Microstructural study of silica-doped zirconia ceramics

Acta mater. 48 (2000) 4647–4652 www.elsevier.com/locate/actamat

MICROSTRUCTURAL STUDY OF SILICA-DOPED ZIRCONIA CERAMICS L. GREMILLARD*, T. EPICIER, J. CHEVALIER and G. FANTOZZI GEMPPM, CNRS 5510, Baˆtiment 502, INSA-Lyon, F-69621 Villeurbanne Cedex, France

Abstract—The aim of this study was to show the effects of small silica additions on the microstructures and mechanical properties of 3 mol% yttria-stabilised zirconia (3Y-TZP) ceramics. Experiments were conducted on different batches of 3Y-TZP (pure to 2.5 wt% silica-doped). Microstructures were characterised mainly by transmission electron microscopy (TEM), but also by scanning electron microscopy (SEM) and X-ray diffraction (XRD). Silica was found at triple junctions, but neither at grain boundaries nor in the lattice. Undoped zirconia ceramics exhibited faceted grains and significant internal stresses, while doped zirconias showed a much more rounded microstructure and a lower level of internal stresses. Low-temperature degradation (LTD) and slow crack growth (SCG) measurements were conducted on the different batches. The addition of silica strongly increases LTD resistance without affecting the SCG behaviour. The microstructural origins of the different behaviours are discussed.  2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved. Re´sume´—Cette e´tude a pour but de montrer l’effet d’un dopage a` la silice sur la microstructure et les proprie´te´s me´caniques de diffe´rentes zircones stabilise´es avec 3% molaire d’oxyde d’yttrium (3Y-TZP) (pures ou dope´es a` la silice (jusqu’a` 2.5% en masse)). Les microstructures ont e´te´ observe´es principalement au Microscope Electronique en Transmission (MET), mais aussi au Microscope Electronique a` Balayage (MEB), et en diffraction des rayons X. La silice se retrouve principalement aux joints triples, mais ni aux joints de grains ni dans le re´seau. La zircone non dope´e est constitue´e de grains pre´sentant des angles vifs et pre´sente des contraintes internes significatives, alors dans les zircone dope´es les grains sont beaucoup plus ronds et les contraintes internes moins importantes. Des mesures de vieillissement et de propagation sous critique des fissures ont e´te´ mene´es sur les diffe´rentes zircones. L’addition de silice accroıˆt la re´sistance au vieillissement sans de´te´riorer les proprie´te´s me´caniques. L’origine microstructurale de ces diffe´rents comportements est analyse´e.  2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved. Keywords: Transmission electron microscopy (TEM); X-ray diffraction (XRD); Microstructure; Structural ceramics; Zirconia

1. INTRODUCTION

3 mol% Yttria-stabilised zirconia (3Y-TZP) ceramics have been extensively studied [1] and used in many applications because of their unique advantages at room temperature. Indeed, it was shown in the 1970s that zirconia exhibited a phase transformation toughening acting to resist to crack propagation. The stressinduced phase transformation of metastable tetragonal phase towards the monoclinic phase at the crack tip is accompanied by a volumetric expansion that induces compressive stresses, reducing the driving force for crack propagation. 3Y-TZP ceramics can exhibit toughness, KIC, and strength, sr, of more than 6 MPa m1/2 and 2 GPa, respectively. However, 3Y-TZP cer-

* To whom all correspondence should be addressed. E-mail address: [email protected] (L. Gremillard)

amics are susceptible to low-temperature degradation (LTD) and slow crack growth (SCG) when exposed to humid atmospheres. SCG corresponds to a crack propagation for stress intensity factors KI below KIC, and is experimentally described by plotting the crack velocity versus stress intensity factor (V–KI diagram). Experimental studies [2, 3] have shown that SCG in zirconia ceramics is a consequence of stress-assisted corrosion by water molecules, as for silica glass, which involves a thermally activated reaction between Zr–O and H–O bonds. A threshold value, KI0, under which no propagation occurs, defines a safety stress intensity factor. Its value is about 50% of the KIC. Recent studies have shown that the whole V–KI diagram is shifted towards high KI values when increasing the grain size [3]. Thus, grain size increases SCG resistance by increasing both KIC and KI0. LTD proceeds by a slow, tetragonal-to-monoclinic

1359-6454/00/$20.00  2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 9 - 6 4 5 4 ( 0 0 ) 0 0 2 5 2 - 4

