass ceramics has revealed three distinctly different 12T. J. Headley and R. E. Loehman, ‘‘Crystallization of a Glass Ceramic by microstructures of silica (quartz and crystabolite): alumina, ens- Epitaxial Growth,’’ J. Am. Ceram. Soc., 67, 620–5 (1984). 13V. Maier and G. Muller, ‘‘Mechanism of Oxide Nucleation in Lithium Alu- tatite, and celsian. Crystal growth at low temperature (10101C) minosilicate Glass Ceramics,’’ J. Am. Ceram. Soc., 70, 976 (1987). is fibrillar, indicating that the growth is diffusion controlled. The 14G. H. Beall and D. A. Duke, ‘‘Glass Ceramic Technology,’’ p. 403 in Glass crystallization is cellular at 10801C. At the highest temperature Science and Technology, Vol. 1 Edited by D. R. Uhlmann and N. J. Kreidl, Ac- (12001C), fragmented dendrites grow and disintegrate into mul- ademic Press, New York, 1983. 15M. H. Lewis and G. Smith, ‘‘Spherulitic Growth and Recrystallization in tiple cube-shaped crystals, indicating that crystallization be- Barium Silicate Glasses,’’ J. Mater. Sci., 11, 2015–26 (1976). comes interface controlled. 16K. Watanabe, E. A. Giess, and M. W. Shafer, ‘‘The Crystallization Mecha- nism of High Cordierite Glass,’’ J. Mater. Sci., 20, 508–15 (1985). 17K. Matusita and S. Sakka, ‘‘Kinetic Study on Crystallization of Glass by Dif- Acknowledgments ferential Thermal Analysis—Criterion on Application of Kissinger Plot,’’ J. Non- Cryst. Solids, 38–39, 741–6 (1980). We gratefully acknowledge Dr. D. V. Hivascu for initiating the re- 18J. S. Langer, ‘‘Instabilities and Pattern Formation in Crystal Growth,’’ Rev. search and for the work carried out by O. Nabutovsky and by our students Modern Phys., 52, 1–28 (1980). &