MATERIALS iEnE& EGMEERNG A SEVIER laterals Science and Engineering A328(2002)267-276 www.elsevier.comflocate/msea Stress-induced阝B→∝- cristobalite phase transformation in (Na2O+Al2O3)-codoped silica Chin-Hsiao Chao, Hong-Yang Lu* Institute of Materials Science and Engineering, National Sun Yat-Sen Unicersity, Kaohsiung 80424, Taiwan Received 9 April 2001: received in revised form 25 June 2001 Abstract Colloidal gel-derived silica(SiO,) powder codoped with(Na, 0+Al,O3 was sintered at 1100C. The crystalline phase content and phase transformation of the sintered ceramics have been studied via X-ray diffractometry and scanning electron microscopy The amount of B-cristobalite retained metastably in the mixture to room temperature is found to depend on the level of additives Samples codoped with Na, O and Al,O3, both of 6.30 mol%o, were found to contain only B-cristobalite in the crystalline mixture. and which is known as the chemically stabilized cristobalite(CSC). Multiple liquid phase separation is also observed in the codoped samples. The lattice spacing dola of a-cristobalite increases with the doping level while dib of the B-phase remains almost unchanged in all compositions studied. Surface grinding or pulverizing of the sintered samples into powder induces the B-a-cristobalite phase transformation. The mechanism of the high temperature B-cristobalite stabilization to room temperature associated with both the chemical and mechanical terms is discussed. o 2002 Elsevier Science B.v. all rights reserved Keywords: Sintering: Nucleation and growth; Phase transformation 1. Introduction perature B-cristobalite is a truly dynamic disordered phase [3]. Others [4, 5]. however, proposed models Crystalline silica(SiO2)experiences a series of poly- which suggested that the p-structure is composed of morphic phase transformation from cristobalite to low-symmetry domains tridymite, and to quartz upon cooling to room temper The stabilization of B-cristobalite can occur [6,7 by ature in atmospheric pressure. Other crystalline phases (a) altering chemical composition, and/or(b)imposing of stoichovite and coesite also exist under high pres- mechanical constraints In the first case, it appears that sures. Both the higher temperature polymorphs of acceptor impurities in solid solution with silica stabilize cristobalite and tridymite are known to exist metastably the p-phase in a manner similar [6] to that of the fully to room temperature in atmospheric pressure, although stabilized cubic (c)-ZrO2 [8]. The former named accord quartz is believed to be the thermodynamically most ingly the chemically stabilized cristobalite(CSC)has stable phase under such conditions. It remains uncer been synthesized [9-ll] by incorporating stuffing tain if the stabilization of cristobalite and tridymite in catons of Ca+ cu+ and sr+ into the tectosilicate natural minerals, which often incorporate impurity ox framework, or alternatively [ll] by directly replacing ides, is imparted by cation substitutions for silicon. The the Sio4 tetrahedra with AIPOa. However, the fact that B-o-cristobalite phase transformation is accompanied the chemically modified silica did not result in any with a volume contraction of &5 vol. when the significant change in the lattice parameters [4, 6, 7] had crystal changes from the cubic Fd3m to tetragonal gued against the stuffed B-cristobalite structure and P4,2,2 symmetry [1]. Recent studies supported by so the chemical stabilization. It suggested [7 that molecular simulations suggested [2] that the high-tem- any impurities would have affected the lattice parame ters determined experimentally. Contradictorily, it was ding author.Tel.:+886-7-525-4052;fax:+886-7-525- also reported [12] that the lattice constant of a-cristo balite decreased with heating the silicic-acid-derived E-mail address: hyl@mailnsysu. edu. tw(H.-Y. Lu) powders from 1080 to 1420C, in which the 101- -peak 0921-5093/02/s.see front matter c 2002 Elsevier Science B.V. All rights reserved PI:S0921-5093(01)01703-8
Materials Science and Engineering A328 (2002) 267–276 Stress-induced -cristobalite phase transformation in (Na2O+Al2O3)-codoped silica Chin-Hsiao Chao, Hong-Yang Lu * Institute of Materials Science and Engineering, National Sun Yat-Sen Uniersity, Kaohsiung 80424, Taiwan Received 9 April 2001; received in revised form 25 June 2001 Abstract Colloidal gel-derived silica (SiO2) powder codoped with (Na2O+Al2O3) was sintered at 1100 °C. The crystalline phase content and phase transformation of the sintered ceramics have been studied via X-ray diffractometry and scanning electron microscopy. The amount of -cristobalite retained metastably in the mixture to room temperature is found to depend on the level of additives. Samples codoped with Na2O and Al2O3, both of 6.30 mol%, were found to contain only -cristobalite in the crystalline mixture, and which is known as the chemically stabilized cristobalite (CSC). Multiple liquid phase separation is also observed in the codoped samples. The lattice spacing d101 of -cristobalite increases with the doping level while d111 of the -phase remains almost unchanged in all compositions studied. Surface grinding or pulverizing of the sintered samples into powder induces the -cristobalite phase transformation. The mechanism of the high temperature -cristobalite stabilization to room temperature associated with both the chemical and mechanical terms is discussed. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Sintering; Nucleation and growth; Phase transformation www.elsevier.com/locate/msea 1. Introduction Crystalline silica (SiO2) experiences a series of polymorphic phase transformation from cristobalite to tridymite, and to quartz upon cooling to room temperature in atmospheric pressure. Other crystalline phases of stoichovite and coesite also exist under high pressures. Both the higher temperature polymorphs of cristobalite and tridymite are known to exist metastably to room temperature in atmospheric pressure, although quartz is believed to be the thermodynamically most stable phase under such conditions. It remains uncertain if the stabilization of cristobalite and tridymite in natural minerals, which often incorporate impurity oxides, is imparted by cation substitutions for silicon. The -cristobalite phase transformation is accompanied with a volume contraction of 5 vol.% when the crystal changes from the cubic Fd3m to tetragonal P41212 symmetry [1]. Recent studies supported by molecular simulations suggested [2] that the high-temperature -cristobalite is a truly dynamic disordered phase [3]. Others [4,5], however, proposed models which suggested that the -structure is composed of low-symmetry domains. The stabilization of -cristobalite can occur [6,7] by: (a) altering chemical composition, and/or (b) imposing mechanical constraints. In the first case, it appears that acceptor impurities in solid solution with silica stabilize the -phase in a manner similar [6] to that of the fully stabilized cubic (c)-ZrO2 [8]. The former named accordingly the chemically stabilized cristobalite (CSC) has been synthesized [9–11] by incorporating ‘stuffing’ cations of Ca2+, Cu2+ and Sr2+ into the tectosilicate framework, or alternatively [11] by directly replacing the SiO4 tetrahedra with AlPO4. However, the fact that the chemically modified silica did not result in any significant change in the lattice parameters [4,6,7] had argued against the stuffed -cristobalite structure and so the chemical stabilization. It was suggested [7] that any impurities would have affected the lattice parameters determined experimentally. Contradictorily, it was also reported [12] that the lattice constant of -cristobalite decreased with heating the silicic-acid-derived powders from 1080 to 1420 °C, in which the 101-peak * Corresponding author. Tel.: +886-7-525-4052; fax: +886-7-525- 6030. E-mail address: hyl@mail.nsysu.edu.tw (H.-Y. Lu). 0921-5093/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 0 9 3 ( 0 1 ) 0 1 7 0 3 - 8
268 C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 shifted towards higher 20-angles. Second, the transfor- hot zone. The estimated cooling rate(of air-cooling) mation of B-cristobalite to a may be hindered mechani- was x 110C min- for pressureless-sintering in a cally by matrix constraint when the volume change conventional box furnace equipped with Sic heating accompanied with the phase transformation is sup- elements pressed. Indeed, tensile stresses generated across the Crystalline phases were analysed in the 20 range of interface of glass and B-cristobalite during cooling was 20-80 using X-ray diffractometry(XRD)(Siemens thought [13] to have triggered off the transformation of D5000, Karlsruhle, Germany) operating at 40 kV and the metastable B-cristobalite to o in crystallized silica. 