4648

GREMILLARD et al.: SILICA-DOPED ZIRCONIA CERAMICS

(t→m) transformation at the surface in a moist atmosphere, followed by microcracking and a loss of strength. The main features of LTD are the following [4–7]: 1. transformation is time-dependent—it proceeds at room temperatures at very low rates and the most rapidly at temperatures of 200–300°C; 2. water or water vapour enhances the transformation; and 3. transformation proceeds from the surface to the bulk of zirconia materials by a nucleation and growth mechanism. Transformation of some grains occurs first, followed by propagation to the neighbouring grains by the formation of microcracks and residual stresses.

were ground and polished down to 3 µm with a diamond paste. They were then heated to 1350°C to reveal the grain boundaries. Grain sizes were measured by scanning electron microscopy (SEM; Philips XL 20 instrument), using the linear intercept method. Microanalysis was performed with an energy-dispersive X-ray (EDX) analyser mounted on the scanning electron microscope. Results are presented for samples sintered at 1450°C. This temperature offered several advantages: on the one hand, no zircon was found in any batch (zircon appeared only at temperatures higher than 1550°C) and no quartz grains could be seen; on the other hand, it gave rise to a fine and almost constant grain size of 0.6 µm (as measured by SEM) for all batches. Hence, the only varying microstructural parameter was the amount of dopants. 2.2. Microstructural characterisation

Different mechanisms were proposed to explain LTD in humid atmosphere, but there is no clear experimental evidence for those models. However, the role of the combination of Zr–O with O–H bonds, as well as the presence of internal stresses, are generally emphasised [4–8]. To avoid degradation, the stability of the tetragonal phase can be increased by either decreasing the grain size or increasing the yttria content. However, both solutions reduce the phase transformation toughening, resulting in a reduced crack propagation resistance. Other approaches have been proposed, among them the addition of dopants such as silica and alumina [9– 11]. A limited amount of literature indicates that silica additions should increase the LTD resistance, without providing further explanations [12]. Thus the aim of the present study was to better understand the effect of SiO2 on the microstructure of 3Y-TZP ceramics, and on SCG and LTD, in order to investigate whether LTD resistance can be increased without decreasing mechanical properties. 2. MATERIALS AND METHODS

2.1. Processing A high-purity powder (Tosoh TZ3Y-S, with less than 100 ppm of impurities, and a 3 mol% Y2O3 content) was used as starting material. Three different materials were manufactured by a slip-casting method, all slurries with 80 wt% solids content. Zirconia ceramics with different silica contents were prepared by adding different amounts of colloidal silica (Ludox HS40, Aldrich Chemical Company) to the slurries: 0 wt%, 0.5 wt% and 2.5 wt% in batches T, T-S5 and T-S25 respectively. The different slurries were placed in plaster moulds in order to eliminate water and to form green compacts, which were then dried at 25°C in air for 7 days. Organic compounds were removed by heating at 600°C (heating rate: 15°C/h). Compacts were then sintered at different temperatures for 5 h in air (heating and cooling rate: 300°C/h). Sintered samples

Conventional transmission electron microscopy (CTEM) was performed on a JEOL 200 CX instrument, while high-resolution work, EDX and electron energy-loss spectroscopy (EELS) were performed on a JEOL 2010 F instrument owned by the CLYME (Consortium Lyonnais de Microscopie Electronique). TEM thin foils were prepared by the classic method: 100 µm thick slices were first cut with a diamond wire saw, then ground with a dimpler down to a thickness ranging between 30 and 10 µm, and finally ion-milled with argon ions at 4.5 keV. CTEM observations of the glassy phases were done both in bright field and dark field, using a pseudo “diffuse dark-field method” [13]. Nano-probe analyses were conducted along and across grain boundaries, and in the lattice. Owing to partial overlap of the Si-K, YL and Zr-L peaks, only qualitative analysis of EDX spectra was attempted; extensive line scans across the boundaries and the triple points were then recorded, in order to evidence any possible relative variation of the silicon content. X-ray diffraction (XRD) measurements were conducted on batches T and T-S25 in order to investigate possible lattice distortions, which can be due to either internal stresses, owing to the preparation route, or to the presence of silicon in the lattice. Two methods were employed to analyse the diffractograms. 1. The first method consists in measuring lattice parameters. Diffractograms were recorded with ˚ and Cutwo types of radiation: Cu-Kα at 1.54 A ˚ . Since the Cu-Kβ radiation is monoKβ at 1.39 A chromatic, peak treatment is easier in the latter case owing to the limited convolution of peaks. Peaks were fitted by a pseudo-Voigt function; the Gaussian and Lorentzian widths, the quantity of the Gaussian component, the intensity and position of each peak were calculated. 2. The second method, derived from the Williamson– Hall plot [14], consists in using the equation: bG cos q = e sin q,

(1)

