Marys Medicine

Doi:10.1016/j.jnoncrysol.2003.08.075

Journal of Non-Crystalline Solids 331 (2003) 217–227 TEM and XRD study of early crystallization of lithium disilicate glasses P.C. Soares Jr. a, E.D. Zanotto a,*, V.M. Fokin a, H. Jain b a LaMaV – Vitreous Materials Laboratory, Universidade Federal de S ao Carlos, 13595-905 S ao Carlos, SP, Brazil b Department of Materials Science and Engineering, Lehigh University, 18015 Bethlehem, PA, USA Received 1 April 2003 Numerous researchers have speculated the precipitation of metastable phase(s) in the early stages of crystallization of lithium disilicate glass to explain the large discrepancies between the predictions of the classical nucleation theory(CNT) and experimental data. Therefore, we have investigated the early and intermediate stages of crystallization ofthree glasses on both sides of the stoichiometric composition through direct observations by transmission electronmicroscopy (TEM)/selected area electron diffraction (SAED) and X-ray diffraction (XRD). In samples heat-treated at
Tg ¼ 454 °C, two distinct crystalline phases, stable lithium disilicate (LS2) and metastable lithium metasilicate (LS)coexist up to 120 h at 454 °C (crystalline fraction <1 vol.%). For longer treatments (240–600 h) only the stable phase(LS2) was observed. These results suggest that in the early stages, simultaneous homogeneous nucleation of both LS andLS2 takes place. As treatment time and crystallized fraction increase, the relative number of LS crystals decreases.
Therefore, the precipitation of the LS phase does not disturb the nucleation of the stable LS2 phase and thus cannotexplain the failure of CNT in predicting the nucleation rates in this glass.
Ó 2003 Elsevier B.V. All rights reserved.
ergy [2,5] and the elastic stresses that arise oncrystal nucleation [6,7]. Moreover, there is a pos- It is well established that the classical nucleation sibility that metastable crystalline phases of small theory (CNT) underestimates the steady-state nu- surface energy could nucleate first in the early cleation rates in glasses by many orders of mag- stages inducing heterogeneous nucleation of the nitude [1,2]. Several assumptions in the theory may stable phase or transform into the stable form at be responsible for such discrepancy [3], including later times [8].
its failure for the smallest nuclei for which the The nucleation and crystallization kinetics of crystal/glass interface is not sharp [4], a possible lithium silicate glasses close to Li2OÆ2SiO2 (LS2) temperature or size dependence of the surface en- composition have been studied intensively for de-cades, because this glass can be made easily, detailed thermodynamic data are available in the Corresponding author. Tel.: +55-16 260 8556/260 8527; fax: literature, and internal crystal nucleation and +55-16 261 5404.
E-mail address: (E.D. Zanotto).
growth kinetics can be measured conveniently.
0022-3093/$ - see front matter Ó 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2003.08.075

P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 Fig. 1. Phase diagram of the Li2O–SiO2 system according to Migge [9], showing solid-solution regions determined by West and Glasser[10] (dashed area), and Deubener [11] (circles). The figure also shows the liquid–liquid phase separation region, according to Zanotto[12].
Fig. 1 shows a phase diagram of this system. In have been used (SAXS, dielectric relaxation, equilibrium at 454 °C, both slightly hypo-stoi- Raman spectroscopy, XPS, e.g. [14–19]) they only chiometric (<33.3 mol% Li2O) and slightly hyper- give indirect evidence of a metastable phase for- stoichiometric LS2 compositions (>33.3 mol% mation, and hence do not resolve the controversy Li2O) should crystallize as LS2-ss (solid-solution).
unambiguously. Reviews on this issue can be While several authors suggested the possibility found in Refs. [20–23].
of a metastable phase precipitation (discussed Towards direct observations, James and Keown below), Zanotto and Leite [13] concluded that [24] used TEM/SAED for studying the nucleation metastable phases, if present at all, do not have of a stoichiometric lithium disilicate glass heat- any significant impact on the overall crystalliza- treated in the 450–490 °C temperature range, for tion kinetics of LS2 glass. This is one of the major up to 150 h. Electron diffraction patterns showed controversies concerning crystallization kinetics in only the stable phase LS2. On the other hand, in this system and has not been resolved conclusively another TEM study, Deubener et al. [25] observed a transient phase in a slightly hyper-stoichiometric In addition, the different phases reported to (33.5 mol% Li2O) sample heat-treated at 454 °C precipitate in the early stages of crystallization in for 7 h. That phase was indexed on the basis of LS this system have not been clearly identified yet.