30 mA using CuKo radiation with Ni filter; a scanning We have investigated the crystalline phases in the speed of 0.6 20 min was adopted. A scanning speed intered samples prepared from colloidal gel-derived of the 20-angle at 0.005- and a time constant of 10 SiO2 powders co-doped with Na2O and Al2O3. The s were always ensured for step-scanning. Reflections sintered samples re-ground to fine powders are found to between the 20 angles of 20.5-23.5 for a-and B-cristo- have induced the phase transformation of metastable balite and those of 49-51 for a-quartz were used for B-cristobalite to a-phase. Besides the chemical stabiliza- both the crystalline phase identification and quantita- tion of B-cristobalite, the likelihood of particle size tive analysis. The relative content of a-to B-cristobalite effects on the B-0-cristobalite transformation is also in sintered discs was determined by measuring the peak discussed areas of (lll and(101)x using a calibrated curve established by a modified external standard method [4] 2. Experimental procedures Surface morphology and polished sections of the sintered samples were observed using a scanning elec Chemically stabilized B-cristobalite was synthesized tron microscope (SEM) (JEM6400, JEOL, Tokyo, in the Na2O-AlO3-SiO2 system adopting the powder Japan) operating at 20 kV. Sintered samples were compositions reported by Perrotta et al. [9]. The solute ground and polished mechanically with Sic papers cations of Al+ and Na+ were added in the equal successively before diamond lapping to I um surface concentration I molar ratio. The Na2O-doped roughness. Polished sections were then chemicall colloidal silica sols of Ludox HS-30 containing 2 30 etched using HF-HNO3 solution to delineate grain wt% of silica particles were supplied by E I du Pont de boundaries and other microstructural features Nemours(Wellington, DE). Appropriate quantities of Al, (SO4)3.13H,o and Na, CO,(reagent grade, Katayama, Osaka, Japan) dissolved in de-ionized water 3. Results was blended in the colloidal silica sols according to the oppositions of (Na,0+Al,O3)-ySiO2 given in 3. 1. Identification of crystalline phases Table 1. the mixed solution contained in a beaker was dried in an oven at 120 oC for 12 h to a cake of The(Na2O+Al,O3)-ySiO2 powders of four compo- amorphous powder. It was then deagglomerated using sitions with y= 13.9, 19.5, 24.6 and 32. 4 were sintered an agate mortar and pestle before passing through a and determined for their crystalline phases. The 177 um sieve. The ground and sieved powder was Na,O+Al2O3 contents of the initial compositions are lry-pressed at x 50 MPa in a wC-inserted steel die listed in Table I giving the dopant concentrations in assembly to discs of 10 mm()x 1.5 mm. The green both mol% and wt % compacts accommodated in an alumina boat were pres sureless-sintered at 1100C before cooling to room 3. 1. Composition with y= 19.5 temperature by rapid withdrawal of the boat from the As-sintered sample surface and sintered samples re- round to powder have both been analyzed. Fig. 1(a) Table I presents the representative XRD trace of the as-sintered Na,O-AL,O, contents of the initial colloidal silica compositions surface of the y= 19.5 samples, which have been fired at 1100C for 48 h. The reflection peaks in the le range AL,O -,, of 20= 17-37 are designated to C-l at 20= 21.8, C-2 13.9 at27.3°,C-3at31.2°andC-4at35.8°. The lattice spacings calculated from C-2 at 20=27.3 and C-3 at 3.76 4.65 6.29 31.20 agree very well with those of (lll)a and (102)B 4.65 6.29 respectively, as compared with those given by JCPDS 2 mol% 87.42 Al-O, wt% 39-1425 for a-cristobalite. These two reflections repre- 64 sent characteristically that the sintered mixture consist SiO, wt 87.73 83.59 predominantly of a-cristobalite. They may also contain other silica polymorphs but only of negligible quantities
268 C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 shifted towards higher 2-angles. Second, the transformation of -cristobalite to may be hindered mechanically by matrix constraint when the volume change accompanied with the phase transformation is suppressed. Indeed, tensile stresses generated across the interface of glass and -cristobalite during cooling was thought [13] to have triggered off the transformation of the metastable -cristobalite to in crystallized silica. We have investigated the crystalline phases in the sintered samples prepared from colloidal gel-derived SiO2 powders co-doped with Na2O and Al2O3. The sintered samples re-ground to fine powders are found to have induced the phase transformation of metastable -cristobalite to -phase. Besides the chemical stabilization of -cristobalite, the likelihood of particle size effects on the -cristobalite transformation is also discussed. 2. Experimental procedures Chemically stabilized -cristobalite was synthesized in the Na2O–Al2O3 –SiO2 system adopting the powder compositions reported by Perrotta et al. [9]. The solute cations of Al3+ and Na+ were added in the equal concentration of 1:1 molar ratio. The Na2O-doped colloidal silica sols of Ludox® HS-30 containing 30 wt.% of silica particles were supplied by E.I. du Pont de Nemours (Wellington, DE). Appropriate quantities of Al2(SO4)3 · 13H2O and Na2CO3 (reagent grade, Katayama, Osaka, Japan) dissolved in de-ionized water was blended in the colloidal silica sols according to the compositions of (Na2O+Al2O3)−ySiO2 given in Table 1. The mixed solution contained in a beaker was dried in an oven at 120 °C for 12 h to a cake of amorphous powder. It was then deagglomerated using an agate mortar and pestle before passing through a 177 m sieve. The ground and sieved powder was dry-pressed at 50 MPa in a WC-inserted steel die assembly to discs of 10 mm()×1.5 mm. The green compacts accommodated in an alumina boat were pressureless-sintered at 1100 °C before cooling to room temperature by rapid withdrawal of the boat from the hot zone. The estimated cooling rate (of air-cooling) was 110 °C min−1 for pressureless-sintering in a conventional box furnace equipped with SiC heating elements. Crystalline phases were analysed in the 2 range of 20–80° using X-ray diffractometry (XRD) (Siemens D5000, Karlsruhle, Germany) operating at 40 kV and 30 mA using CuK radiation with Ni filter; a scanning speed of 0.6° 2 min−1 was adopted. A scanning speed of the 2-angle at 0.005° s−1 and a time constant of 10 s were always ensured for step-scanning. Reflections between the 2 angles of 20.5–23.5° for - and -cristobalite and those of 49–51° for -quartz were used for both the crystalline phase identification and quantitative analysis. The relative content of - to -cristobalite in sintered discs was determined by measuring the peak areas of (111) and (101) using a calibrated curve established by a modified external standard method [14]. Surface morphology and polished sections of the sintered samples were observed using a scanning electron microscope (SEM) (JEM6400, JEOL, Tokyo, Japan) operating at 20 kV. Sintered samples were ground and polished mechanically with SiC papers successively before diamond lapping to 1 m surface roughness. Polished sections were then chemically etched using HF-HNO3 solution to delineate grainboundaries and other microstructural features. 3. Results 3.1. Identification of crystalline phases The (Na2O+Al2O3)–ySiO2 powders of four compositions with y=13.9, 19.5, 24.6 and 32.4 were sintered and determined for their crystalline phases. The Na2O+Al2O3 contents of the initial compositions are listed in Table 1 giving the dopant concentrations in both mol% and wt.%. 3.1.1. Composition with y=19.5 As-sintered sample surface and sintered samples reground to powder have both been analyzed. Fig. 1(a) presents the representative XRD trace of the as-sintered surface of the y=19.5 samples, which have been fired at 1100 °C for 48 h. The reflection peaks in the range of 2=17–37° are designated to C-1 at 2=21.8°, C-2 at 27.3°, C-3 at 31.2° and C-4 at 35.8°. The lattice spacings calculated from C-2 at 2=27.3° and C-3 at 31.2° agree very well with those of (111) and (102) respectively, as compared with those given by JCPDS 39-1425 for -cristobalite. These two reflections represent characteristically that the sintered mixture consist predominantly of -cristobalite. They may also contain other silica polymorphs but only of negligible quantities Table 1 Na2O–Al2O3 contents of the initial colloidal silica compositions Al2O3–Na2O–ySiO2 y 32.4 24.6 19.5 13.9 Al2O3 mol% 3.76 6.29 2.91 4.65 Na2O mol% 6.29 2.91 3.76 4.65 SiO2 mol% 94.19 92.48 90.70 87.42 Al2O3 wt.% 4.83 6.21 7.63 10.21 Na2O wt.% 6.20 2.94 4.64 3.77 SiO2 wt.% 92.23 90.02 83.59 87.73
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 temperature polymorph of crystalline silica, B-cristo 19.5 balite, has apparently been stabilized and retained 1100℃/48h metastably to room temperature in the y= 19.5 samples with 4.65 mol% of both Na,O and Al,O3(Table 1) The lattice spacing of d=0.4069 nm(C-1)does not match exactly with that of either the a- or B-phase given by the JCPDS files, but falls between dIor C-2C- 0.4039 nm and d11B=0.4110 nm. However, the C 102aC4 reflection in Fig. l(a)can be deconvoluted to two peaks corresponding to the lattice spacings of 0.4084 and 0.4058 nm and representing(101)a and(lID),respec- 20(deg) tively. This is shown in Fig. 1(b). The co-existence of a-cristobalite and the untransformed B-cristobalite can be discerned. It is also noted that no peak splitting(of the C-l reflection in Fig. I(c) was detected before the 20 1100℃/48h sintered samples had been pulverized again to powders a-cristobalite (C-l(a) in Fig. I(c). The C-4 peak(Fig. I(a))at 26 35.80 corresponding to a lattice spacing of d=0. 