Abstract

Using Eshelby's energy-momentum tensor, it is shown that the elastic configurational force acting on a moving crack tip does not necessarily point in the direction of crack propagation. A generalization of Griffith's approach that takes into account this fact is proposed to describe dynamic crack propagation in two dimensions. The model leads to a critical velocity below which motion proceeds in a pure opening mode, while above it, it does not. The possible relevance of this instability to recent experimental observations is discussed.