GREMILLARD et al.: SILICA-DOPED ZIRCONIA CERAMICS

4649

where bG and q are respectively the measured Gaussian width and the angle of peaks, and e the strain in the lattice. Plotting (bG cos q) versus (sin q) should give a straight line, the slope of which is the strain e. 2.3. Mechanical properties SCG experiments were conducted with the double torsion (DT) test, the advantage of which is to allow direct visualisation of the crack propagation on the sample surface. The whole V–KI law can then be obtained without any assumption. All samples were polished with diamond pastes down to 3 µm, then heated to 1250°C to remove residual grinding stresses. Samples were completely in the tetragonal state. Two loading configurations were used: constant load and relaxation tests (i.e., constant strain test) allowed low (down to 10⫺12 m/s) and fast (up to 10⫺3 m/s) propagation speeds to be measured, respectively [15, 16]. 2.4. Ageing According to a previous work [7], samples must be polished down to 1 µm so as to reach a very small roughness (Ra of about 3 nm); a heat treatment at 1250°C allows us to ensure that zirconia is in the tetragonal state. Steam sterilisation ageing tests were performed at 134°C and under a pressure of 2 bar. XRD measurements, conducted every 2 h, allowed the LTD to be quantified by measuring the evolution of the monoclinic content at the surface with time. The penetration depth of X-rays within zirconia was estimated to be 5 µm from [17].

Fig. 1. HRTEM micrographs showing the absence of non-crystalline layers at grain boundaries in the zirconia material doped with 2.5 wt% silica (T-S25): (a) “general” grain boundary; (b) coincidence grain boundary; (c) evidence for a glassy pocket at a triple junction.

3. RESULTS

3.1. TEM According to Clarke [18, 19], thermodynamical equilibrium at the sintering temperature should lead to a uniform layer, with a constant thickness (about 1 nm), of glassy phase at grain boundaries. The present high-resolution TEM (HRTEM) investigation clearly shows that no glassy phase is present at grain boundaries (for each batch, about 20 grain boundaries were observed edge-on). Figure 1 summarises the results. In Fig. 1(a), a “general” (i.e., without any specific orientation relationship between the two grains), glass-free grain boundary is shown. Figure 1(b) shows a “special” grain boundary without any intergranular non-crystalline film; such coherent boundaries were frequently observed in all materials. Figure 1(c) shows a fine glassy pocket at a triple junction, which does not extend along adjacent grain boundaries. Figure 2 shows EELS spectra which confirm that the glassy pockets are made of silica, and that no detectable silica is observed along grain boundaries. EDX spectra (not shown here) lead to the same con-

Fig. 2. 1.5 nm probe EELS spectra showing the presence of silicon (Si-K edge) in a silica-rich glassy pocket (lower spectrum) and the absence of this element on a grain boundary (upper spectrum).

clusions: both EELS and EDX line scans failed to reveal any silica (silicon) enrichment at grain boundaries. However, EELS line scans showed a yttrium enrichment near the grain boundaries. This result will be published elsewhere. CTEM observations revealed that a glassy phase

4650

GREMILLARD et al.: SILICA-DOPED ZIRCONIA CERAMICS

Fig. 4. Williamson–Hall plot of XRD results for batches T and T-S25; linear regression of experimental data leads to a lattice distortion parameter e equal to 0.8 and 0.4%, respectively.

racy, which supports the idea that very little (if any) silica is present within the zirconia tetragonal phase. The Williamson–Hall plot (Fig. 4) shows an interesting feature regarding the average degree of distortion of both materials. The slope of the diagram for the compound T is significantly larger than that corresponding to compound T-S25; according to the meaning of that slope (see above), this clearly indicates a stronger distortion of the zirconia lattice in the undoped material.

Fig. 3. “Diffuse dark-field” images showing the general microstructure of (a) T, (b) T-S5 and (c) T-S25 ceramics; note the absence of any glassy pockets in micrograph (a) (the contrast of which is then very poor), contrary to cases (b) and (c).

was almost systematically present in triple points in batches T-S5 and T-S25, but never in batch T (see Fig. 3). An important result is the change in grain shape with the silica addition. Batch T, with no silica, exhibits highly faceted grains; in batch T-S5, and even more in T-S25, the grains are rounded owing to the presence of glassy pockets at triple junctions.