unit cell parameters, but with doubled c-lattice Some authors, contradicting others who could not constant. After further heat treatment for 40 h at detect any phase prior to stable LS2, have sug- 454 °C, two phases, LS2 and the transient phase, gested the appearance of a metastable phase for were indexed. However, Deubener [26] mentioned certain heat treatments. Though many techniques later that their results were subjected to some un- P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 certainty due to fast degradation of the crystals phase, referred as a0-LS2. After longer heat treat- under the electron beam. Crystal degradation was ment times at the same temperature, most of the also observed in Ref. [27].
peaks of this phase were present though with de- Soares Jr. [27] also observed a phase different creasing intensity, along with the stable LS2 phase.
from LS2 using TEM, in a hypo-stoichiometric This metastable phase was not observed by TEM/ glass (32.5 mol% Li2O) nucleated at 454 °C for SAED. After 551 h, apart from LS2, they observed 5–20 h. Heat-treated samples were prepared by small traces of stable LS phase by XRD, which chemical thinning. Two different phases were was attributed to the slight hyper-stoichiometry of found: LS2 and LS, but the LS crystals were in- the glass. According to the phase diagram (Fig. 1), dexed only in the [0 0 1] direction. These results only LS2-ss may crystallize as a stable phase in the allowed him to identify the a and b lattice pa- glass with 33.9 mol% Li2O at 454 °C. Another rameters only. Samples nucleated for 20 h at 454 metastable phase, b0-LS2, was identified by TEM/ °C were given a growth treatment at 610 °C for 10 SAED in samples nucleated for 331 and 502 h at min, yielding only the stable LS2 phase.
454 °C. In recent papers, Burgner et al. [20,21] Iqbal et al. [22] followed the crystallization have repeated the XRD experiments of Iqbal et al.
stages in hyper-stoichiometric LS2 glasses with [22] in a series of LS2 glasses heat-treated for long 33.9 mol% Li2O by TEM and XRD and showed times (120–600 h) at 454 and 465 °C. No phase that two metastable phases (different from LS) other than the stable lithium disilicate was ob- persisted to very long times at 454 °C. In their investigation, glass samples were heat-treated at In summary, there is some evidence that dif- 454 °C for time periods of 50–551 h. After 120 h at ferent metastable phases may form during the 454 °C, they observed by XRD only a metastable early stages of crystallization in LS2 glasses, but time at 454°C (h) Fig. 2. Estimated volume fraction of crystals, a, in stoichiometric lithium disilicate glass heat-treated at 454 °C. Nucleation rate datafrom Zanotto [12] and growth rate measured from TEM sizes. The volume fraction of crystals was determined using the JMAKequation for crystals of elongated shape [13].
P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 Table 1Composition of hypo, hyper and stoich glasses obtained by DSC and by inductive plasma spectrometry (ICP) Average composition Mol% Li2O (±0.7) Mol% Li2O (±0.5) these were not clearly identified so far and the role Li2O (hyper) as shown in Table 1. They are hypo- metastable phases play with respect to the crys- stoichiometric, stoichiometric and hyper-stoichio- tallization mechanism remains unclear. In par- metric, respectively. These chemical analyses were ticular, there is considerable uncertainty over repeated twice, and to obtain more accurate results whether metastable phases form and serve as we also performed differential scanning calori- precursors for subsequent precipitation of the metric (DSC) experiments according to a proce- stable LS2 phase (which could partially explain dure proposed by Fokin and Ugolkov [28]. Bulk the apparent failure of CNT), or if the precipita- pieces of glass (40 mg) were analyzed in a Net- tion of the stable phase occurs concurrently and zsch 404 DSC, using Pt crucibles and 10 K/min independently of the formation of metastable heating rate. The crystallization peak temperatures were related to molar fractions of Li2O. Results In this paper we shed further light on the are shown in Fig. 3.