2506 nm again lies amongst d12a =0.2467 nm, d200 =0.2487 nm and d220B=0.2530 nm(JCPDS 27-605 for B-cristo balite). Similarly, it implies the co-existence of a-and B-cristobalite in the sintered mixture. The later is appar 20.521.021.5 022523.023 ently the untransformed high temperature B-phase 20(dee) which has been retained metastably to room 35 temperature. Sintered samples re-ground to powder were passed 1100℃/48h through a 45 um sieve prior to XRD analysis. The characteristic feature of the re-ground samples is the peak splitting of the initially Gaussian-type reflections of C-1(also given in Fig. I(c))and C-4. The C-l peak 20=21.8%(of Fig. 1(a)) has splitted of dup=0.4086 nm at 20=21.75 and d1ola=0.4058 nm at 20=21.90. They are again not completely in 20.521.021.522.022.52 3.524.0 accordance with the d-spacings(of d1B=0.4110 nm 20(deg) and dIola=0.4039 nm)given by the JCPDS files. The 111)B reflection shown in Fig. 1(c)for both as-sintered ng. I. XRD traces of (a)as-sintered surface of the y= 19.5 sample and re-ground samples remains almost unchanged in C for 24 h,(b) deconvoluded peak, and (c)re- position. However, the (101), reflection shifting to- higher 20-angles) from that (of <5 wt %)and beyond the detection by XRD, if of the as-sintered sample surface can be easily discerned they exist at all in the mixture from Fig. I(c) The reflections of C-l at 20=21.8 and C4 at 35.8 3. 1.2. Composition with y=24.6 may include both a- and B-cristobalite since the respec For powder compacts containing a smaller amount tive 20-angles are very close to each other(as indicated of Al,Oa and Na,o(e.g. y=24.6 of 3. 76 mol% as given in Fig. I(a). Investigating the peak areas of C-I and in Table 1), sintering at 1100C for 24 h results in C-4 in Fig. I(a)reveals that the sample may also mixture of a-and B-cristobalite. The XRd trace resem contain B-cristobalite. The relative intensities of C-l to bles that of the y=19.5 samples (of Fig. I()). The C-4 by integrating the peak areas are 1045: 24: 41: 236 20=21.8 reflection approximated to a Gaussian-type Normalizing them on the basis of the C-2 peak (i.e. peak locating between(101)a and(Ill)e is again ob (lID)a) intensity following the JCPDS file 39-1425, the tained. and which is also designated to C-1 in Fig. 2 mixture containing only a-cristobalite would give a Similar peak splitting and shift are observed from the peak ratio of 300: 24: 27: 51. The discrepancy of excess y=246 samples re-ground to 45-38 Hm, shown by intensities by A=745 for the C-l peak and 4=185 for C-1(a)in Fig. 2. When the same sample was pulverized the C-4 peak indicates the co-existence of B-cristobalite further down to particle size of 38 um, not only that with the a-phase in the crystalline mixture. The high the peak splitting had become more distinctive, the
C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 269 Fig. 1. XRD traces of (a) as-sintered surface of the y=19.5 sample sintered at 1100 °C for 24 h, (b) deconvoluded peak, and (c) reground powder showing -cristobalite. temperature polymorph of crystalline silica, -cristobalite, has apparently been stabilized and retained metastably to room temperature in the y=19.5 samples with 4.65 mol% of both Na2O and Al2O3 (Table 1). The lattice spacing of d=0.4069 nm (C-1) does not match exactly with that of either the - or -phase given by the JCPDS files, but falls between d101= 0.4039 nm and d111=0.4110 nm. However, the C-1 reflection in Fig. 1(a) can be deconvoluted to two peaks corresponding to the lattice spacings of 0.4084 and 0.4058 nm and representing (101) and (111), respectively. This is shown in Fig. 1(b). The co-existence of -cristobalite and the ‘untransformed’ -cristobalite can be discerned. It is also noted that no peak splitting (of the C-1 reflection in Fig. 1(c)) was detected before the sintered samples had been pulverized again to powders (C-1(a) in Fig. 1(c)). The C-4 peak (Fig. 1(a)) at 2= 35.8° corresponding to a lattice spacing of d=0.2506 nm again lies amongst d112=0.2467 nm, d200=0.2487 nm and d220=0.2530 nm (JCPDS 27-605 for -cristobalite). Similarly, it implies the co-existence of - and -cristobalite in the sintered mixture. The later is apparently the untransformed high temperature -phase which has been retained metastably to room temperature. Sintered samples re-ground to powder were passed through a 45 m sieve prior to XRD analysis. The characteristic feature of the re-ground samples is the peak splitting of the initially Gaussian-type reflections of C-1 (also given in Fig. 1(c)) and C-4. The C-1 peak at 2=21.8° (of Fig. 1(a)) has splitted into two peaks of d111=0.4086 nm at 2=21.75° and d101=0.4058 nm at 2=21.90°. They are again not completely in accordance with the d-spacings (of d111=0.4110 nm and d101=0.4039 nm) given by the JCPDS files. The (111) reflection shown in Fig. 1(c) for both as-sintered and re-ground samples remains almost unchanged in position. However, the (101) reflection shifting towards smaller d-spacings (higher 2-angles) from that of the as-sintered sample surface can be easily discerned from Fig. 1(c). 3.1.2. Composition with y=24.6 For powder compacts containing a smaller amount of Al2O3 and Na2O (e.g. y=24.6 of 3.76 mol% as given in Table 1), sintering at 1100 °C for 24 h results in a mixture of - and -cristobalite. The XRD trace resembles that of the y=19.5 samples (of Fig. 1(a)). The 2=21.8° reflection approximated to a Gaussian-type peak locating between (101) and (111) is again obtained, and which is also designated to C-1 in Fig. 2. Similar peak splitting and shift are observed from the y=24.6 samples re-ground to 45–38 m, shown by C-1(a) in Fig. 2. When the same sample was pulverized further down to particle size of 38 m, not only that the peak splitting had become more distinctive, the (of 5 wt.%) and beyond the detection by XRD, if they exist at all in the mixture. The reflections of C-1 at 2=21.8° and C-4 at 35.8° may include both - and -cristobalite since the respective 2-angles are very close to each other (as indicated in Fig. 1(a)). Investigating the peak areas of C-1 and C-4 in Fig. 1(a) reveals that the sample may also contain -cristobalite. The relative intensities of C-1 to C-4 by integrating the peak areas are 1045:24:41:236. Normalizing them on the basis of the C-2 peak (i.e. (111)) intensity following the JCPDS file 39-1425, the mixture containing only -cristobalite would give a peak ratio of 300:24:27:51. The discrepancy of excess intensities by =745 for the C-1 peak and =185 for the C-4 peak indicates the co-existence of -cristobalite with the -phase in the crystalline mixture. The high
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 (101) reflection at 20=21.95 designated to C-1(b) also appeared to have higher intensity than(IlDB- The suggestion is that re-grinding the sintered samples to powder and to finer particles favours the formation of C-1(b)(<38um) a-cristobalite. The B-+al-phase transformation has ap parently been encouraged in the mixture by the reduc C1a)(4538um) tion of particle sizes from pulverizing or simply the relief of the matrix constraints. It is clear that the(101) eflection has shifted towards smaller d-spacing as pow 21,021.5 2.523.023.5 der particle size is further reduced. The temperature of the cubic-tetragonal phase transformation dep upon the particle size is not uncommon in aces of surface(C-1), re-ground to The retention of high-temperature polymorphs has powders of different sizes(C-1(a)and C-l(b)of y=24. 6 sintered at been reported for various ferroelastic and ferroelectric 00°cfor24h. ceramics of, e.g. BaTiO3, [15, 16] ZrO2[17] and indeed SO2[6,15,18] 1100℃/48h 3.1.3. Composition with y= 13.9 For samples with higher Al2O3- and NayO-contents (e.g. y= 13.9 of 6.29 mol% as given in Table 1), sinter ing at 1100C for 48 h has produced predominantly B-cristobalite even after samples are re-ground to fine powders of <45-38 um. It is shown clearly in Fig 3 where the characteristic a-cristobalite reflections of 18212427303336394245 (1l1) and (102), are absent from the XRD trace Judging from the peak intensity, the air-cooled sample Fig. 3 ith y-3. 9 sintered at l100" C for 48 h showing The indication is that crystallization of B-cristobalite from the sintered gel-derived powder compact is still continuing upon cooling(from the sintering tempera- 1100℃/4 ture of 1100C) to room temperature. The high-tem (20=22.00 perature B-cristobalite in the sintered mixture has been retained metastably to room temperature by the higher doping level of 6.29 mol% The XRD trace for the y= 13.9 sample pulverized to 45-38 um powder did not show any peak splitting at 20=21 all(as evidenced form Fig 4(a)). In fact, it is the fully stabilized sample in which the B-a-cristobalite phase transformation has been prevented all together 21,0 23.0 Sintering at 1100C for 48 h, samples with y=19.5, 20(deg) 24.6 and 32.4 always yielded the crystalline mixture of (a+ B)-cristobalite. The XRD traces in the proximity of Cristobalite (d 20=21.75 for the three compositions and for y= 13.9 are given in Fig. 4(a). Taking the peak height to represent the amount of crystalline phases in the mix ture, it appears that the quantity of B-cristobalite(of 0.406 20=21.75) remains almost unaltered when that of a-cristobalite has increased very appreciably e.g. by N1.6 times from samples containing Al,O3 and Na2O y=324 of 4.65 mol%(=19.5)to 2.91 mol%(=32. 4). It is illustrated in Fig. 4(b) that the lattice spacing of B- cristobalite(duB) als increasing Al,O3- and Na2O-content. The data point for the y= 32. 4 sample is missing from Fig. 4(b) since Fig4.(a) Peak shift registered by(111)B and (101)a, and (b) change its B-cristobalite content is negligible from the XRD of the lattice spacings of dola and dinp in samples of four dopant levels sintered at 1100 C for 48 h trace(as indicated in Fig. 4(a)). On the other hand
