| Abstract: | This paper reviews the literature on the mechanisms by which cracks are nucleated in polycrystalline ice. In the absence of preexisting cracks, crack nucleation is the first step in mechanical failure. A variety of mechanisms have been discussed, mostly involving the propagation of microcracks or precursors from initial sizes below the level of convenient detection to easily observable cracks. True crack nucleation without preexisting precursors requires that stresses be locally concentrated to levels matching the theoretical cleavage or cohesive strength. Candidate mechanisms for this concentration of stress include: dislocation glide (leading to pile-ups of dislocations on particular slip planes); grain-boundary-sliding (leading to stress concentrations at the triple junctions at the edge of grain-boundary facets); thermal expansion of extraneous inclusions (such as produced by progressive freezing of brine pockets upon cooling); and elastic anisotropy of the ice crystals. Both dislocation glide and boundary sliding are kinetic processes in which the stress redistribution occurs at a finite, temperature-dependent rate. These processes also contribute temperature-dependent internal friction and anelasticity and their operation can therefore be independently measured. Elastic anisotropy leads to stress concentrations in an athermal manner and therefore becomes more important at very low temperatures and high loading rates. It has been shown, however, that in the absence of stresses due to brine pockets, the inherent elastic anisotropy of ice is not sufficient to nucleate cracks in purely two-dimensional models, such as perfectly columnar grains in plane strain. Calculated nucleation energies are far beyond available thermal activation energies and some additional stress concentrating effects are required. Irregularities in the third dimension (e.g. jogs in column boundaries) have not been completely investigated. Although inclusions (e.g. brine pockets or particulate inclusions) may sometimes be important, we should not neglect the kinetic processes mentioned above. Various experiments on fresh-water ice have provided evidence that crack nucleation is usually associated with grain boundaries in a manner consistent with grain-boundary sliding and the associated stress concentration fields. |
