Nonuniversality of roughness exponent of quasistatic fracture surfaces

Numerous experiments have indicated that the fracture front (in three dimensions) and crack lines (in two dimensions) in disordered solids and rocklike materials is rough. It has been argued that the roughness exponent ζ is universal. Using extensive simulations of a two-dimensional model, we provide strong evidence that if extended correlations and anisotropy-two features that are prevalent in many materials-are incorporated in the models that are used in the numerical simulation of crack propagation, then ζ will vary considerably with the extent of the correlations and anisotropy. The results are consistent with recent experiments that also indicate deviations of ζ from its supposedly universal value, as well as with the data from rock samples.