270 C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 Fig. 2. XRD traces of the as-sintered surface (C-1), re-ground to powders of different sizes (C-1(a) and C-1(b)) of y=24.6 sintered at 1100 °C for 24 h. (101) reflection at 2=21.95° designated to C-1(b) also appeared to have higher intensity than (111). The suggestion is that re-grinding the sintered samples to powder and to finer particles favours the formation of -cristobalite. The -phase transformation has apparently been encouraged in the mixture by the reduction of particle sizes from pulverizing or simply the relief of the matrix constraints. It is clear that the (101) reflection has shifted towards smaller d-spacing as powder particle size is further reduced. The temperature of the cubictetragonal phase transformation depending upon the particle size is not uncommon in ceramics. The retention of high-temperature polymorphs has been reported for various ferroelastic and ferroelectric ceramics of, e.g. BaTiO3, [15,16] ZrO2 [17] and indeed SiO2 [6,15,18]. 3.1.3. Composition with y=13.9 For samples with higher Al2O3- and Na2O-contents (e.g. y=13.9 of 6.29 mol% as given in Table 1), sintering at 1100 °C for 48 h has produced predominantly -cristobalite even after samples are re-ground to fine powders of 45–38 m. It is shown clearly in Fig. 3 where the characteristic -cristobalite reflections of (111) and (102) are absent from the XRD trace. Judging from the peak intensity, the air-cooled sample contains more -cristobalite than the water-quenched. The indication is that crystallization of -cristobalite from the sintered gel-derived powder compact is still continuing upon cooling (from the sintering temperature of 1100 °C) to room temperature. The high-temperature -cristobalite in the sintered mixture has been retained metastably to room temperature by the higher doping level of 6.29 mol%. The XRD trace for the y=13.9 sample pulverized to 45–38 m powder did not show any peak splitting at all (as evidenced form Fig. 4(a)). In fact, it is the fully stabilized sample in which the -cristobalite phase transformation has been prevented all together. Sintering at 1100 °C for 48 h, samples with y=19.5, 24.6 and 32.4 always yielded the crystalline mixture of (+)-cristobalite. The XRD traces in the proximity of 2=21.75° for the three compositions and for y=13.9 are given in Fig. 4(a). Taking the peak height to represent the amount of crystalline phases in the mixture, it appears that the quantity of -cristobalite (of 2=21.75°) remains almost unaltered when that of -cristobalite has increased very appreciably e.g. by 1.6 times from samples containing Al2O3 and Na2O of 4.65 mol% (y=19.5) to 2.91 mol% (y=32.4). It is illustrated in Fig. 4(b) that the lattice spacing of - cristobalite (d111) also remains unchanged with the increasing Al2O3- and Na2O-content. The data point for the y=32.4 sample is missing from Fig. 4(b) since its -cristobalite content is negligible from the XRD trace (as indicated in Fig. 4(a)). On the other hand, Fig. 3. Samples with y=13.9 sintered at 1100 °C for 48 h showing predominantly -cristobalite by water-quenching and air-cooling. Fig. 4. (a) Peak shift registered by (111) and (101), and (b) change of the lattice spacings of d101 and d111 in samples of four dopant levels sintered at 1100 °C for 48 h.
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 diola of a-cristobalite increases monotonically with presented in Fig. 5(b)(for feature-l in Fig 5(a). Some higher dopant levels. of these particles have coalesced to form doublets and The peak shift of (101)a from 20=21.9(of y= 19.5) triplets. Second, extensive microcracks, feature-2, can to 20= 220(of y= 32. 4)for the three samples is also not be differentiated for its trans- or inter-granular an indication of different solid solubilities (of both nature with certainty, since the matrix has been disinte- AP+ and Na*), as shown in Fig. 4(a). The lattice grated after grinding and polishing and grains are no spacings exhibit an increasing trend from the lower longer discernible from Fig. 5(b). Third, the residual (Al,O3+ Na2O)-doping levels of y= 32. 4 to y= 19.5 of glassy phase contains two distinctive types of crystalline higher doping levels. It suggests that the higher con- particles, dendritic and spherical in sha The SeM tents of Al,O3 and Na,O in the initial silica sols pro- backscattered electron image(BED) provides a distin duce sintered samples containing more a-cristobalite guishable contrast. It represents the nucleation at th with the larger dyol. Nevertheless, dug of B-cristobalite early stage as shown by higher magnification in F remains unchanged at 20=21.75 of d11B=0.4086 nm. 5(c) for feature-3. The indication is that two types of It is demonstrated unambiguously that the y= 13.9 nucleation have taken place from the melts. Some of samples with 6.29 mol%(Al,O3+Na2O)-doping con- the spherical particles have also formed multiplets as tains solely of B-cristobalite(Fig 4(a)in the crystalline indicated in Fig. 5(c). Here, the atomic contrast(bright mixture. It appears that silica samples grown in stilbite versus grey) revealed by SEM-BEI suggests that these NaCa2AlSSi13O36)[6] could have dissolved Na,O and two types of crystals may have different chemical Al,, of sufficient quantities to render the stabilization compositions ofβ- cristobalite The cracked matrix contains both a- and B-cristo- The results suggest that the content of Al2O3 and balite(Fig. 5(a) and(b)as evidenced from XRD(for Na,o determine the crystalline phases of the sintered y=19.5 shown in Fig. 4(a)where the (101) and (IID colloidal gel-derived powder. Higher additive content peaks are almost equal in height). It is of course not appears to favour the formation of the high tempera- known from SEM which are the B-cristobalite crystals ture B-cristobalite. The B-cristobalite in samples with Nevertheless, microstructure analysis using TEM [19] y=13.9 stabilized and retained metastably to room has identified B-phase in the sintered(Al,O3+Na2O)- temperature has also sustained the phase transforma- codoped samples, and it will appear in another tion induced by re-grinding of the sintered samples to manuscript. Along the glass-matrix interface(indicated powder. With higher(Al,O3+ Na2O)-doping levels, e. g. by arrow in Fig. 5(a)and shown also by higher magnifi y=13.9, not only the transformation to a-tridymite is cation in Fig. 5(d)), it can be seen that cracking has not completely suppressed, but also the B-cristobalite con- extended into the residual glass. The indication is that tent is increased in the phase assemblage. This ha the cracks in Fig. 5(b) are likely to have been generated become more conspicuous as the(Al,O3+ Na2O)-con- by grinding(or indeed the applied stress) because of the tent increases from y= 24.6 (i.e. 3.76 mol%)to y= 13.9 volume contraction accompanying with the induced (i.e. 6.29 mol%)and also for longer sintering hours(of B-a-cristobalite phase transformation. The character- 48 h as in Fig 4(a)) istic feature of the a-cristobalite twin lamellae [20] better ned [19 by TEM has been smeared out 3. 2. Microstructure after grinding and polishing after grinding. Consequently, unlike those in Fig 6(b), they can no longer be identified by SEM. The Phase transformation has been induced on sintered spherical particles, however, remain intact (and still sample surfaces by grinding and polishing when prepar- attached to the cracked matrix) when allowing cracks ing the polished sections for microstructure analysis. to deflect into the matrix containing predominantly the The characteristic microstructures of samples contain- transformed a-phase(Fig. 5(b). It suggests that me- ing(a) both(a+ B)-cristobalite and(b) only B-cristo- chanical grinding induces the B-+a-cristobalite phase balite in the crystalline mixture, as evidenced by XRD transformation, but some of the B-cristobalite in the ( Fig. 4(a)), are described in the following form of spherical particles, successfully resisted the transformation, have retained its high temperature cu 3.2.1. y=19.5 samples containing (+B)-cristobalite bic symmetry. These B-cristobalite particles of 300-500 a typical microstructure of samples with y=19, e unyielding to mechanical grinding would have been stabilized by the doping solutes and become'non containing Na,O and Al,O3, each of 4.65 mol%, and transformable at room temperature adopting the term sintered at 1100C for 48 h, is shown by SEM-SEl in used for the Y2O3-doped Zro2[8] Fig. 5(a) where three specific features(labelled as 1, 2 Dendritic and cellular growth(Fig. 6(a)) are both and 3)are discernible at low magnifications. First of all observed in one of the residual pores from the sintered spherical particles of 300-500 nm are found to disperse and polished sample. No induced cracks such as those biquitously in the porous and cracked matrix, occurred in Fig. 5(b)can immediately be discerned