complex crystallization mechanism of lithium di- Heat treatments were carried out in a vertical silicate glasses. We observe the very early stages of tube furnace with the temperature controlled crystallization (<1 vol.% crystallized fraction, as within ±1 °C. Samples were given single stage heat shown in Fig. 2) using TEM/SAED, and present treatments at the temperature of maximum nu- XRD results for samples with higher crystalline cleation rate, Tg ¼ 454 °C for periods of 2.5, 5, 10, fractions (up to 35 vol.%), thus overlapping both 20, 50, 120 and 312 h for the TEM; and 120, 240, techniques. The main objective is to systematically 360, 480 and 600 h for the XRD experiments. The identify the crystallization pathways and the pos- crystallized surface layer of the heat-treated glass sible formation of metastable phases in the glass samples, which may have foreign phases, was re- volume, in an attempt to explain the controversies moved before characterization.
of the previous studies.
We used two methods to prepare TEM samples.
The crushing method, based on Ref. [29], where afew grams of glass were crushed into fine powder using an agate mortar. The resulting fine powderwas then dispersed in pure ethyl alcohol in a 50 ml beaker by keeping the solution in an ultrasonic Li2OÆ2SiO2 composition were prepared using bath for a few minutes. A carbon coated 200-mesh standard reagent grade Li2CO3 (Aldrich Chem.
TEM copper grid was placed on a wire support Co., 99+%) and ground Brazilian quartz, >99.9% inside the beaker. The suspension decanted on the SiO2. The 200 g batches were melted in a Pt cru- grid. After its removal from the beaker the alcohol cible at 1450 °C for 2 h in an electric furnace. To evaporated, leaving the glass particles adhering to ensure homogeneity, the poured glasses were the grid surface. We also used the chemical thin- ground and remelted at the same temperature for ning method, where 150 lm thick discs of 3 mm one additional hour. The melts were quenched by diameter were dimpled up to 10 lm, and then pressing between steel plates. Subsequent chemical chemically thinned in a solution of 15HF–5HCl– analysis revealed the composition of the prepared 75H2O (by vol.%). Once perforation occurred, the glasses to be 32.5 ± 0.5 mol% Li2O (hypo), samples were washed in ethyl alcohol and distilled 33.3 ± 0.5 mol% Li2O (stoich) and 34.6 ± 0.5 mol% P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 1 - glass hypo (32.98 %) 2 - glass stoich (33.17 %) 3 - glass stoich (33.20 %) 4 - glass hyper (34.26 %) 5 - glass hyper (34.55 %) solid circles [28] Fig. 3. DSC crystallization peak temperature as a function of glass composition: (d) Fokins glasses [28]; (s) glasses hypo, hyper andstoich.
TEM analysis was carried out at 120 kV using a Among all known phases listed in Table 2, we Philips EM-420T microscope at Lehigh Univer- observed through TEM two different crystalline sity, USA and a Philips CM120 microscope at the phases in all samples heat-treated from 2.5 to 120 Universidade Federal de S ao Carlos, Brazil.
h at 454 °C, which were indexed as LS2 (from XRD data were collected on monolithic sam- Liebau [34] and De Jong et al. [35]) and LS (from ples on two different equipments: (i) a conven- Hesse [37]). According to the phase diagram, at temperatures below 550 °C, the LS phase should CuKa radiation (k ¼ 1:54  A) at 50 kV and 100 not appear as a stable phase in the three glasses mA, where the samples were scanned from 15° to (hypo, stoich and slightly hyper). Hence LS is a 45°, in steps of 0.02° for 3 s; (ii) a synchrotron metastable phase in these glasses.
radiation source at LNLS, Campinas, Brazil, Examples of bright field TEM micrographs of where the samples were scanned from 10° to 32° crystals and their related electron diffraction pat- terns are shown in Figs. 4 and 5. Fig. 4(a) shows alithium metasilicate crystal observed in a stoichglass sample treated for 20 h at 454 °C. The cor- responding electron diffraction (Fig. 4(b)) revealsthat the crystal was oriented with the [0 0 1] plane The selected area diffraction patterns were in- parallel to the beam direction. Fig. 5(a) shows a dexed using a commercial software [30] based on lithium disilicate crystal observed in a sample of the method described by Goodhew [31] and Bee- hyper glass (nucleated at 454 °C for 312 h), indexed ston et al. [32], with structure details from the with the [3 1 2] plane parallel to the beam direction.
