Fracture Toughness of Yttria-Stabilized Cubic Zirconia (8Y-CSZ) Doped with Pure Silica

Transcription

1 Materials Transactions, Vol. 45, No. 12 (2004) pp to 3329 #2004 The Japan Institute of Metals EXPRESS REGULAR ARTICLE Fracture Toughness of Yttria-Stabilied Cubic Zirconia (8Y-CSZ) Doped with Pure Silica Keijiro Hiraga, Koji Morita, Byung-Nam Kim and Yoshio Sakka Materials Engineering Laboratory, National Institute for Materials Science, Ibaraki , Japan In yttria-stabilied cubic irconia, the addition of mass% pure silica introduces a glass phase dispersing uniformly along grainboundary facets and at multiple junctions. For a grain sie of 0.75 or 1.7 mm, the dispersion of the glass phase decreases the elastic modulus, the Vickers hardness and the elastic modulus-to-hardness ratio, whereas it affects little in the value of fracture toughness measured by an indentation method. The latter result arises because the decrease in the elastic modulus-to-hardness ratio is compensated by a decrease in the crack length for a given indentation load. Inspection of crack-propagation paths indicates that the glass phase with sies smaller than those of the matrix grains is not a site for easy crack-propagation, but provides a site for a crack-deflection mechanism. (Received September 2, 2004; Accepted October 18, 2004) Keywords: cubic irconia, silica addition, glass phase, fracture toughness, crack propagation path, crack deflection 1. Introduction In yttria-stabilied cubic irconia (Y-CSZ), which is widely used for solid electrolytes in high-temperature electrochemical devices such as solid oxide fuel cells, 1,2) some benefits can be expected from a small amount of pure silica addition. Recent studies have shown that, while pure silica addition less than 5 mass% does not noticeably affect the total ionic conductivity, 3) it is effective in suppressing grain growth during sintering 4) and enhancing superplasticlike deformation at high temperatures. 5) This result may lead to an opportunity of net-shaping the electrolytes. 3,5) For an application to electrolytes, however, information is necessary on the fracture strength of silica doped Y-CSZ. This is because the solid electrolytes suffer thermal and external stresses and because the fracture strength of Y-CSZ is relatively low 6 9) owing to lack of the transformation toughening 10,11) that works in tetragonal irconia. However, little information has been obtained on the fracture strength of silica-doped Y-CSZ. For the effects of silica addition on mechanical properties, earlier studies have addressed superplasticity, 12 16) low-temperature degradation 17 19) and transformation toughening 17,20) in yttria-stabilied tetragonal irconia (Y-TZP). The present study was undertaken to obtain systematic information on the fracture toughness of silica-doped Y-CSZ. For this purpose, the evolution of microstructure and fracture toughness for a given grain sie was investigated as a function of pure silica addition. 2. Experimental Procedure A high-purity cubic irconia powder containing 8 mol% Y 2 O 3 and 20 mass ppm SiO 2 (TZ-8Y, Tosoh, Japan) was used to prepare materials doped with 0.15, 0.7, 2.0 and 5.0 mass% colloidal silica (SNOWTEX-O, Nissan Chemical, Japan). The raw materials were mixed for 24 h in a polyethylene ball-mill together with high-purity Y-TZP balls (Nikkato, Japan) and ethanol. The nominal impurity contents of the mixture were 700 Na 2 O, 50 > Al 2 O 3 and 50 Fe 2 O 3 in mass ppm. After drying and granulation, the mixture was cold isostatically pressed at 400 MPa. The pressed compacts were densified in air by single-step sintering or by two-step sintering 21) at C for 2 38 h. The combination of sintering temperature and time was adjusted so that the relative density would exceed 99.8% for two different grain sies: smaller than and larger than 1.0 mm. The grain sie, d, was