The fracture of boundaries was investigated in  = 25 bicrystalline 4-point bend specimens which were stressed using constant loads at 550 or 600C. It was concluded that the overall stress level which was applied to the specimens was insufficient to fracture the boundaries. Thus, fracture began only at dislocation pile-ups of sufficient intensity. This was supported by the observation that fracture never occurred before the first dislocations had reached the boundary. As the temperature and the applied stress were increased, the dislocation mobility increased and the time that was required to form a pile-up decreased. It was proposed that dislocation transmission at the grain boundary ahead of the pile-up required both a minimum stress (which was less than that for fracture) and a minimum time (that was suggested to be due to some thermally activated process and a change in slip plane) before the process would start. At high stresses, the pile-up built up too rapidly for transmission to begin before fracture occurred. At low stresses, transmission had time to begin before the pile-up became intense enough to cause fracture. The stress on the boundary was then expected to be the result of 2 processes. One of these was a pile up in the first grain, which increased the stress, and the other was an activated source across the grain boundary, which decreased the stress. The net result was a slower stress build-up at the boundary, and the occurrence of fracture after a delay.

A.Jacques, S.G.Roberts: Materials Science Forum, 1996, 207-209, 153-6