C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 271 d101 of -cristobalite increases monotonically with higher dopant levels. The peak shift of (101) from 2=21.9° (of y=19.5) to 2=22.0° (of y=32.4) for the three samples is also an indication of different solid solubilities (of both Al3+ and Na+), as shown in Fig. 4(a). The lattice spacings exhibit an increasing trend from the lower (Al2O3+Na2O)-doping levels of y=32.4 to y=19.5 of higher doping levels. It suggests that the higher contents of Al2O3 and Na2O in the initial silica sols produce sintered samples containing more -cristobalite with the larger d101. Nevertheless, d111 of -cristobalite remains unchanged at 2=21.75° of d111=0.4086 nm. It is demonstrated unambiguously that the y=13.9 samples with 6.29 mol% (Al2O3+Na2O)-doping contains solely of -cristobalite (Fig. 4(a)) in the crystalline mixture. It appears that silica samples grown in stilbite (NaCa2Al5Si13O36) [6] could have dissolved Na2O and Al2O3 of sufficient quantities to render the stabilization of -cristobalite. The results suggest that the content of Al2O3 and Na2O determine the crystalline phases of the sintered colloidal gel-derived powder. Higher additive content appears to favour the formation of the high temperature -cristobalite. The -cristobalite in samples with y=13.9 stabilized and retained metastably to room temperature has also sustained the phase transformation induced by re-grinding of the sintered samples to powder. With higher (Al2O3+Na2O)-doping levels, e.g. y=13.9, not only the transformation to -tridymite is completely suppressed, but also the -cristobalite content is increased in the phase assemblage. This has become more conspicuous as the (Al2O3+Na2O)-content increases from y=24.6 (i.e. 3.76 mol%) to y=13.9 (i.e. 6.29 mol%) and also for longer sintering hours (of 48 h as in Fig. 4(a)). 3.2. Microstructure after grinding and polishing Phase transformation has been induced on sintered sample surfaces by grinding and polishing when preparing the polished sections for microstructure analysis. The characteristic microstructures of samples containing (a) both (+)-cristobalite and (b) only -cristobalite in the crystalline mixture, as evidenced by XRD (Fig. 4(a)), are described in the following: 3.2.1. y=19.5 samples containing (+)-cristobalite mixture A typical microstructure of samples with y=19.5 containing Na2O and Al2O3, each of 4.65 mol%, and sintered at 1100 °C for 48 h, is shown by SEM-SEI in Fig. 5(a) where three specific features (labelled as 1, 2 and 3) are discernible at low magnifications. First of all, spherical particles of 300–500 nm are found to disperse ubiquitously in the porous and cracked matrix, as presented in Fig. 5(b) (for feature-1 in Fig. 5(a)). Some of these particles have coalesced to form doublets and triplets. Second, extensive microcracks, feature-2, can not be differentiated for its trans- or inter-granular nature with certainty, since the matrix has been disintegrated after grinding and polishing and grains are no longer discernible from Fig. 5(b). Third, the residual glassy phase contains two distinctive types of crystalline particles, dendritic and spherical in shape. The SEM backscattered electron image (BEI) provides a distinguishable contrast. It represents the nucleation at the early stage as shown by higher magnification in Fig. 5(c) for feature-3. The indication is that two types of nucleation have taken place from the melts. Some of the spherical particles have also formed multiplets as indicated in Fig. 5(c). Here, the atomic contrast (bright versus grey) revealed by SEM-BEI suggests that these two types of crystals may have different chemical compositions. The cracked matrix contains both - and -cristobalite (Fig. 5(a) and (b)) as evidenced from XRD (for y=19.5 shown in Fig. 4(a) where the (101) and (111) peaks are almost equal in height). It is of course not known from SEM which are the -cristobalite crystals. Nevertheless, microstructure analysis using TEM [19] has identified -phase in the sintered (Al2O3+Na2O)- codoped samples, and it will appear in another manuscript. Along the glass-matrix interface (indicated by arrow in Fig. 5(a) and shown also by higher magnifi- cation in Fig. 5(d)), it can be seen that cracking has not extended into the residual glass. The indication is that the cracks in Fig. 5(b) are likely to have been generated by grinding (or indeed the applied stress) because of the volume contraction accompanying with the induced -cristobalite phase transformation. The characteristic feature of the -cristobalite twin lamellae [20] better discerned [19] by TEM has been smeared out after sample grinding. Consequently, unlike those in Fig. 6(b), they can no longer be identified by SEM. The spherical particles, however, remain intact (and still attached to the cracked matrix) when allowing cracks to deflect into the matrix containing predominantly the transformed -phase (Fig. 5(b)). It suggests that mechanical grinding induces the -cristobalite phase transformation, but some of the -cristobalite in the form of spherical particles, successfully resisted the transformation, have retained its high temperature cubic symmetry. These -cristobalite particles of 300–500 nm unyielding to mechanical grinding would have been fully-stabilized by the doping solutes and become ‘nontransformable’ at room temperature adopting the term used for the Y2O3-doped ZrO2 [8]. Dendritic and cellular growth (Fig. 6(a)) are both observed in one of the residual pores from the sintered and polished sample. No induced cracks such as those occurred in Fig. 5(b) can immediately be discerned.
C -H. Chao, H.-Y. Lu/Materials Science and Engineering 4328 (2002)267-276 dendritc partial 000038o a :500nm b m (d) cracked matrix(SEM-SED) (c)early stage of nucleation and (d) interface between glass and the cracked matrix(SEM-BElh ticles dispersed in Fig. 5. General structure of the y= 19.5 sample sintered at 1100C for 48 h,(a) three general features, (b) spherical par Glassy phases amongst the dendritic arms etched away formed to a-cristobalite(Fig. 6(b)), although they are hen preparing for the polished sections have become a found only occasionally. They may have escaped from recess, which makes chemical analysis for their content the grinding stress when located within a residual pore by energy-dispersive X-ray spectroscopy(EDS)imprac- during preparing for the polished sections. a-Grains tical. Although crystallized features are similar to those larger than a l um can clearly be identified from the commonly found in metals, and in the silicate system of characteristic twinning. These grains confined Li,O-SiO2 or BaO-Sio2 [21], preferred crystallo- glassy matrix must have exceeded the critical size for graphic orientations of (101)a reported [13] for devit- phase transformation to a-cristobalite. rified glass surface was not seen in the secondary dendrites. Fig. 6(a) reveals the primary and secondary 3. 2. 2. y= 13.9 samples containing only B-cristobalite dendritic arms where their crystallization by constitu SEM microstructures of the samples sintered at tional supercooling is indicative by the gaps(shown by 1100C for 48 h, containing Na,O and Al, O3, each of arrow) existed between them when viewing along the 6.29 mol%(i. e y= 13.9), are presented in Fig. 7(a)-(c) longitudinal direction. The gap would have contained XRD results(Figs. 3 and 4(a)) have shown that these glass of higher Na2O-content and (so)etching rate than samples contain only B-cristobalite. a-Cristobalite char- the dendrites. A crystallization mechanism involving acterized by lamellar twins [19, 20] is not detected in the expulsion of Nat-ion has indeed been proposed [22] these samples with higher doping levels. The sintered for Na,O-containing borosilicate glass microstructure is represented by five distinctive features Not all ground and polished sample surfaces appear as indicated in Fig. 7(a):(1) residual pores located in to contain the spherical particles. Some of the crystals the glassy matrix beco ome more noticeable; (2)spherical hibiting the cellular growth morphology have trans- particles resemble one of the nucleation types from Fig
272 C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 Fig. 5. General microstructure of the y=19.5 sample sintered at 1100 °C for 48 h, (a) three general features, (b) spherical particles dispersed in cracked matrix (SEM-SEI), (c) early stage of nucleation and (d) interface between glass and the cracked matrix (SEM-BEI). Glassy phases amongst the dendritic arms etched away when preparing for the polished sections have become a recess, which makes chemical analysis for their content by energy-dispersive X-ray spectroscopy (EDS) impractical. Although crystallized features are similar to those commonly found in metals, and in the silicate system of Li2O–SiO2 or BaO–SiO2 [21], preferred crystallographic orientations of (101) reported [13] for devitrified glass surface was not seen in the secondary dendrites. Fig. 6(a) reveals the primary and secondary dendritic arms where their crystallization by constitutional supercooling is indicative by the gaps (shown by arrow) existed between them when viewing along the longitudinal direction. The gap would have contained glass of higher Na2O-content and (so) etching rate than the dendrites. A crystallization mechanism involving the expulsion of Na+-ion has indeed been proposed [22] for Na2O-containing borosilicate glass. Not all ground and polished sample surfaces appear to contain the spherical particles. Some of the crystals exhibiting the cellular growth morphology have transformed to -cristobalite (Fig. 6(b)), although they are found only occasionally. They may have escaped from the grinding stress when located within a residual pore during preparing for the polished sections. -Grains larger than 1 m can clearly be identified from the characteristic twinning. These grains confined in a glassy matrix must have exceeded the critical size for phase transformation to -cristobalite. 3.2.2. y=13.9 samples containing only -cristobalite SEM microstructures of the samples sintered at 1100 °C for 48 h, containing Na2O and Al2O3, each of 6.29 mol% (i.e. y=13.9), are presented in Fig. 7(a)–(c). XRD results (Figs. 3 and 4(a)) have shown that these samples contain only -cristobalite. -Cristobalite characterized by lamellar twins [19,20] is not detected in these samples with higher doping levels. The sintered microstructure is represented by five distinctive features as indicated in Fig. 7(a): (1) residual pores located in the glassy matrix become more noticeable; (2) spherical particles resemble one of the nucleation types from Fig.