ICSD database [33]. Gold standard samples were After 312 h at 454 °C (a  3%), we observed sev- used for the camera length calibration. Indexing of eral crystals with sizes ranging from 0.5 to 7 lm in the phases was performed using the lattice con- chemically thinned samples of the hyper glass. All stants listed in Table 2.
the crystals observed were identified as stable LS2.

P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 Table 2Crystallographic data for lithium silicate phases Crystalline system 5.82, 14.66, 4.79, 90.08 5.81, 14.58, 4.77 De Jong et al. [35] 5.73, 14.64, 4.79 Deubener et al. [25] 5.68, 4.78, 14.65 Smith et al. [36] 5.14, 6.1, 5.3, 90.5 Vollenkl et al. [38] Vollenkl et al. [39] Fig. 4. TEM micrograph of the stoichiometric glass (33.3 mol% Li2O) heat-treated for 20 h at 454 °C: (a) bright field image;(b) corresponding electron diffraction pattern; (c) indexed as LS phase in the [0 0 1] direction. The bar denotes 150 nm.
One could imagine that the LS crystals were in crystallized 1%). These peaks were identified as some way dissolved chemically, but then they were the stable LS2 phase (JCPDS 40-0376). After fur- observed clearly in chemically thinned samples ther heating, at 360, 480 and 600 h (5, 16 and 35 heat-treated at 454 °C for 10 and 20 h (hypo and vol.% crystalline fraction, respectively) the inten- hyper glasses).
sity of these peaks increased. Results from syn- Fig. 6(a) shows the XRD results of the hyper chrotron radiation XRD showed only the LS2 glass samples heat-treated at 454 °C for longer phase (Fig. 6(b)).
periods than the TEM treatments. There is nocrystalline material detected in the sample heated for 120 h (a  0:07%). Diffraction peaks start toemerge against the amorphous background after Fig. 7 shows the observed crystal size of the heating for 240 h (estimated volume fraction largest crystals of both phases as a function of heat

P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 Fig. 5. TEM micrograph of the hyper-stoichiometric glass (34.6 mol% Li2O) heat-treated for 312 h at 454 °C: (a) bright field image;(b) corresponding electron diffraction pattern; (c) indexed as the LS2 phase in the [3 1 2] direction. The bar denotes 2 lm.
treatment time at 454 °C in hypo, hyper and stoich composition. For the three glasses LS is a meta- glasses. From Fig. 7 we find that the metastable LS stable phase. Therefore, the following two possible phase (open symbols) appears simultaneously with situations could occur: the stable LS2 phase in all samples up to 120 h (<1vol.% fraction crystallized). The growth rate of LS i(i) There is simultaneous homogeneous nucleation (slope of the curve) is nearly zero. On the other of LS2 and LS phases, but LS phase disappears hand, the size of the LS2 crystals follows reason- at some point during the heat treatment. In this ably well the estimated values of Dmax (maximum case LS is a metastable phase that does not af- diameter of crystals) using extrapolated crystal fect the nucleation path of LS2; growth rate data from higher temperatures [40].
(ii) The LS crystals nucleate first and then the LS2 Fig. 8 shows the relative frequency of occur- crystals nucleate heterogeneously over the LS.
rence (number of crystals of each phase divided by This mechanism could partially explain the the total number of crystals) of both phases with failure of the classical nucleation theory.
heat treatment time, including all indexed crystalsof all samples. Despite the poor statistics (on av- The fact that lithium metasilicate forms initially erage only a dozen crystals were detected in the in a larger number than the disilicate phase was TEM samples for each treatment time), we observe suggested before by Hench et al. [14]. A reasonable that up to 20 h, the relative number of LS crystals argument is that [–SiO2–] units (Q2 chains, typical is larger than LS2. This number decreases with of LS) are more readily organized than the Q3 increasing treatment time and, in spite of this de- layers present in the disilicate (here Qn refers to creasing LS/LS2 ratio, LS crystals are still present SiO4 units with n bridging oxygens). It has also at 120 h (0.07 vol.% crystallized fraction). After been suggested that the formation of the lithium 312 h (3 vol.% crystallized fraction), only the sta- metasilicate phase enriches the glass matrix in SiO2 ble LS2 phase is observed. These results show that increasing the glass viscosity in its vicinity, which both phases coexist in the early stages of crystal- presumably restricts the growth. Based on the lization in glasses around the lithium disilicate present results, the mechanism proposed is that P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 Fig. 6. XRD diffraction patterns of the hyper glass samples heated at 454 °C for 120–600 h. Vertical lines correspond to LS2 patternfrom JCPDS 40-0376. Using: (a) conventional diffractometer, (b) synchrotron radiation.
