Microstructure, toughness and flexural strength of self-reinforced silicon nitride ceramics doped with yttrium oxide and ytterbium oxide

Authors


Dr Y. S. Zheng, Department of Physics, Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong, China. Tel.: +852 2358 8860; fax: +852 2358 1652; e-mail: phyesha@usk.hk

Abstract

Self-reinforced silicon nitride ceramics with additions of either yttrium oxide or ytterbium oxide have been investigated at room temperature after various processing heat treatments. Devitrification of the intergranular phase in these materials is very sensitive to the heat treatment used during processing and does not necessarily improve their strength and toughness. Hot-pressed ceramics without a subsequent devitrification heat treatment were the strongest. The ytterbium oxide-doped silicon nitride ceramics were consistently tougher, but less strong, than the yttrium oxide-doped silicon nitride ceramics. In all the ceramics examined, the fracture toughness showed evidence for R-curve behaviour. This was most significant in pressureless sintered ytterbium oxide-doped silicon nitride ceramics. A number of toughening mechanisms, including crack deflection, bridging, and fibre-like grain pull-out, were observed during microstructural analysis of the ceramics. In common with other silicon nitride-based ceramics, thin amorphous films were found at the grain boundaries in each of the ceramics examined. Arrays of dislocations left in the elongated silicon nitride grains after processing were found to belong to the {101¯0}<0001> primary slip system.

Introduction

Silicon nitride-based ceramics have many excellent properties: high strength and relatively high fracture toughness, good wear resistance, good oxidation resistance and good corrosion resistance. For some time, they have been under consideration as potential high-performance structural materials because of their superior thermal shock resistance relative to oxide ceramics (Weaver et al., 1975). The use of sufficiently damage-tolerant silicon nitride components in turbine engines in the future could lead to weight savings and an increase in engine operating temperatures, which would in turn increase fuel efficiency.

However, silicon nitride is intrinsically difficult to sinter because of the covalent character of the Si-N bonds and the extremely low self-diffusion coefficients of silicon and nitrogen. Thus, in practice, to achieve nearly complete densification of powder compacts, sintering aids which promote liquid-phase assisted sintering are required for such ceramics (Bonnell et al., 1987; Mitomo & Petzow, 1995; Kleebe et al., 1996). Moreover, it is a material which fails by brittle fracture and in the absence of any special processing procedures it has an unacceptably low fracture toughness, typically about 3 MPa m1/2.

Recent reports on the development of silicon nitride-based ceramics have shown that an increase in the aspect ratio (length/diameter) of silicon nitride grains is able to increase the fracture toughness of silicon nitride to up to 10 MPa m1/2 during conventional sintering (Lee et al., 1997; Becher et al., 1998; Sun et al., 1998). Early work by Lange (1979) showed that the development of elongated grains is related to the αβ phase transformation in silicon nitride. These newly developed ceramics are termed in-situ composites or self-reinforced ceramics. The elongated grains grow in a fine, uniform matrix during sintering, so that the resultant microstructures are similar to those of whisker-reinforced ceramics, in which whiskers are added separately to the silicon nitride powders.

The heavier lanthanide oxides (i.e. the oxides of samarium → ytterbium) have relatively high melting points and are in principle attractive candidate oxides for use as sintering aids, the aim being to provide a highly refractory, viscous, remnant amorphous phase residing at the grain boundaries after processing. Sanders & Mieskowski (1985) have established that grain boundary liquid phases with rare-earth additives usually have a high nitrogen content, a high viscosity and a high glass transition temperature, all of which can improve the high-temperature strength and creep resistance of the ceramics.

Nakayasu et al. (1998) have found from first principles calculations that rare-earth ion additions to silicon nitride are likely to segregate to grain boundaries rather than diffuse into the grains. If the ionic radius of the rare-earth ions is too large, the grain boundaries will be adversely weakened. Ytterbium has the smallest ionic radius of the rare-earth elements and thus is the most suitable candidate rare-earth element to incorporate as a sintering aid on this basis. During sintering, ytterbium oxide reacts with silicon nitride and native surface silica on the silicon nitride powder particles, forming a liquid which promotes both particle rearrangement and solution-reprecipitation processes, and which leads to densification of powder compacts.