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 (b):3)dendrites have grown to impinge one an- other where the primary and secondary dendrite arms have no apparent crystallographic relationships under SEM(Fig. 7(b);(4)cellular growth produces grains of submicron size, which appear to have two distinc tive morphologies of dendrites and cellular grains (Fig. 7(c); and (5)the liquid immiscibility featuring multiple phase separation (as indicated in Fig. 7(b) and (c)) is seen extensively over the polished and etched sample surface the proximity of the B-cristobalite crystals which ex- hibit both the dendritic and cellular morphology indi- cated in Fig. 7(c). It is the continuous type [23] of μm iquid phase separation, where one (i.e. the droplet (a) phase indicated by feature (5)in Fig. 7(c)) of the two liquid phases is separated from the other by a contin- uous glass matrix. The observation is very significant since the extensive microcracking in the y= 19.5 sam- dendrite (b) ecelldiangro 10μm (a) μm FI ve distinctive features of the y= 13.9 sample,(b) th, (c)multiple liquid-phase separation in the vicinity of dene nd cellular growth(SEM-SED) um ples of Fig. 5(b)is not detected from samples with (b) y=13.9 of higher doping level after they have both been subjected to similar grinding and polishing. This Fig.6.(a)Dendritic and cellular growth and(b) characteristic Implies that B-a-cristobalite phase transformation vinning feature of the a-crisotablite within residual pores (SEM has not occurred in the y= 13.9 samples even after surface grinding and polishing
C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 273 5(b); (3) dendrites have grown to impinge one another where the primary and secondary dendrite arms have no apparent crystallographic relationships under SEM (Fig. 7(b)); (4) cellular growth produces grains of submicron size, which appear to have two distinctive morphologies of dendrites and cellular grains (Fig. 7(c)); and (5) the liquid immiscibility featuring multiple phase separation (as indicated in Fig. 7(b) and (c)) is seen extensively over the polished and etched sample surface. Multiple liquid phase separation is often found in the proximity of the -cristobalite crystals which exhibit both the dendritic and cellular morphology indicated in Fig. 7(c). It is the continuous type [23] of liquid phase separation, where one (i.e. the droplet phase indicated by feature (5) in Fig. 7(c)) of the two liquid phases is separated from the other by a continuous glass matrix. The observation is very significant since the extensive microcracking in the y=19.5 samFig. 7. (a) Five distinctive features of the y=13.9 sample, (b) dendritic growth, (c) multiple liquid-phase separation in the vicinity of dendrites and cellular growth (SEM-SEI). Fig. 6. (a) Dendritic and cellular growth and (b) characteristic twinning feature of the -crisotablite within residual pores (SEMSEI). ples of Fig. 5(b) is not detected from samples with y=13.9 of higher doping level after they have both been subjected to similar grinding and polishing. This implies that -cristobalite phase transformation has not occurred in the y=13.9 samples even after surface grinding and polishing.
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 4. Discussion lized B-phase. The indications are:(a) the codoping of Na,O+Al,O3) enables the stabilization of the high 4.1. Chemical and mechanical terms in the temperature cubic phase to room temperature, and(b) B-a-cristobalite phase transformation particle size [18, 25] is also a determining factor in the metastable retention, as discussed in the literature [6,7 The volume contraction upon the B-a phase trans- The stabilization mechanism is therefore likely to asso- formation of cristobalite and quartz results in dusting ciate with both the chemical and mechanical terms [7] of fused silica prepared from quartz sand as in the Again, it is similar to the stabilization of c-and tetrag Osram process [24]. For samples with y2 19.5, at-cristo- onal(t)-ZrO2 to room temperature. A critical mechani- balite is the predominant crystalline phase in the mix- cal constraint imposed by the matrix of either glass or ture(Fig. 4(a)). Although large cracks may have other (a+ B)-cristobalite grains would have existed as already existed in the as-sintered microstructure(as in the partially stabilized zro2(PSz)[8], ZrO2-tough arrowed in Fig. 5(a)), extensive microcracking only ened Al,O3(ZTA)[26] and t-ZrO2 polycrystals(TZP occur after sample grinding(e.g y=19.5 in Fig. 5(b)). [27]. The relief of the constrained stresses by surface Large cracks generated thermally are attributed to the nding has resulted in the transformation of the em- 5 vol. volume contraction upon the B-a-cristo- bedded t-ZrO2 particles to the monoclinic(m)-phase by balite phase transformation since no quartz was found a stress-induced martensitic phase transformation in these samples. Microcracks, however, are more likely mechanism [28]. Similarly, the B-+a-cristobalite pha to be of the stress-induced nature. They are originated transformation also occurs when the matrix constraint from the B-+a-cristobalite phase transformation, trig- has been removed by external stresses. Those particles gered off by the stress applied in grinding. It is evi- in the(Na++Al+)-codoped SiO2 ceramics containing denced from the increased amount of a-cristobalite solute cations below the critical concentration for full after sintered samples have been re-ground to fine stabilization then transform to a-cristobalite during powders(of 38-45 and 0. Since the The B-cristobalite survived from pulverizing to pow- inhomogeneity of solute oxides exists in the initial der, as represented by (111B in C-1(a)of Fig. I(c), must mixture of Na2O-containing colloidal silica sols doped have been fully stabilized to the extent that they have with additive salts, some of the cristobalite crystallized become 'non-transformable' meaning its resistance of with higher solute contents therefore become the fully the phase transformation to a-cristobalite even under stabilized B-phase. The lower limit of Na,O and Al2O3 grinding stresses. In fact, particles larger than 300-500 solute oxides would be an overall concentration of 6.29 nm, and not completely detached from the cracked mol% since samples with y= 13.9 contain only B-cristo matrix(Fig. 5(b)) after surface polishing, are probably balite as detected by XRD(Fig. 4(a). For the Na2O the survived non-transformable B-cristobalite. The ma- and Al2O3-additive content of 4.65 mol% (=19.5), trix transformed to a-phase and experienced the accom- below the critical solute level, crystallization(probably panied volume contraction has resulted in extensive to the spherical particles indicated in Fig. 5(b)may microcracking. It is also clear from Fig 4(a) that higher have also occurred from regions of higher solute con- doping level (with y= 13.9) has led to the fully stabi- tent than the overall concentration of the initial mix-
274 C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 4. Discussion 4.1. Chemical and mechanical terms in the -cristobalite phase transformation The volume contraction upon the phase transformation of cristobalite and quartz results in dusting of fused silica prepared from quartz sand as in the Osram process [24]. For samples with y19.5, -cristobalite is the predominant crystalline phase in the mixture (Fig. 4(a)). Although large cracks may have already existed in the as-sintered microstructure (as arrowed in Fig. 5(a)), extensive microcracking only occur after sample grinding (e.g. y=19.5 in Fig. 5(b)). Large cracks generated thermally are attributed to the 5 vol.% volume contraction upon the -cristobalite phase transformation since no quartz was found in these samples. Microcracks, however, are more likely to be of the stress-induced nature. They are originated from the -cristobalite phase transformation, triggered off by the stress applied in grinding. It is evidenced from the increased amount of -cristobalite after sintered samples have been re-ground to fine powders (of 38–45 and 38 m in size), as shown by C-1(a) and C-1(b) in Fig. 2 for y=24.6. The y=24.6 samples initially contain a mixture of - and -phase (C-1 in Fig. 2) (although not immediately discernible from the XRD trace, it has been demonstrated for samples with y=19.5 by deconvulsion (Fig. 1(b)). It is plausible that some of the initially metastable -cristobalite confined in the matrix has transformed to the -phase because the mechanical constraint is relieved by grinding and polishing. In fact, phase transformation to -cristobalite was also attributed [13] to the tensile stresses induced upon cooling due to differential thermal expansion. The actual stress state involving E (differential elastic moduli), (differential thermal expansion) and the sample grinding technique at present can not be assessed easily. The -phase found in Breuckener’s samples [13] was probably due to the stabilization by the chemical impurities associated with the initial colloidal silica (Vitresil® containing impurities of Al3+, Na+ and OH−). The -cristobalite survived from pulverizing to powder, as represented by (111) in C-1(a) of Fig. 1(c), must have been fully stabilized to the extent that they have become ‘non-transformable’ meaning its resistance of the phase transformation to -cristobalite even under grinding stresses. In fact, particles larger than 300–500 nm, and not completely detached from the cracked matrix (Fig. 5(b)) after surface polishing, are probably the survived non-transformable -cristobalite. The