both phases nucleate concurrently at the very early number of LS crystals decreases, and they are no stages of nucleation (up to 10 h at 454 °C), but LS longer observed after 120 h at 454 °C. As signifi- overcomes LS2 due to the ease of its formation. As cant amount of glassy phase remains after 312 h at the treatment time increases, nucleation and 454 °C (97 vol.% glassy), we can thus safely state growth of LS becomes restricted, as shown by its that equilibrium has not been reached and that almost zero growth rate (Fig. 7). The relative lithium metasilicate is a metastable phase that will P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 LS2 - glass hypo LS - glass hypo LS - glass stoich LS - glass stoich LS - glass hyper LS - glass hyper U [40] time at 454 C (h) Fig. 7. Maximum dimension (Dmax) of the largest crystals observed by TEM in samples of hypo, stoich and hyper glasses as a functionof heat treatment time (in hours) at 454 °C. U refers to calculated Dmax using extrapolated crystal growth rate data from highertemperatures [40].
time at 454 C (h) Fig. 8. Ratio of number of crystals observed for each phase by the total number of crystals found in all samples as a function oftreatment time.
disappear at some stage. We thus have no evidence As in general, nucleation treatments are carried that the lithium metasilicate crystals induce het- out for short times, typically less than 20 h at  Tg erogeneous nucleation of lithium disilicate.
before the growth treatment, our results show that P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 up to 20 h at 454 °C the LS phase is dominant and, [3] E.G. Rowlands, P.F. James, Phys. Chem. Glasses 20 (1) therefore, the measured nucleation rates often re- [4] S. Sen, T. Mukerji, J. Non-Cryst. Solids 246 (1999) 229.
ported in the literature could, in principle, reflect [5] M.C. Weinberg, E.D. Zanotto, S. Manrich, Phys. Chem.
the kinetics of precipitation of LS and not of LS2.
Glasses 33 (3) (1992) 99.
However, XRD results for double-treated samples oller, J. Schmelzer, I. Gutzow, J. Non-Cryst. Solids (as in most published research) reveal only LS 240 (1–3) (1998) 131.
sample of the stoich glass was treated at 454 °C for [7] J. Schmelzer, I. Gutzow, J. M€ oller, J. Glass Phys. Chem. 24 (3) (1998) 244.
10 h (temperature with the highest LS/LS2 ratio) [8] A. Hishinuma, D.R. Uhlmann, J. Non-Cryst. Solids 95&96 and then at 600 °C for 30 min. Only the LS2 phase was detected, showing that LS dissolves in the [9] H. Migge, J. Nucl. Mater. 151 (1988) 101.
second heating path.
[10] A.R. West, F.P. Glasser, in: Advances in Nucleation and Crystallization in Glasses, American Ceramic Society,Columbus, OH, 1971.
[11] J. Deubener, J. Non-Cryst. Solids 274 (1–3) (2000) 195.
[12] E.D. Zanotto, PhD thesis, Sheffield University, UK, We studied the early and intermediate stages of [13] E.D. Zanotto, M.L.G. Leite, J. Non-Cryst. Solids 202 crystallization of hypo, hyper and stoichiometric lithium disilicate glasses heat-treated at Tg ¼ 454 [14] L.L. Hench, S.W. Freiman, D.L. Kinser, Phys. Chem.