defined as 1.56 times 22) the average area-equivalent circle diameter of grains on polished sections. Undoped 8Y-CSZ was withdrawn from the present examination, since grain growth was significant in both single- and two-step sintering: the grain sie exceeded 5 mm already when the relative density reached 95%. Scanning electron microscopy (SEM, JSM-6500F, JEOL, Japan) and image-analying software (Optimas 6.5, Media Cybernetics, Maryland, USA) were used for microstructural examination on polished surfaces. The surfaces polished to a 1-mm-diamond finish were heat-etched in air and coated with a thin platinum film. Heat etching was conducted at 1175 C for 1 h for the grain-sie measurement. For the examination of a glass phase formed, etching was performed at 1125 C for 1 h in order to suppress the grain-boundary grooving that may be miscounted as the glass phase. After image processing described elsewhere, 23) the measurement was carried out on the glass phase with area-equivalent circle diameters larger than 25 nm on the heat-etched surfaces. Fracture toughness was measured by a Vickers-indentation method 24) for 4-mm-thick specimens polished to a 1-mmdiamond finish and annealed at 1175 C for 1 h. The specimens were indented at a load of P ¼ 98 N for 15 s. The fracture-toughness value was calculated as 24) K IC ¼ 0:016ðE=H v Þ 1=2 ðp=c 3=2 Þ; where E is the elastic modulus, H v is the Vickers hardness and C is half the radial crack length. The value of E was determined from the elastic recovery of Knoop indentations, 25) where the in-surface dimensions of the indentations introduced at 49 N were measured by an optical image analysis 20,23) with a resolution of 0.6 mm. In all samples, the ratio of C to half the diagonal of the indentation exceeded ð1þ

2 Fracture Toughness of Yttria-Stabilied Cubic Zirconia (8Y-CSZ) Doped with Pure Silica 3325 Fig. 1 Microstructural evolution accompanying pure SiO 2 addition: scanning electron micrographs of materials with (a) 0.15, (b) 0.7, (c) 2.0 and (d) 5.0 mass% SiO 2. The average grain sie was 2.4 mm for (a) and 0.75 mm for (b) through (d) The formation of median/radial cracks was confirmed from a linear relationship between P and C 2=3 and on the surfaces cleaved after the test. 3. Results and Discussion 3.1 Sintered microstructure The aimed combination of grain sie and relative density was attained for SiO 2 additions of 0.7, 2.0 and 5.0 mass%: the density measured by the Archimedes method in distilled water reached 99.8% or higher for grain sies of 0:75 0:03 mm and 1:67 0:03 mm. For 0.15 mass% SiO 2, the minimum grain sie obtained was 2.4 mm when the relative density reached 99.8%. X-ray diffraction using Cu-K radiation showed that the matrix of the sintered materials fully consisted of a cubic phase. As shown in Fig. 1, the addition of pure SiO 2 introduced a silicate phase that appears darker than does the irconia grain in SEM. Examination by X-ray diffraction and energy dispersive X-ray spectroscopy confirmed that this phase is amorphous and enriched with Si. The dispersion of the glass phase was uniform. The pockets of the glass phase located mainly at multiple grain junctions and some pockets lay along grain boundaries. The latter became more frequent with an increase in the SiO 2 addition (Figs. 1(c) and (d)). With increasing SiO 2 addition, although the sie and the amount of the glass pocket also increased, the pocket sie remained smaller than those of the matrix grains. Figure 2 shows the quantitative data for the occurrence of the grass phase. The measured volume fraction of the glass phase, V g, is not significantly affected by the exclusion of Fig. 2 Volume fraction of the glass phase as a function of pure SiO 2 addition. The dashed line (V g(cal.) ) is a calculation assuming the saturation of Si ions along grain boundaries. glass pockets smaller than the resolution limit of 25 nm. As an upper bound estimation of this error, we obtain V g 0:2% for d ¼ 0:75 mm using the following equation: 26) V g ¼ 8A g =d 2 ; under an assumption that all triple junctions are filled with SiO 2 having a cross section of A g ¼ð25=2Þ 2 nm 2. ð2þ