In this work, a comparison has been made of two types of self-reinforced silicon nitride, ytterbium oxide-doped and yttrium oxide-doped silicon nitride, the former of which is more expensive to produce, but is the more refractory relative to the ‘benchmark’in-situ yttrium oxide-doped composite. Ceramics were examined as a function of consolidation procedure (either pressureless sintering or uniaxial hot pressing) and a subsequent heat treatment aimed at devitrifying the remnant grain boundary glassy phases. Specific attention has been paid to the microstructural characterization of these ceramics by transmission electron microscopy (TEM) and to their strength and fracture toughness.

Experimental procedure

Materials

The materials used in this investigation were pressureless sintered silicon nitride and hot-pressed silicon nitride. The chemical compositions of the ceramics were: (i) Si3N4 : SiO2 : Y2O3 : Al2O3 in the molar ratios 85 : 8: 4 : 3, and (ii) Si3N4 : SiO2 : Yb2O3 in the molar ratios 85 : 10 : 5. In both cases the Si3N4 in the starting powders was the α-form. In (i), the presence of Y2O3 stimulates the nucleation and growth of crystalline phases in the amorphous grain boundary material during devitrification. Si3N4 ceramics produced with appropriate additions of Y2O3 and Al2O3 densify more readily at lower sintering temperatures than ceramics produced with only one of these two oxide additives, so that small quantities of Al2O3 promote densification of Y2O3-doped Si3N4 significantly at 1550 °C (Knutson-Wedel et al., 1991). Aluminium and oxygen can also be accommodated in the β-Si3N4 crystal structure in appreciable amounts (Oyama, 1972).

Sintering and heat-treatment

Pressureless sintering was carried out in a boron nitride crucible heated by a graphite heating element with an initial constant heating rate of 10 °C min−1 from room temperature to 1000 °C, then to 1750 °C at a heating rate of 5 °C min−1, held at 1750 °C for 2 h under a nitrogen pressure of 0.1 MPa, followed by cooling to 1000 °C at a rate of 10 °C min−1 and finally cooling to room temperature at 40 °C min−1.

Uniaxial hot-pressing was carried out for 60 min at 1650 °C under a pressure of 25 MPa. The hot pressed pellets had final densities of 3.25 g cm−3 and above, as shown in Table 1. The heating rate was 50 °C min−1 from room temperature to 1100 °C, and 30 °C min−1 from 1100 to 1650 °C.

Table 1.  Density, hardness and flexural strength of self-reinforced Si3N4 ceramics under various sintering conditions.

Components

Sintering program

Temp/time
Density
(g/cm3)
Hardness
(GPa)
3-point flexural
strength (MPa)
Y2O3-Si3N4Pressureless sintered1750 °C, 2 h3.1411.9898 ± 45
Pressureless sintered + 1750 °C, 2 h +3.1412.2942 ± 30
 heat-treatment1400 °C, 24 h   
Hot-pressed1650 °C, 1 h3.2515.21161 ± 15
Hot-pressed +1650 °C, 1 h +3.2113.9856 ± 41
 heat-treatment1400 °C, 24 h   
Yb2O3-Si3N4Pressureless sintered1750 °C, 2 h3.238.8633 ± 10
Pressureless sintered +1750 °C, 2 h +3.2410.0810 ± 38
 heat-treatment1400 °C, 24 h   
Hot-pressed1650 °C, 1 h3.3913.9912 ± 28
Hot-pressed +1650 °C, 1 h +3.3914.7794 ± 42
 heat-treatment1400 °C, 24 h   

After densification, half of the samples from both hot-pressed and pressureless sintered systems were heat treated at 1400 °C for 24 h under a nitrogen pressure of 0.1 MPa to crystallise the remnant amorphous material within the ceramics.

Microstructural characterization

The densities of all the samples were determined via immersion in mercury using Archimedes' principle. X-ray diffraction (XRD) with Cu Kα radiation was used to analyse the crystalline phases present. Further microstructural characterization was performed by scanning electron microscopy (SEM) and TEM. TEM foil preparation followed standard ion-beam thinning techniques for ceramic materials, after which a carbon coating was applied to minimize specimen charging under the electron beam. TEM observations were carried out in a JEOL 2000FX at 200 kV and a Hitachi H-9000NA at 300 kV.

Flexural strength

All of the samples tested were pre-machined into flexural bars of 22–24 mm length and 3 × 4 mm2 rectangular cross-section. All four longitudinal faces of the test samples were polished to a mirror finish using diamond abrasives decreasing in size from 15 µm down to 0.5 µm through sizes of 6 µm and 1 µm. To prevent failure caused by edge flaws, all the bars were chamfered prior to bend testing. A detailed study on the effect of the polishing process on the strength of these bars is reported elsewhere (Zheng et al., 2000).