matrix transformed to -phase and experienced the accompanied volume contraction has resulted in extensive microcracking. It is also clear from Fig. 4(a) that higher doping level (with y=13.9) has led to the fully stabilized -phase. The indications are: (a) the codoping of (Na2O+Al2O3) enables the stabilization of the hightemperature cubic phase to room temperature, and (b) particle size [18,25] is also a determining factor in the metastable retention, as discussed in the literature [6,7]. The stabilization mechanism is therefore likely to associate with both the chemical and mechanical terms [7]. Again, it is similar to the stabilization of c- and tetragonal (t)-ZrO2 to room temperature. A critical mechanical constraint imposed by the matrix of either glass or other (+)-cristobalite grains would have existed as in the partially stabilized ZrO2 (PSZ) [8], ZrO2-toughened Al2O3 (ZTA) [26] and t-ZrO2 polycrystals (TZP) [27]. The relief of the constrained stresses by surface grinding has resulted in the transformation of the embedded t-ZrO2 particles to the monoclinic (m)-phase by a stress-induced martensitic phase transformation mechanism [28]. Similarly, the -cristobalite phase transformation also occurs when the matrix constraint has been removed by external stresses. Those particles in the (Na++Al3+)-codoped SiO2 ceramics containing solute cations below the critical concentration for full stabilization then transform to -cristobalite during surface grinding and/or pulverizing to powder. The -cristobalite (e.g. Fig. 2) that sustained pulverizing or polishing can therefore be perceived as being equivalent to the fully stabilized c-ZrO2 (FSZ). It still retains metastably to room temperature is therefore chemically ‘non-transformable’. The full stabilization of the cubic symmetry in cristobalite may require a lower limit of solute content similar to that at 13.2 mol% MgO for the fully stabilized c-ZrO2 [29]. Those transformed to -cristobalite upon surface grinding (Fig. 1(c) and 2) are only ‘partially’ stabilized when the modification of chemical free energy (Gchem term) [26] alone by solute cations (e.g. Mg2+ for ZrO2 and (Na+ +Al3+) for SiO2) is not sufficient to render full stabilization. The -crystals containing less solute content can only become metastable if augmented by the mechanical constraint from the matrix. It is necessary for these crystals to remain confined to a matrix (of glass and (+)-cristobalite mixture) in order to fulfil the thermodynamic requirement of Gtotal0. Since the inhomogeneity of solute oxides exists in the initial mixture of Na2O-containing colloidal silica sols doped with additive salts, some of the cristobalite crystallized with higher solute contents therefore become the fully stabilized -phase. The lower limit of Na2O and Al2O3 solute oxides would be an overall concentration of 6.29 mol% since samples with y=13.9 contain only -cristobalite as detected by XRD (Fig. 4(a)). For the Na2Oand Al2O3-additive content of 4.65 mol% (y=19.5), below the critical solute level, crystallization (probably to the spherical particles indicated in Fig. 5(b)) may have also occurred from regions of higher solute content than the overall concentration of the initial mix-
C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 ture. In particular when compositions studied in the formation to a-cristobalite, since its content was in- Al2O3-Na2O-SiO2 system could produce both the creased when particle size was reduced to 19.5(i.e. containing less amount of solute oxides lodal silica ceramics codoped with Na,O+Al,O,can than y=19.5)given in Fig. 1(a)(=19.5)and Fig. 2 be attributed to both the chemical and mechanical factors. Particle size effects similar to the ZrO-con taining ceramics appear to exist also with the stabi- 4.2. Critical particle size for B +a-cristobalite lized B-cristobalite phase embedded matrix a-cristobalite and the residual Na20-Al,O3- Sio2 glass A critical size appears to exist in the present sam- Codoping with Na,O and Al2O3, both at 6.29 mol%, ples for the metastable retention of B-cristobalite, be- has successfully produced crystalline phase of only B- low which the particles have survived the grinding cristobalite in the mixture by sintering at 1100C for stress. Reduction in the particle size favours the trans- 48h
C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 275 ture. In particular when compositions studied in the Al2O3 –Na2O–SiO2 system could produce both the ternary and binary eutectic liquids at temperatures between 740 and 867 °C [30]. Liquid-phase sintering at 1100 °C assisted by the eutectic liquids has indeed occurred [19] with the characteristic rounded grains of -cristobalite containing twin lamellae and glassy grain-boundary phase of a continuous nature. Since the ternary eutectic [30] at 740 °C is richer in Na2O and leaner in both SiO2 and Al2O3 than the compositions investigated here, the crystallization to -cristobalite incorporating less amount of Na+ is highly possible upon cooling. In fact, liquid-phase separation is only observed from samples with y=13.9 in Al2O3 –Na2O–ySiO2, in which the crystalline phase contains only -cristobalite (as revealed by XRD (Fig. 3)). When phase separation occurs in samples with y=13.9, solute content in the separated droplet phase (y=13.9 in Fig. 7(c)) appears to have adjusted to favour the crystallization to -cristobalite (Fig. 4(a)). Nevertheless, the actual content of (Al2O3+Na2O) in the retained and untransformable -phase can not be deduced from the present data before detailed chemical microanalysis. The lattice spacing of -cristobalite d101 increases with higher contents of Al2O3+Na2O (Fig. 4(b)), contrary to what was previously reported [7]. However, it leaves little doubt [9] now that taking up more stuffing cations [9–11] has led to the stabilization of the -phase. A possible interpretation for the reverse effect [12] when -critobalite was obtained is that impurities associated with the starting powder may have already stabilized the -phase chemically, but XRD failed to differentiate the (+)-cristobalite mixture (as demonstrated for 101 and 111 in Fig. 1(a) and 2). When the 101-peak at d101=0.4062 nm stops increasing its intensity and the 111-peak at d111= 0.4110 nm becomes predominant, the sintered sample would have incorporated impurities of higher content than that of y=19.5 (i.e. samples with y19.5 which contain both solute oxides greater than 4.65 mol% each and probably also with d-spacing larger than d101=0.4058 nm, as indicated by Fig. 4(a)). In fact, the predominant -peak (i.e. 101) could easily obscure any chemically stabilized -cristobalite of a minor amount, such as those detected in samples with y19.5 (i.e. containing less amount of solute oxides than y=19.5) given in Fig. 1(a) (y=19.5) and Fig. 2 (y=24.6). 4.2. Critical particle size for +-cristobalite A critical size appears to exist in the present samples for the metastable retention of -cristobalite, below which the particles have survived the grinding stress. Reduction in the particle size favours the transformation to -cristobalite, since its content was increased when particle size was reduced to 38 m (Fig. 2). This suggests that the relief of the confined particles, which would otherwise be the metastable -cristobalite, from the matrix constraints initiates the -phase transformation. However, those particles with sizes less than the critical value still retain - phase. Particles of 300–500 nm have become fully stabilized in the y=19.5 samples with 4.65 mol% Al2O3 and Na2O, even after the relief of matrix constraint by surface grinding (Fig. 5(b)). It may be argued by considering the total thermodynamic free energy change (Gtotal) for the size-dependent phase transformation of the unconstrained particles as in ZrO2 [26]. A critical particle size (rcritical) for just retaining the cubic -cristobalite phase to room temperature can be determined from the endpoint thermodynamic calculation of Gtotal=0 [26]. rcritical= −3(Gtotal)/(sv) where Gtotal is the difference in total free energy per unit volume, and sv is the difference in surface-tovapour surface energy of crystal. Since Gtotal=G−G is negative for the - cristobalite phase transformation to occur, sv, must be positive in order to keep rcritical rational. Total free energy per unit volume (Gtotal) consists of the chemical term (Gchemical) and mechanical term (Gstrain) of the crystal. The critical particle size would therefore fall in a range depending on the exact solute content in the -cristobalite crystals (i.e. depending on Gchemical). Indeed, -cristobalite crystals larger than 100 m have been reported [18] in commercial silica glass containing both Al2O3 and Na2O. For solutefree particles exceeding rcritical to retain -cristobalite, however, sufficient strain energy (Gstrain) would have to be induced by the differential thermal expansion (of =5×10−6 °C−1 ) [25] between -cristobalite and the vitreous silica matrix to compensate for a negative Gchemical. 5. Conclusions The metastable retention of the high temperature -cristobalite to room temperature in sintered colloidal silica ceramics codoped with Na2O+Al2O3 can be attributed to both the chemical and mechanical factors. Particle size effects similar to the ZrO2-containing ceramics appear to exist also with the stabilized -cristobalite phase embedded in a matrix of -cristobalite and the residual Na2O-Al2O3-SiO2 glass. Codoping with Na2O and Al2O3, both at 6.29 mol%, has successfully produced crystalline phase of only - cristobalite in the mixture by sintering at 1100 °C for 48 h.