3 3326 K. Hiraga, K. Morita, B.-N. Kim and Y. Sakka The measured amount of the glass phase increases almost linearly from 0.30 vol% (0.15 mass% SiO 2 )to12:5 vol% (5 mass% SiO 2 ). For a given SiO 2 addition, there is a trend of a slight increase in the glass phase with increasing grain sie. These data can be explained from the saturation of Si ions at grain boundaries, for a negligibly small solubility of Si ions in irconia grains. 27) Using a saturation value of 25 pmolmm 2 for a high-purity tetragonal irconia (3Y- TZP) 20) and assuming that the solubility of Si is also negligible in the grains of 8Y-CSZ, we estimated the amount of the glass phase that occurs by SiO 2 addition. In this estimation, the grain-boundary area in unite volume is given by stereology as 2N l, where N l is the number density of intersections with grain-boundary lines. As presented with the dashed line in Fig. 2, this estimation reproduces the experimental data. 3.2 Fracture toughness Figure 3 shows the dependence of the elastic modulus, Vickers hardness and fracture toughness on the amount of the glass phase. These properties are insensitive to the variation in the examined grain sie, except that a larger grain sie for a given SiO 2 addition shifts the data points slightly toward a larger V g -region. The V g -dependence is different between the elastic modulus or hardness and the fracture toughness. The elastic modulus and hardness are decreased with the increasing amount of the dispersed glass phase (Fig. 3(a)). This result is compatible with the commonly accepted idea that is based on rules-of-mixtures. Since the values of E and H v for vitreous silica 28) or silicate glasses 24,25) are about 30% and 50% of the values for undoped 8Y-CSZ, respectively, the decreasing E and H v values with increasing V g are reasonable. In Fig. 3(a), the decrease in E is particularly steeper than that expected from the usual rule-of-mixtures (the action-inparallel model 29) ) that gives an upper (Voight) bound. A similar result was reported for 3Y-TZP doped with silicate glasses. 12) There is a possibility that the steep decrease measured by the present method would be associated with the saturation of intergranular Si-segregation and/or an increase in the frequency of irconia/glass-phase interfaces, and further examination is necessary on this point. Another explanation is obtained from the action-in-series model that gives a lower (Reuss) bound. 29) For a given load, a trend of steeply decreasing E appears when the stress carried by the glass phase is the same as that carried by the matrix. The trend is intensified when the former stress is higher than the latter one by a factor of (>1:0), that is, stress concentrations occur in the glass phase. The rule-of-mixtures of this case is represented as E mg ¼ E m E g =fð1 V g ÞE g þ V g E m g; ð3þ where mg, m and g denote the glass-dispersed irconia, the irconia matrix and the glass phase, respectively. As shown with the dashed line of Fig. 3(a), the present data can closely be approximated by assuming E g ¼ 55 GPa without stress concentrations ( ¼ 1:0) or by using a value of vitreous SiO 2 (E g ¼ 73 GPa 28) ) with an assumption of ¼ 1:4. To the glass-phase dispersion, on the other hand, the fracture toughness of 8Y-CSZ is insensitive (Fig. 3(b)). There appears even a trace of increasing toughness with an increase in V g to about 12 vol%. This result is contrary to the common idea in the following points. First, studies on Y- TZP 12,18) have pointed out that the dispersion of a glass phase may provide an easy crack-propagation