higher oxygen pressures was also discontinuous (on the sub-micron scale), but the time intervals were much shorter, and the crack remained essentially sharp. Similar crack-arrest marks have been seen in all cases of dynamic embrittlement in polycrystalline specimens, but the cracking of the Cu-Sn bicrystals was continuous [10]. Our interpretation [17] of the fracture-surface observations was as follows: Cracking in polycrystals at sufficiently high surface concentration of the embrittling element occurs by the advance of sharp intergranular cracks caused by oxygen-induced decohesion, but at different rates along the crack front, depending on the local grain-boundary diffusivity. At any instant the main crack is advancing in some places, but not in others, and the load is therefore carried by the non-cracking regions. This shields the cracking regions from stress and allows the existence of sharp crack fronts. However, at any region of cracking the local stress relaxes as the crack moves forward, and the crack then has to stop and wait for creep of the non-cracking regions to raise the stress again. Thus, the cracking of a polycrystal is discontinuous because of the constraint imposed by non-cracking regions at any moment in time. The result is that large stress intensities are borne by the specimen (Fig. 3), but this does not reflect the stress intensity at the points of sharp-crack advance. In addition, the temperature dependence of cracking is found [17] to be consistent with that of self-diffusion in the alloy, rather than with the intergranular diffusion of the embrittling element in from the surface, because the cracking is constrained by the rate of power-law creep in the noncracking regions. Continuing this interpretation, in the case of cracking at very low oxygen pressures, oxygen atoms arrive at the crack tip so slowly that they have time to diffuse away along the grain boundary, down the gradient of chemical potential provided by the local stress. The crack then has time to blunt, and the oxygen collects in the region of maximum stress ahead of the blunt crack. When the concentration becomes high enough, cracking occurs back to the crack tip, and the process is repeated. Oxygen-induced cracking in nickel-base alloys has been studied mainly in the context of cyclic loading and the effects of hold-time and loading rate on intergranular fatigue-crack growth [16,18]. The results obtained under fixed displacement conditions [17] indicate that the cyclic nature of the loading in the previous studies is incidental to the crack-growth process. Experiments to study cracking in liquid metals using the same fixed-displacement, loadrelaxation approach as outlined above have been only partly successful. It has been found with mercury and several copper-base alloys that the cracking is either too rapid to control or is so slow that it has to be driven with a moving cross head on the testing machine [19,20]. Much more work is needed in this area. SUMMARY It is now evident that dynamic embrittlement is a generic form of brittle fracture that involves decohesion caused by the inward diffusion of surface-adsorbed embrittling
RkJQdWJsaXNoZXIy MjM0NDE=