276 C -H. Chao, H.Y. Lu/Materials Science and Engineering 4328(2002)267-276 Acknowledgements [I E.S. Thomas, J.G. Thompson, R L. Withers, M. Sterns, Y Xiao, RJ. Kirkpartick, J. Am. Ceram Soc. 77(1994)49 This research was financed by the National Science 12A.G. Verduch, J. Am. Ceram. Soc. 41(1958)327. Council of Taiwan through NSC-82-0405-D-110-003 (14 B.D. Cullity, Introduction to X-ray Diffraction, Addison and83-0405-D-110-001 [5 E.H. Bogardus, R. Roy, J. Am. Ceram Soc. 48 [16 H.I. Hsiang, F.S. Yen, J. Am. Ceram. References [7 A H. Heuer, J. Am. Ceram Soc. 70(1987)689. [18 F.E. Wagstaff, J. Am. Ceram. Soc. 51(1968)449 P.J. Heaney, in: P.J. Heaney, C.T. Prewitt(Eds ) Silica-Physical [19 C.H. Chao, PhD thesis, National Sun Yat-Sen University, Tai- Behaviour, Geochemistry and Material Applications, Mineral wan,1998 Soc. Am, Washington, DC, 1994, pp. 1-40 [20 C H. Chao, H.Y. Lu, Mater. Sci. Eng. A282(2000)123 2]KD. Hammonds, M.T. Dove, A P. Giddy, V. Heine, B. Win- 21]M.H. Lewis, J. Metcalf-Johansen, P.S. Bell, J. Am. Ceram Soc kler. Am. Miner. 81(1996)1057. 62(1979)278 3M.T. Dove, M. Bambhir, K D. Hammonds, V. Heine, A K.A. 222 J.H. Jean, T.K. Gupta, J. Mater. Res. 7(1992)3013 Pryde, Phase Transition 58(1996)121 [23W. Vogel, Chemistry of Glass, Am. Ceram Soc, Columbus, OH 4 A.F. Wright, A. Leadbetter, Philos. Mag. 31(1975)1391 1985 5R.L. Withers, J.G. Thompson, T.R. Welberry, Phys. Chem. 24]G H. Beall, in: P.J. Heaney, C.T. Prewitt(Eds ) Silica-Physical minerals.16(1989)517 Behaviour, Geochemistry and Material Applications, Mineral [6 F. Aumento, Am. Miner. 51(1966)1167. Soc. Am, Washington, D. C, 1994, pp. 469-505 [7] I.P. Swainson, M.T. Dove, J. Phys. Condens. Matter 7(1995) [25] V.M. Castano, T. Takamori,MW.Shafer, Am.CeramSoc.70 [8]A H. Heuer, M. Ruhle, in: N. Claussen, M. Ruhle. A H. Heuer [26]DJ. Green, R H.J. Hannink, M V. Swain, Transformation (Eds ) Advances in Ceramics, Science and Technology of Zirco- Toughening of Ceramics, CRS Press. Boca Raton, FL. 1989 nia (I), vol. 12, Am. Ceram. Soc, Westerville, OH, 1988, pp. 27 H.Y. Lu, S.Y. Chen, J. Am. Ceram. Soc. 70(1988)53 [28R.C. Garvie. R H.J. Hannink, R.T. Pascoe, Nature (London 9)AJ. Perrotta, D K. Grubbs, E.S. Martins, N.R. Dando, H.A. 258(1976)70 McKinstry, C.Y. Huang, J. Am. Ceram Soc. 72(1989)441 229R. H.. Hannink, J. Mater. Sci. 18(1983)457 [0J M.A. Saltzberg, S L. Bors, H. Bergna, S.C. Winchester, J. Am B0 E.M. Levin, C.R. Robbins, H F. McMurdie, Phase Diagrams for Ceram.Soc.75(1992)89 Ceramists (D), Fig 501, Am. Ceram Soc., Columbus, OH, 1964
276 C.-H. Chao, H.-Y. Lu / Materials Science and Engineering A328 (2002) 267–276 Acknowledgements This research was financed by the National Science Council of Taiwan through NSC-82-0405-D-110-003 and 83-0405-D-110-001. References [1] P.J. Heaney, in: P.J. Heaney, C.T. Prewitt (Eds.), Silica-Physical Behaviour, Geochemistry and Material Applications, Mineral. Soc. Am, Washington, DC, 1994, pp. 1–40. [2] K.D. Hammonds, M.T. Dove, A.P. Giddy, V. Heine, B. Winkler, Am. Miner. 81 (1996) 1057. [3] M.T. Dove, M. Bambhir, K.D. Hammonds, V. Heine, A.K.A. Pryde, Phase Transition 58 (1996) 121. [4] A.F. Wright, A.J. Leadbetter, Philos. Mag. 31 (1975) 1391. [5] R.L. Withers, J.G. Thompson, T.R. Welberry, Phys. Chem. Minerals. 16 (1989) 517. [6] F. Aumento, Am. Miner. 51 (1966) 1167. [7] I.P. Swainson, M.T. Dove, J. Phys. Condens. Matter 7 (1995) 1771. [8] A.H. Heuer, M. Ru¨hle, in: N. Claussen, M. Ruhle, A.H. Heuer (Eds.), Advances in Ceramics, Science and Technology of Zirconia (II), vol. 12, Am. Ceram. Soc, Westerville, OH, 1988, pp. 3–20. [9] A.J. Perrotta, D.K. Grubbs, E.S. Martins, N.R. Dando, H.A. McKinstry, C.Y. Huang, J. Am. Ceram. Soc. 72 (1989) 441. [10] M.A. Saltzberg, S.L. Bors, H. Bergna, S.C. Winchester, J. Am. Ceram. Soc. 75 (1992) 89. [11] E.S. Thomas, J.G. Thompson, R.L. Withers, M. Sterns, Y. Xiao, R.J. Kirkpartick, J. Am. Ceram. Soc. 77 (1994) 49. [12] A.G. Verduch, J. Am. Ceram. Soc. 41 (1958) 327. [13] R. Brueckner, J. Non-Cryst. Solids 5 (1970) 123. [14] B.D. Cullity, Introduction to X-ray Diffraction, AddisonWesley, New York, 1978. [15] E.H. Bogardus, R. Roy, J. Am. Ceram. Soc. 48 (1965) 205. [16] H.I. Hsiang, F.S. Yen, J. Am. Ceram. Soc. 78 (1996) 1053. [17] A.H. Heuer, J. Am. Ceram. Soc. 70 (1987) 689. [18] F.E. Wagstaff, J. Am. Ceram. Soc. 51 (1968) 449. [19] C.H. Chao, PhD thesis, National Sun Yat-Sen University, Taiwan, 1998. [20] C.H. Chao, H.Y. Lu, Mater. Sci. Eng. A282 (2000) 123. [21] M.H. Lewis, J. Metcalf-Johansen, P.S. Bell, J. Am. Ceram. Soc. 62 (1979) 278. [22] J.H. Jean, T.K. Gupta, J. Mater. Res. 7 (1992) 3013. [23] W. Vogel, Chemistry of Glass, Am. Ceram. Soc, Columbus, OH, 1985. [24] G.H. Beall, in: P.J. Heaney, C.T. Prewitt (Eds.), Silica-Physical Behaviour, Geochemistry and Material Applications, Mineral. Soc. Am, Washington, D.C, 1994, pp. 469–505. [25] V.M. Castano, T. Takamori, M.W. Shafer, Am. Ceram. Soc. 70 (1987) C77. [26] D.J. Green, R.H.J. Hannink, M.V. Swain, Transformation Toughening of Ceramics, CRS Press, Boca Raton, FL, 1989. [27] H.Y. Lu, S.Y. Chen, J. Am. Ceram. Soc. 70 (1988) 537. [28] R.C. Garvie, R.H.J. Hannink, R.T. Pascoe, Nature (London) 258 (1976) 703. [29] R.H.J. Hannink, J. Mater. Sci. 18 (1983) 457. [30] E.M. Levin, C.R. Robbins, H.F. McMurdie, Phase Diagrams for Ceramists (I), Fig. 501, Am. Ceram. Soc., Columbus, OH, 1964