path and thereby decrease fracture toughness or increase the rate of crack propagation. Second, since K IC ¼ 0:6{0:8 MPam 1=2 and E ¼ 60{80 GPa for silicate glasses 24,25,30,31) are noticeably smaller than K IC ¼ 1:3{2:2 MPam 1=2 and E ¼ 210{220 GPa for undoped 8Y-CSZ, 6 9) the glass-phase dispersion should decrease fracture toughness. The negative effect that should arise from the glass-phase dispersion can be estimated from the following rule-ofmixtures: 32) J IC,mg ¼ð1 V g ÞJ IC,m þ V g J IC,g; ð4þ Fig. 3 Evolution of (a) the elastic modulus, E, and the Vickers hardness, H v and (b) fracture toughness, K IC, as a function of the amount of the glass phase. The dashed lines drawn in (a) and (b) are the calculations of eqs. (3) and (5), respectively. where J IC is the toughness (the energy-release rate). This estimation gives an upper bound of J IC,mg and does not include any additional negative or positive 32) effects that may arise from the mixing of the phases. Using J IC ¼ K 2 IC ð1 vþ=e, where v is Poisson s ratio, and assuming that variation in v due to SiO 2 addition is negligible, eq. (4) is rewritten as K IC,mg fe mg ½ð1 V g ÞK 2 IC,m =E m þ V g K 2 IC,g =E g Šg 1=2 : ð5þ Taking the K IC and E values for 0.15 mass% SiO 2 as those of the matrix and using the values for vitreous SiO 2 (K IC,g ¼ 0:79 MPam 1=2 31) and E g ¼ 73 GPa), we obtain the dashed line of Fig. 3(b). The estimation predicts decreasing fracture toughness with the increasing amount of the glass phase. As presented in Fig. 4, inspection of indentation data show that the glass-phase dispersion causes a decrease both in the elastic modulus-to-hardness ratio, E=H v, and in the crack length, C, for a given indentation load. In eq. (1), a decrease in ðe=h v Þ 1=2 is accordingly compensated with an increase in

4 Fracture Toughness of Yttria-Stabilied Cubic Zirconia (8Y-CSZ) Doped with Pure Silica 3327 certain mechanism that gives a resistance to crack-propagation. Fig. 4 Evolution of the indentation-crack length and the modulus-tohardness ratio as a function of the amount of the glass phase. P=C 3=2, and thereby the value of K IC becomes insensitive to the dispersion of the glass phase. Thus, Figs. 3(b) and 4(a) indicate that the glass-phase dispersion should introduce a 3.3 Crack-propagation behavior As presented in Fig. 5, the radial crack was observed to pass through multiple routes: the irconia grain (), the irconia grain-boundary (), the interface between the irconia grain and the glass phase (/g) and the glass phase (g). Branching of the radial crack and microcracking around the major crack were rarely observed in the present materials. Comparison of the cracks presented in Fig. 5 shows that the frequency of each path is changed with the amount of SiO 2 addition. While transgranular crack propagation is dominant in the material with 0.15 mass% SiO 2 (Fig. 5(a)), for which the glass-phase formation is very limited, the other paths appear more frequently with increasing SiO 2 addition (Figs. 5(b) (d)). The latter is a new aspect in cubic irconia, since the earlier studies 6 9,33) have persistently reported that cracks propagate dominantly through the matrix grains in Y-CSZ and Y-CSZ-base composites. Furthermore, the microscopic shape of the radial crack is also changed with SiO 2 addition. Crack propagation is smooth and/or straight for 0.15 mass% SiO 2, whereas it tends to be jagged along some grain boundaries and irconia/glass-phase interfaces for the increasing SiO 2 addition. To examine the observed aspects more closely, we evaluated the fraction of crack path i, g /g g /g /g g Fig. 5 SEM observation of crack paths in materials doped with (a) 0.15, (b) 0.7, (c) 2.0 and (d) 5.0 mass% SiO 2 for d ¼ 2:4 mm ((a)) and d ¼ 0:75 mm ((b) (d)). In (a), the crack passes through the matrix grains except for two short grain-boundary segments ().