The flexural strengths of the specimens were determined at room temperature using a three-point flexural testing facility. The outer knife edges were 18 mm apart. The cross-head speed of the testing machine was 0.5 mm min−1. Between three and eight specimens were tested for each condition.

Fracture toughness

Fracture toughness data were obtained by both the indentation fracture method (IF) and the indentation strength method (IS). Such techniques can be applied to relatively small ceramic specimens and can help to determine the toughness of materials for cracks of the order of the grain size in these materials. Prior to indentation, the polishing technique described above for the flexural strength data was used for each specimen. In the IF method, Vickers hardness indentations were produced via indentation loads of either (i) 98 N, (ii) 294 N or (iii) 588 N. Crack lengths were measured from the centre of the indents to the crack tip. Between three and eight identical specimens were tested for each condition.

The fracture toughness values were computed from the IF data using the equation proposed by Anstis et al. (1981):

image(1)

and from the IS data using the equation proposed by Dusza & Sajgalik (1995):

image(2)

where E is Young's modulus, H is hardness, c is the measured initial crack length after indentation, but before fracture, σf is the fracture strength of the specimen with an indented crack and P is the indentation load, which was 98 N for all the IS data.

Results

Microstructural characterization

(i) X-ray diffraction analysis

Phase relationships at 1750 °C for the Y2O3– SiO2– Si3N4 and Yb2O3– SiO2– Si3N4 systems are shown in Figs 1(a) and (b), respectively (Gauckler et al., 1980; Hoffmann & Petzow, 1993; Nishimura & Mitomo, 1995), from which it is apparent that there are a number of compounds which can form during processing at high temperature within these two systems.

Figure 1.

(a) Phase diagram of the Y2O3-SiO2-Si3N4 system at 1750 °C (Gauckler et al., 1980; Hoffmann & Petzow, 1993). (b) Phase diagram of the Yb2O3-SiO2-Si3N4 system at 1750 °C (Nishimura & Mitomo, 1995). In both cases axes along the sides of the triangle are in equivalent percentage composition.

X-ray diffractometer traces of the ceramic compacts made in this investigation are shown in Fig. 2. In all cases but one, these show that the α-Si3N4 converts completely to β-Si3N4 during the processing procedure. The exception is the hot pressed yttrium oxide-containing samples: the minor peaks of trace 1 of Fig. 2b, such as the peaks at 2θ values of 20.6° and 31.0°, all index satisfactorily to α-Si3N4 (International Centre for Diffraction Data (ICDD) card 41–0360).

Figure 2.

X-ray diffractometer analyses of yttrium and ytterbium oxide-doped silicon nitride before (trace 1) and after (trace 2) devitrification heat treatments. (a) Yttrium oxide-doped pressureless sintered samples, with prominent silicon nitride peaks labelled in trace 2 in the conventional three-index hkl notation, (b) yttrium oxide-doped hot-pressed samples, (c) ytterbium oxide-doped pressureless sintered samples, and (d) ytterbium oxide-doped hot-pressed samples.

Only the pressureless sintered Y2O3-doped Si3N4 is seen by XRD to be free of secondary crystalline phases (trace 1 of Fig. 2a). All the other processing procedures generate minority secondary crystalline phases. There are clear differences between each of the traces shown in Figs 2(a) and (b), particularly in the 2θ range of 20–50°, which indicate the sensitivity of the devitrification products to the specifics of the processing and post-processing procedures. It is also notable that while the two traces of Fig. 2(c) are similar, the detail of the diffraction peaks of the minority second phases are distinctly different from the traces in Fig. 2(d). An analysis of the X-ray diffractometer traces with ICDD data was able to identify Y10(SiO4)6N2 unambiguously in trace 2 of Fig. 2(a)(ICDD card 30–1462), Yb4Si2N2O7 (ICDD card 31–1455) in traces 1 and 2 of Fig. 2(c) and Yb2Si2O7 (ICDD card 25–1345) in trace 2 of Fig. 2(d). The minority phases in the remaining two X-ray traces, trace 2 of Fig. 2(b) and trace 1 of Fig. 2(d), could not be assigned to a single crystalline phase or to two or more crystalline phases unambiguously.

