The measurements and the cavity observations are reported in the rst section. The observations are then compared with the predi ons of di fusive cavitation models, to establish the merit of such models for characterizing cavitation in this class of polycrystal. Finally, the models are used to examine the influence of cavi tion on the creep deformation under various conditions of stress and temperature.
RESULTS AND OBSERVATDeformation tests were conducted on exural specimens at constant displacement rates. The plastic strain levels achieved in the were determined from the separation of scribes placed on the side surface of tThe material was a hot pressed Al 2 0 3 /0o25% MgO with an average grain size of rv2]Jm.
T.he Cavity ShapeThe near·~ti p sha ) the upper bound cavity width, w , can be predicted. For the cavity shown in figure 6 the upper bound value of the cavity width is 224 nm. compared th a meas nm. For each case, examined in this way, the measured width was always less than the upper bound value and within a factor of 3. Considering the uncertainty in the diffusion parameter, these correspondences are considered to be sufficiently close to confirm that the observed crack-like cavities had indeed formed by a diffusive mechanism~ with surface diffusion as the ra limiting step.
The TransitionThe transition from the equilibrium to crack-like cavity roorphology is dependent upon the stress level. the relative cavity size and the ratio 6 of the surface to boundary diffusivities 8 • 13 • The location of this transition is plotted in figure 7 for a temperature of l550K (0.15<6<0.5) 3 .The experimentally observed cavities can be placed on the transition diagram to examine conformance with the prediction. This is achieved by adopting implied by the cavity growth models. The observed non-linearity of the creep ra n ~1.8 ·is thus concluded to originate primarily from a non-l'i near c mechanismo The most likely origin of the non-linearity resiconcerned with the vacancy creati nihilation process at boundar-ies, or with impediments to grain boundary sl·idh1g (as , unspecified)16. It is tempting to invoke the non-linearities con ined in dislocation models of grain boundary creep 17 • But. it is di cult to envisa on the basis of the present observations that a sufficient fra ion of boundaries can be described by a discrete dislocation array permit such models to be applied with confidence. This is clearly an issue that meri further investigation.The extent of the deformation attributed to the cavitation can be predicted from models of uniform cavity arrays. The predictions are summa~ zed in figure 8. in terms of the parameter R • the cavitation creep rate relative to the Coble creep rate and f • the ratio of the crack length to crack spacing. For a typical value of I' J (i'J l). figure 8 shows that the proportion of the cavitation creep attributed to the growth of crack-like or full-facet cavities is less than that derived from equilibrium shaped cavities, by a factor of ~3. This prediction is