3.3. SCG and LTD The DT results are plotted in a crack propagation speed versus KI diagram for all batches (Fig. 5). The

3.2. XRD The measured lattice parameters for T and T-S25 zirconia are given in Table 1. Both sets differ significantly from that of the pure tetragonal phase: the presence of 3 mol% of yttria (i.e., 6 mol% of yttrium) makes the lattice parameters change from 0.364 to 0.361 nm for the a parameter, and from 0.527 to 0.518 nm for the c parameter. Since the difference between ionic radii is larger for zirconium and silicon than for zirconium and yttrium, one may expect an easily detectable variation of lattice parameters when adding 2.5 wt% of silica (i.e., 6.6 mol%) to zirconia, if silica penetrates homogeneously within the tetragonal ZrO2 lattice. It is clear from Table 1 that T and T-S25 parameters are identical within a good accu-

Fig. 5. V–KI laws for batches T, T-S5 and T-S25 (the different propagation stages—see text for details—are labelled I, II and III).

Table 1. Average lattice parameters of TZP and silica-doped zirconia compared with that of pure tetragonal ZrO2 ˚) Cu-Kβ radiation (l = 1.39 A

˚) Cu-Kα radiation (l = 1.54 A

Sample a (nm) T T-S25 Pure t-ZrO2

0.3612 0.3611

c (nm) 0.5180 0.5185

a (nm) 0.3606 0.3604 0.364

c (nm) 0.5176 0.5177 0.527

GREMILLARD et al.: SILICA-DOPED ZIRCONIA CERAMICS

V–KI law can be divided in three stages, in agreement with previous studies [2, 16]. According to the commonly accepted idea [20], the first propagation stage (labelled I in Fig. 5) is attributed to stress-assisted corrosion kinetics; it is steeper in doped batches and crack velocities are slightly higher. Stage II is controlled by the diffusion of corrosive water molecules [20]; it begins at the same speed for all batches, but is quite more extended for doped zirconia ceramics (it begins at smaller KI) because of the larger slope of their first propagation stage. Stage III occurs at very high crack velocity, and has not been specifically investigated here. No significant differences were found between the three compounds for the KIC and KI0 values. The ageing study of the T and T-S5 batches is illustrated in Fig. 6. The monoclinic content versus time curve for T material shows a typical sigmoidal shape, consistent with a nucleation and growth micromechanism for the tetragonal-to-monoclinic transformation at the surface [7]. Nucleation and growth could also be observed by means of optical interferometry [7]. Transformation at the surface of doped T-S5 zirconia was much more limited, which was only shown by a slight increase of monoclinic content with time. Optical interferometer observations further showed that both nucleation and growth of the monoclinic phase were limited. 4. GENERAL DISCUSSION

The “diffuse dark-field” images allow statistical measurements of the local curvature radius R of grains near triple junctions; more than 60 grains were examined in each batch. Figure 7(a) shows the average value of R plotted versus the silica weight fraction, F. It is interesting to notice that the variation of R with F is compatible with the following relation: R⬀F1/3.

(2)

Assuming that the triple junctions have an ideal geometry (angle of 120° between each set of two

Fig. 6. Ageing of zirconia: monoclinic content versus time in batches T and T-S5.

4651

grain boundaries), one can consider that glassy pockets can be included in tetrahedra of side 2R, where R is found to be the common curvature radius of the grains meeting at the triple point [see illustration in Fig. 7(b)]; the volume of each pocket is then proportional to R3. Since it is expected that the total volume of glassy pockets is proportional to F, relation (2) is meaningful. The most important result concerning the location of the silica-rich glassy phase within silica-doped zirconia is that it is not present along grain boundaries. Such a finding was reported previously by Ikuhara et al. [21], who however showed possible analytical evidence for silica segregation at grain boundaries. Attempts to reveal such silica enrichment at grain boundaries failed in the present study; a fortiori, no detectable silica was found in the central region of the grains. The latter result is in agreement with XRD measurements, which showed that average lattice parameters did not change from the T to the T-S25 material. There is controversy in the literature regarding the presence of continuous intergranular films in SiO2doped zirconia: some authors claim that, according to Clarke’s model [18, 19], a silica phase is homogeneously located in grain boundaries (e.g., [22]) while others defend the contrary (e.g., [21]). The development of an equilibrium non-crystalline layer of constant thickness at grain boundaries must in fact be strongly affected by the thermodynamical conditions (i.e., temperature and cooling rate after sintering). Figure 4 clearly shows that lattice distortions are stronger in the case of undoped zirconia; this proves that internal stresses are accommodated by the presence of silica at triple junctions in the case of the silica-doped compound. These glassy pockets can indeed act to reduce deformation incompatibilities between grains during cooling, but, above all, they decrease the stress concentrations at triple junctions: it was shown in Fig. 1 that grains have angular shapes in undoped zirconia, while they are rounded when silica has been added. It has been shown that low-temperature degradation is much influenced by internal stresses [8] and that the tetragonal-to-monoclinic transformation begins at grain corners (where the stresses are more intense) [23]. Since the grains are rounder, the internal stresses, especially at the grain edges, are lower for silica-doped compounds and transformation is delayed: the resistance to LTD can then be improved without changing either the grain size or the yttria content. As observed during stage I of crack propagation (see Fig. 5), the presence of silica at the triple junctions leads to an increase in crack velocities, since intergranular cracks have to propagate through these glassy pockets. Nevertheless, KI0 and KIC are identical in all batches. KI0 is related to the surface energy of grain