(ii) Microstructure and morphology

As expected, the processing procedures produced Si3N4 ceramics with elongated grains. Most of these had a relatively high aspect ratio of between 3 and 6, but occasionally exaggerated elongated grains could be found with an aspect ratio of about 10 (Fig. 3). The surface of this particular elongated grain shows distinct evidence of crack growth along the surface of the grain in preference to through the elongated grain. Such a growth mode helps to produce a tortuous crack path, which in turn contributes to crack deflection and bridging, thereby improving the toughness of these ceramics. Two examples taken from the TEM work of the interlocking of the elongated grains are shown in Fig. 4 for pressureless sintered and hot-pressed yttrium oxide-doped samples. Such interlocking is also beneficial for improving fracture toughness.

Figure 3.

SEM image of a fracture surface from a ytterbium oxide-doped pressureless sintered sample (without a devitrification heat treatment) showing elongated grains lying in the fracture surface. The crack has grown along the surface of the elongated grains rather than through these grains, causing crack deflection and crack bridging and helping to increase the fracture toughness of the material.

Figure 4.

TEM bright-field images of (a) yttrium oxide-doped pressureless sintered silicon nitride and (b) yttrium oxide-doped hot-pressed silicon nitride showing impingement of elongated grains. These constitute stiff obstacles to crack growth fronts. In neither of these two examples were the specimens heat treated after processing to induce devitrification of the grain boundary amorphous material.

The lack of chemical analysis facilities on the TEMs used meant that information could not be acquired on the chemical compositions at and near the triple junctions and the grain boundaries. However, examination of grain boundaries by bright-field TEM in both yttrium-doped and ytterbium-doped material consistently showed dark contrast at the grain boundaries and dark particles at triple junctions, indicative of the presence of high atomic number elements such as yttrium and ytterbium (Fig. 5), and in line with the expected segregation to grain boundaries and triple junctions of such species (Nakayasu et al., 1998).

Figure 5.

(a) and (b): TEM bright-field images of ytterbium oxide-doped hot-pressed samples. The presence of ytterbium between the silicon nitride grains can be inferred from the dark particles and the dark contrast at the grain boundaries (arrowed).

(iii) Dislocations in βSi3N4 grains

At low magnification, dislocations were frequently observed in the centre of elongated grains, such as in one in Fig. 6 extending from the top left-hand side of the micrograph into the centre of the micrograph. Here the dislocations will have been generated at high temperatures as a consequence of stresses generated during the growth of adjacent elongated grains into these grains, such as the two elongated grains apparent in this figure which impinge on this grain. One particular elongated β-Si3N4 grain with an aspect ratio of 5–6 shown in the middle of Fig. 7a was chosen for a quantitative analysis of these dislocations. The dislocations within the dotted lines in the bright-field image shown in Fig. 7b were studied. Apart from dislocations labelled A, B and C in Figs 7(c) and (d), these dislocations exhibited only faint residual contrast for g = 21¯30 (Fig. 7c) and g = 112¯0 (Fig. 7d), implying from the g.b = 0 invisibility criterion that their Burgers vectors were parallel to [0001].

Figure 6.

Bright-field TEM image from an yttrium oxide-doped hot-pressed sample showing a high density of dislocations in the centre of an elongated grain arising from the impingement of an adjacent elongated grain.

Figure 7.

TEM images from an ytterbium oxide-doped pressureless sintered sample: (a) bright-field image showing the elongated grain in the centre of the image chosen for further study; (b) at higher magnification, an image of the dislocations chosen for further study within the dotted lines indicated, taken with g = 101¯1. (c) and (d) are two beam conditions taken under g = 213¯0 and g = 112¯0, respectively, where most of the dislocations in the boxed area apart from A, B and C can be seen to be invisible.

Further dark-field images shown in Fig. 8 confirmed that A, B and C were different in character from the longer lines of dislocations with Burgers vector parallel to [0001]. A and C are nests of dislocation loops and B is a short length of dislocation line pinned at the top and bottom surfaces of the thin foil, with a distinctly different line direction from the dominant, longer, dislocation lines.

Figure 8.

(a) and (b) Further dark-field images (not two beam) from the grain in Fig. 7 with dominant g-vectors of g = 202¯3 and g = 224¯3, respectively, showing the way in which A, B and C are different in character from the majority long lengths of dislocation lines.

Trace analysis of the longer lines of dislocations shown in Fig. 7(b) and the set of long dislocation lines in Fig. 9, each of which also had Burgers vector parallel to [0001], showed that in both cases the slip planes were the {1¯010} prism planes. [Note that in Fig. 9 neither the bright-field image in Fig. 9(a) nor the dark-field image in Fig. 9(b) are taken in two beam imaging conditions, which is why the dislocations are visible.]