5 3328 K. Hiraga, K. Morita, B.-N. Kim and Y. Sakka Fig. 6 Evolution in the crack paths ((a) and (b)) and the extent of crackjaggedness ((c)) as a function of the amount of the glass phase. The specimens examined are the same as those shown in Fig. 5. F i ¼ L i =L t ; where L i is the total length of crack segments classed to path i and L t is the total crack length examined. We also evaluated the microscopic jaggedness of crack propagation as R ¼ L t =L p ; ð7þ where L p is the total crack length projected in a direction parallel to the macroscopic crack propagation. The value of R is unity for an ideally straight and smooth crack and increasingly exceeds unity with an increasing extent of jaggedness. The examination conducted in a crack-tip region of more than 10 indentations for each SiO 2 addition yielded the results shown in Fig. 6. The examination of more than 10 indentations for each SiO 2 addition yielded the results shown in Fig. 6. The results give the following information on the crack-propagation behavior in SiO 2 -doped 8Y-CSZ. First, while the fraction of crack segments passing through the glass phase, F g, increases almost linearly with the amount of the glass phase, the absolute value of F g is smaller than that of V g (Fig. 6(b)). Hence, the penetration of the radial crack into the glass phase occurs with a frequency lower than that is expected from the random sectioning of the glassdispersed microstructure. This is a confirmation that the glass phase dispersed with sies smaller than those of the matrix grains (Fig. 1) does not act as a site for easy crack propagation. Second, increasing V g results in decreasing F (Fig. 6(a)) and increasing F þ F /g (Fig. 6(b)). Saturation in F appearing for V g 5 vol% can be related to a decrease in the number of irconia grain-boundaries in unit volume. The data as a whole show that, with increasing V g, transgranular crack propagation is increasingly replaced with the propagation through the irconia grain-boundaries and irconia/glassphase interfaces. This result also means that the strength against crack propagation is lower in some grain-boundaries ð6þ and in the interfaces than in the matrix grains. Although detailed studies are desirable on this issue, the present results imply that the lower grain-boundary strength may arise from the intergranular segregation of Si ions. The occurrence of segregation is consistent with Fig. 2, as noted before, and is supported by earlier studies on SiO 2 -doped 8Y-CSZ 3) and (2.5-3)Y-TZP. 14,16,20) The lower strength of the irconia/ glass-phase interfaces may be associated with the elastic modulus and/or thermal expansion mismatch between irconia and the glass phase. Finally, the evolution in the crack paths is accompanied by a change in the microscopic jaggedness of crack propagation. As shown in Fig. 6(c), the extent of jaggedness increases with V g and hence with F þ F /g. As seen in Fig. 5, the jaggedness occurs because the irconia grain-boundaries and irconia/glass-phase interfaces, through which the crack prefers to propagate, are twisted and/or tilted from the direction of macroscopic crack propagation. The extent of jaggedness must accordingly be enhanced with increasing F þ F /g. The second and the last points described above indicate that the irconia grain-boundaries and the irconia/glassphase interfaces can act as sites for a crack-deflection mechanism. Although crack deflection is usually associated with improved fracture strengths in materials containing reinforcements, 29) it works whenever some microstructural factors force the crack front to tilt and/or twist and thereby result in a non-planar crack ) The factors that may cause the non-planar crack are weak grain boundaries 36) or interfaces 29,34,35) and the presence of residual strains due to the elastic and/or thermal expansion mismatch between the matrix and a second phase. 