4652

GREMILLARD et al.: SILICA-DOPED ZIRCONIA CERAMICS

Fig. 7. Statistical analysis of the local curvature radius R of grains at triple junctions: (a) plot of R versus the weight fraction of silica (F); (b) two-dimensional sketch of the idealised geometry of a glassy pocket.

boundaries in the presence of water. Since there is no glassy phase at grain boundaries even with an addition of silica, the surface energy can be considered as constant in average whatever the silica content. The same argument can be used for KIC, which is related to the fracture energy under vacuum conditions. The crack resistance is also related to the tetragonal-to-monoclinic transformation (transformation toughening); although this transformation is delayed during LTD in the silica-doped materials, it is expected to occur at the same level of stresses at the crack front, compared with the undoped material. As zirconia ceramics should be used only at KI⬍ KI0 and as KI0 is the same for all batches, it can be inferred that addition of small quantities of silica does not decrease the mechanical properties of zirconia ceramics. 5. CONCLUSION

Zirconia ceramics with the same grain size, but different amounts of silica-rich glassy phase, were prepared. It was shown that silica forms vitreous pockets at triple points, but no continuous film along grain boundaries. The presence of a glassy phase does not degrade the crack resistance, but greatly improves the LTD resistance; this improvement has been explained in terms of accommodation of internal stresses, especially at grain edges. REFERENCES 1. Stevens, R. Zirconia and Zirconia Ceramics, 2nd ed. Magnesium Elektron Ltd, Leeds, UK, 1983.

2. Chevalier, J., Ph.D. Thesis, INSA de Lyon, France, 1996. 3. Chevalier, J., Olagnon, C. and Fantozzi, G., J. Am. Ceram. Soc., 1999, 82, 3129. 4. Yoshimura, M., Noma, T., Kawabata, K. and Somiya, S. J. Mater. Sci. Lett., 1987, 6, 465. 5. Hirano, M. Br. Ceram. Trans. J., 1992, 91, 139. 6. Lilley, E. Ceram. Trans., 1990, 10, 387. 7. Chevalier, J., Cales, B. and Drouin, J. M. J. Am. Ceram. Soc., 1999, 82, 2150. 8. Schmauder, S. and Schubert, H. J. Am. Ceram. Soc., 1986, 69(7), 534. 9. Tsubakino, H., Sonoda, K. and Nozato, R. J. Mater. Sci. Lett., 1993, 12, 196. 10. Gulino, A., Egdell, R. G. and Fragala, I. J. Mater. Chem., 1996, 6(11), 1805. 11. Li, J. F. and Watanabe, R. J. Mater. Sci., 1997, 32, 1149. 12. McCartney, M. L. J. Am. Ceram. Soc., 1987, 70(1), 54. 13. Cinibulk, M. K., Kleebe, H. J. and Ru¨hle, M. J. Am. Ceram. Soc., 1993, 76(2), 426. 14. Warren, B. E. X-Ray Diffraction, Addison-Wesley, New York, 1969. 15. Willams, D. P. and Evans, A. G. J. Test. Eval., 1973, 1(4), 264. 16. Chevalier, J., Saadaoui, M., Olagnon, C. and Fantozzi, G. Ceram. Int., 1996, 22, 171. 17. Emsley, J. The Elements, 2nd ed. Oxford University Press, Oxford, 1991. 18. Clarke, D. R. J. Am. Ceram. Soc., 1987, 70(1), 15. 19. Clarke, D. R. J. Am. Ceram. Soc., 1989, 72(9), 1604. 20. Lawn, B. Fracture of Brittle Solids, 2nd ed. Cambridge University Press, Cambridge, UK, 1993. 21. Ikuhara, Y., Thavorniti, P. and Sakuma, T. Acta. mater., 1997, 45(12), 5275. 22. Kajihara, K., Yoshizawa, Y. and Sakuma, T. Acta metall. mater., 1995, 43(3), 1235. 23. Zhu, W. H., Lei, T. C. and Zhou, Y. J. Mater. Sci., 1993, 28, 6479.