Stress-dilatancy in very loose sand

Virtually all investigation of liquefaction has used undrained tests, and it has become common to represent the undrained strength in terms of a collapse surface or collapse stress ratio described by an effective friction angle. A difficulty with undrained tests is that they only allow observation of the interaction of elastic and plastic strain because of the imposed boundary condition (i.e., no drainage or zero volume change), precluding a proper understanding of an effective stress criterion for maximum undrained strength. Drained triaxial tests do not suffer from this shortcoming, and stress-dilatancy of dense sands in drained shear is well established as a fundamental aspect of sand behaviour, based on micromechanical considerations. It is particularly interesting to consider the stress-dilatancy behaviour of very loose sands in the context of soil liquefaction. Although there are some data in the literature on loose sand behaviour in drained triaxial compression, the majority of data are actually for sands markedly denser than sands showing static liquefaction in undrained tests. This paper therefore reports some laboratory testing of very loose sands, together with comparative undrained liquefaction data, and compares the loose behaviour to that of dense sand. These data are reduced to stress-dilatancy form so that the fundamental aspects of loose soil behaviour can be seen and compared to flow rules used in constitutive models. The stress-dilatancy of very loose sand shows no limiting stress ratio markedly less than that of the critical state. Moreover, the stress-dilatancy trends of very loose sand are the same as those of dense sand. There is no evidence of structural collapse of the particulate arrangement of very loose sands, contrary to speculation associated with collapse surfaces in the literature. Explanations of sand liquefaction must seek other physical explanations of the soil behaviour.