°C for 2.5 to 600 h (35 vol.% crystallized frac- Glasses 12 (2) (1971) 50.
tion). TEM analyses of specimens provided clear [15] D.L. Kinser, L.L. Hench, in: Proceedings of the American Ceramic Society – 69th Annual Meeting, New York, 1968.
evidence of a second phase, viz. lithium metasili- [16] S.W. Freiman, L.L. Hench, J. Am. Ceram. Soc. 51 (7) cate (LS), which nucleates simultaneously with the stable LS2 phase.
[17] I, Joseph, L.D. Pye, in: Proceedings of the 14th Interna- tional Congress on Glass, New Delhi, 1, 1986, p. 358.
2 ratio decreases with heat treatment time up to 120 h (<0.1 vol.% crystallized fraction).
[18] J.W. Adams, B.H.W.S. De Jong, in: Proceedings of the Fall Meeting Symposium, MRS 321, 1994, p. 239.
These results support the following crystallization [19] E.D. Zanotto, J. Non-Cryst. Solids 219 (1997) 42.
mechanism: in the early stages there is concurrent [20] L.L. Burgner, P. Lucas, M.C. Weinberg, P.C. Soares Jr., homogeneous nucleation of both phases (LS and E.D. Zanotto, J. Non-Cryst. Solids 274 (1–3) (2000) 188.
[21] L.L. Burgner, M.C. Weinberg, P. Lucas, P.C. Soares Jr., but LS phase disappears at some point during heat treatment. In this case, LS is a meta- E.D. Zanotto, J. Non-Cryst. Solids 255 (1999) 264.
[22] Y. Iqbal, W.E. Lee, D. Holland, P.F. James, J. Non-Cryst.
stable phase (in all glasses in the vicinity of lithium Solids 224 (1998) 1.
disilicate composition) that does not play a role in [23] P.C. Soares Jr., P.A.P. Nascente, Phys. Chem. Glasses 43 the nucleation path of LS2, and thus cannot ex- (3) (2002) 151.
plain the discrepancy between the theoretical and [24] P.F. James, S.R. Keown, Philos. Mag. 30 (4) (1974) 789.
experimental crystallization kinetics.
[25] J. Deubener, R. Br€ uckner, M. Sternizke, J. Non-Cryst.
Solids 163 (1993) 1.
[26] J. Deubener, personal communication, Jena, 1994.
[27] P.C. Soares Jr., MSc dissertation, Universidade Federal de ao Carlos, Brazil, 1997.
[28] V.M. Fokin, V.L. Ugolkov, personal communication, S The authors thank the financial support of Carlos, 2002.
[29] B. Roy, H. Jain, S.K. Saha, D. Chakravorty, J. Non-Cryst.
Solids 183 (1995) 268.
[30] CaRIne Crystallography v.3.0., Divergent S.A. Compiegne, [31] P. Goodhew, Electron diffraction, [1] E.D. Zanotto, P.F. James, J. Non-Cryst. Solids 74 (1985) [32] B.E.P. Beeston, R.W. Horne, R. Makham, Practical Methods in Electron Microscopy – Part II: Electron [2] V.M. Fokin, E.D. Zanotto, J. Non-Cryst. Solids 265 (2000) Diffraction and Optical Diffraction Techniques, North Holland, Cambridge, 1972.
P.C. Soares Jr. et al. / Journal of Non-Crystalline Solids 331 (2003) 217–227 [33] ICSD – Inorganic Crystal Structure Database, Fiz Karls- [37] K.F. Hesse, Acta Crystallogr. B 33 (1977) 901.
ruhe and Gmelin-Institut, 1996.
[38] H. Vollenkl, A. Wittmann, H. Nowotny, Monat. Chem.
[34] F. Liebau, Acta Crystallogr. 14 (1961) 389.
101 (3) (1970) 684.
[35] B.H.W.S. De Jong, P.G.G. Slaats, H.T.J. Sup [39] H. Vollenkl, A. Wittmann, H. Nowotny, Monat. Chem. 99 Veldman, A.L. Spek, J. Non-Cryst. Solids 176 (1994) 164.
(4) (1968) 1360.
[36] R.I. Smith, R.A. Howie, A.R. West, Acta Crystallogr. C 46 [40] V.M. Fokin, E.D. Zanotto, J.W.P. Schmelzer, J. Non- Cryst. Solids 278 (2000) 24.

Source: http://www.lamav.ufscar.br/artpdf/jncs331b.pdf