34,35) A lower strength against the crack propagation is just the case in the grain-boundaries and irconia/glass-phase interfaces of SiO 2 -doped 8Y-CSZ. As mentioned before, the elastic and/or thermal expansion mismatch may also contribute to crack deflection through lowering the strength of the irconia/glass-phase interfaces. The present results indicate that the effect of crack-deflection is not so strong as to increase fracture toughness evidently, but enough to compensate the decrease in toughness due to the glass-phase dispersion (Figs. 3(b) and 4(a)). 4. Conclusions (1) In 8Y-CSZ, pure SiO 2 addition introduces a glass phase dispersing uniformly along grain-boundary facets and at multiple junctions. The amount of the introduced glass phase, which reached vol% for mass% SiO 2, can be explained from the saturation of Si ions at the grain boundaries, under a limited solubility of Si ions within the grains. (2) The increasing amount of the glass phase decreases the elastic modulus, Vickers hardness and elastic modulusto-hardness ratio, whereas it causes little change in the value of fracture toughness measured by an indentation method. The latter is due to a decrease in the crack length for a given indentation load; the decreasing crack length compensates the decreasing elastic modulus-tohardness ratio. (3) The glass phase with sies smaller than those of the

6 Fracture Toughness of Yttria-Stabilied Cubic Zirconia (8Y-CSZ) Doped with Pure Silica 3329 matrix grains does not act as a site for easy crackpropagation. Instead, the grain-boundaries and the irconia/glass-phase interfaces of SiO 2 -doped 8Y- CSZ can act as sites for a crack-deflection mechanism. Although the effect of crack-deflection is not so strong as to increase fracture toughness evidently, it is enough to compensate the decrease in toughness due to the glass-phase dispersion. REFERENCES 1) B. C. H. Steele: J. Mater. Sci. 36 (2001) ) N. Q. Minh: J. Am. Ceram. Soc. 76 (1993) ) M. C. Martin and M. L. Mecartney: Solid State Ionics 161 (2003) ) A. A. Sharif, P. H. Imamura, T. E. Mitchell and M. L. Mecartney: Acta Mater. 46 (1998) ) A. A Sharif and M. L. Mecartney: Acta Mater. 51 (2003) ) N.-H. Kwon, C.-H. Kim, H. S. Song and H.-L. Lee: Mater. Sci. Eng. A299 (2001) ) M. Nabbaro, P. Recio, J. R. Jurado and P. Duran: J. Mater. Sci. 30 (1995) ) L. Donel and S. G. Roberts: J. Euro. Ceram. 20 (2000) ) R. A. Cutler, J. R. Reynolds and A. Jones: J. Am. Ceram. Soc. 75 (1992) ) P. M. Kelly and L. R. Francis Rose: Progress in Mater. Sci. 47 (2002) ) O. Vasylkiv, Y. Sakka and V. V. Skorokhod: Mater. Trans. 44 (2003) ) M. Gust, G. Goo, J. Wolfenstine and M. L. Mecartney: J. Am. Ceram. Soc. 76 (1993) (1993). 13) K. Kajihara, Y. Yoshiawa and T. Sakuma: Acta Metall. Mater. 43 (1995) ) Y. Ikuhara, P. Thavorniti and T. Sakuma: Acta Mater. 45 (1997) ) K. Hiraga, H. Y. Yasuda and Y. Sakka: Mater. Sci. Eng. A (1997) ) K. Morita, K. Hiraga and B.-N. Kim: Acta Mater. 52 (2004) ) M. L. Mecartney: J. Am. Ceram. Soc. 70 (1987) ) L. Gremillard, T. Epicier, J. Chevalier and G. Fantoi: Acta Mater. 48 (2000) ) L. Gremillard, J. Chevalier, T. Epicier and G. Fantoi: J. Am. Ceram. Soc. 85 (2002) ) K. Hiraga, K. Morita, T. Uchikoshi and Y. Sakka: submitted to J. Am. Ceram. Soc. 21) I.-W. Chen and X.-H. Wang: Nature 404 (2000) ) M. I. Mendelson: J. Am. Ceram. Soc. 52 (1969) ) K. Hiraga, K. Nakano, T. S. Suuki and Y. Sakka: J. Am. Ceram Soc. 85 (2002) ) G. R. Antis, P. Chanticul, B. R. Lawn and D. B. Marshall: J. Am. Ceram. Soc. 64 (1981) ) D. B. Marshall, T. Noma and A. G. Evans: J. Am. Ceram. Soc. 65 (1982) C-175 C ) R. Raj: J. Am. Ceram. Soc. 64 (1981) ) W. C. Butterman and W. R. Foster: Phase Diagrams for Zirconium and Zirconia Systems, Ed. by H. M. Ondik and H. F. McMurdie, (The American Ceramic Society, Westerville, OH, 1998) pp ) M. Hukuhara and A. Sanpei: Jpn. J. Appl. Phys. 33 (1994) ) K. K. Chawla: Composite Materials, (Springer-New-York, NY, 1998) pp ) E. L. Bourhis and D. Metayer: J. Non-Cryst. Solids 272 (2000) ) J. P. Lucas: Scr. Metall. Mater. 32 (1995) ) M. F. Ashby: Acta Metall Mater. 41 (1993) ) P. Bhargava and B. R. Patterson: J. Am. Ceram. Soc. 80 (1997) ) K. T. Faber and A. G. Evans: Acta Metall. 31 (1983) ) K. T. Faber and A. G. Evans: Acta Metall. 31 (1983) ) B.-N. Kim and T. Kishi: Mater. Sci. Eng. A176 (1994)