Figure 9.

Further examples of the long lengths of dislocation lines seen in an ytterbium oxide-doped pressureless sintered sample at two different zones: (a) a bright-field image taken with the electron beam parallel to [1¯101] with g = 112¯0 and (b) a dark-field image taken with the electron beam parallel to [011¯1] with g = 21¯1¯0. Trace analysis confirmed that the dislocations lie in {101¯0} prism planes.

(iv) Amorphous films

High resolution TEM showed that, as expected, the samples contained amorphous films separating the grains, particularly at triple junctions, such as in the example shown in Fig. 10(a) from a pressureless sintered ytterbium oxide-containing sample. Amorphous films extending along grain boundaries were also evident (e.g. Fig. 10b), but these were very thin – the example shown in Fig. 10(b) from the same pressureless sintered ytterbium oxide-containing sample appears to be ∼ 7 Å in thickness.

Figure 10.

(a) High resolution transmission electron microscope image of the amorphous phase found at a triple junction in a ytterbium oxide-doped pressureless sintered sample, from which a thin film extends along the adjacent silicon nitride grain boundary. The lattice fringes evident in the lower silicon nitride grain are from (112¯0) planes and the lattice fringes seen in the upper right-hand side grain are from (101¯0) planes. (b) A further example of a silicon nitride grain boundary in a ytterbium oxide-doped pressureless sintered sample showing evidence for a thin amorphous layer. The lattice fringes in the upper silicon nitride grain are from (112¯0) planes and the lattice fringes in the lower grain are from (101¯0) planes.

(v) Mechanical properties

Room temperature three-point bend strengths and hardnesses of the ceramics are shown in Table 1. In each of the four different heat treatment schedules the ytterbium oxide-doped ceramics had lower strengths than the ceramics containing both yttrium oxide and alumina. It was significant that the devitrification heat-treatment improved still further the strengths of the weaker pressureless sintered material, but weakened the stronger hot-pressed material.

The fracture toughness data presented in Table 2 show that the fracture toughnesses calculated from the indentation fracture method increase as the indentation load used to initiate cracks increased. The effect of increasing the indentation load from 98 N to 588 N was to cause crack lengths measured in the SEM to increase from ∼ 200 µm to ∼ 900 µm. The set of samples found to be most sensitive to indentation load were the pressureless-sintered ytterbium oxide-doped ceramics, in which the measured value of fracture toughness rose from 6.3 to 9.1 MPa m1/2 as the indentation load rose from 98 N to 588 N. Values of fracture toughness calculated from the indentation strength method for the same indentation load used to generate crack lengths for the indentation fracture method are consistently higher, but since indentation techniques are necessarily a simple way of estimating absolute toughness (Lawn, 1993), the agreement between the results from the IF and IS techniques is satisfactory.

Table 2.  Values of KIC and the parameter m indicating the degree of R-curve behaviour in the various self-reinforced Si3N4 ceramics.


Components


Sintering program


m
KIC
(IF-98N)
(MPa m1/2)
KIC
(IF-294N)
(MPa m1/2)
KIC
(IF-588N)
(MPa m1/2)
KIC
(IS-98N)
(MPa m1/2)
  1. In determining the KIC values from Eqs (1) and (2),E has been taken to be 300 GPa for silicon nitride.

Y2O3-Si3N4Pressureless sintered0.1267.17.58.58.6
Pressureless sintered +
 heat-treatment
0.2016.87.98.77.8
Hot-pressed0.1027.17.98.18.3
Hot-pressed +
 heat-treatment
0.1726.87.48.57.8
Yb2O3-Si3N4Pressureless sintered0.3326.37.99.17.9
Pressureless sintered+
 heat-treatment
0.2156.88.38.99.6
Hot-pressed0.0957.47.88.57.5
Hot-pressed +
 heat-treatment
0.1786.87.38.47.7

The relationship between flexural strength and the fracture toughness values is shown in Fig. 11. It is notable that there is no clear trend between fracture toughness and flexural strength. For example, while there is an increase in fracture toughness with flexural strength for the IF 98 N data, the reverse is true for the IF 588 N data, and neither the IF 294 N data nor the IS 98 N data show any dependency of fracture toughness on flexural strength. This lack of overall correlation is not surprising: three point flexural strength is determined by a combination of the distribution of flaw sizes and the stress distribution introduced by the three-point bend test, while fracture toughness is a measure of the resistance of a material to crack propagation during brittle fracture.

Figure 11.

Relationship between flexural strength and fracture toughness for the yttrium oxide-doped and ytterbium oxide-doped silicon nitride ceramics.

The sensitivity of the fracture toughness to crack extension (R-curve behaviour) has long been recognized in the literature and arises from crack shielding processes (Lawn, 1993). For materials such as soda-lime glass, the fracture resistance is not dependent on the flaw size, so that the value of fracture toughness calculated using Eq. (1) is independent of the load P applied, because the flaw size, c, of the radial crack generated is such that P/c3/2 remains a constant as P increases. However, it is apparent from Table 2 that, for these self-reinforced silicon nitride ceramics, the calculated KIC values from the IF method consistently increase as P increases.

A consideration by Krause (1988) of the rising fracture toughness seen from the bending strength of indented alumina bars led him to propose a power-law representation of the fracture resistance, Kr, of the material:

image(3)

where Δc is the crack extension from a pre-existing, traction-free, notch. For indentation experiments, the magnitude of any pre-existing, traction-free, notch is necessarily small in comparison to the size of crack introduced and can sensibly be taken to be zero. The exponent m is zero when the material does not exhibit R-curve behaviour and is higher the more a material exhibits R-curve behaviour.

The IF data in Table 2 can be fitted conveniently to the form of Eq. (3) by plotting ln KIC against ln 2c and estimating the gradient m of the data, such as in the examples shown in Fig. 12. Values of m for each ceramic investigated are given in Table 2. The value of m is highest for the pressureless sintered ytterbium oxide-doped material, but it is noteworthy that apart from the pressureless sintered samples, values of m in the yttrium-doped and ytterbium-doped samples are similar for similar heat treatments.

Figure 12.

Plots of ln KIC (with KIC in MPa m1/2) against ln c (with c in µm) for pressureless sintered yttrium oxide-doped and ytterbium oxide-doped silicon nitride ceramics.

Discussion and conclusions

It is apparent from the microstructural work that both the yttrium-doped and ytterbium-doped silicon nitride samples have elongated grains present. The observations of dislocations with Burgers vectors parallel to [0001] on {1¯010} prism planes is consistent with the expected primary slip system operating during plastic deformation processes in silicon nitride (Evans & Sharp, 1971; Kawahara et al., 2000), and can be rationalized in terms of the stresses introduced during the impingement of adjacent elongated silicon nitride grains during sintering.

Values of fracture toughness for ceramic materials quoted in the literature tend to be sensitive to the measurement technique used, irrespective of whether or not the materials exhibit R-curve behaviour. For example, Mukhopadhyay et al. (1999) examined four different procedures for estimating the fracture toughness of sintered silicon nitride samples and found that the value of KIC obtained is sensitive to the measurement technique and to the experimental parameters of a given technique. They found that the chevron notched beam method gives a greater value of KIC than the single edge notched beam (SENB) method for a given blade width and a constant loading rate. The value of KIC from the SENB method was about 1.7 times higher than that deduced from IF, whereas fractographic methods produced values which compared favourably with the SENB in at least 80% of cases. In addition, KIC values deduced from the SENB increased with the notch tip radius.

Here, we have chosen to used the IF and IS methods because of their ease of use. As Lawn (1993) has noted, both methods are limited when attempting to determine absolute values of toughness. We can therefore extrapolate from the work of Mukhopadhyay et al. (1999) that absolute values of fracture toughnesses for both sets of dopants are likely be in excess of 10 MPa m1/2. The marginal increase in fracture toughness of the ytterbium-doped material relative to the yttrium-doped material is offset slightly by a decrease in strength, the latter of which is determined by the sizes of the largest flaws in the materials and their positions in three point bend tests relative to the point of highest stress. Devitrification clearly has the effect of increasing flaw size in the denser hot pressed samples examined in this study. The microstructural observations of elongated grains in all sets of samples is encouraging, providing evidence for the relevant toughening processes such as crack deflection and crack bridging, and it is clear that there is scope for further optimization of starting compositions and heat treatments based on ytterbium doping of silicon nitride to produce self-reinforced silicon nitride ceramics which retain high values of fracture toughness without compromising room temperature strength.

Acknowledgements

This work was supported by UIMC-Research Unit of Ceramic Materials, Praxis Contract no. 53, and by the Praxis XII program A Foundation for Science and Technology, Lisbon, Portugal under Praxis XXI/BPD/18876/98 and research contract P3/3.1/MMA1777/95, and by a Royal Society Sino-British Fellowship Trust Award.

Ancillary