Experiments on sandboxes provide a useful tool for understanding the deformation of continental crust. To understand the mechanics at the grain to grain level, we idealized the sand as an assembly of steel balls in contact. We performed a carefully constrained suite of experiments in two and three dimensions by stretching various assemblies of steel balls placed on a rubber sheet attached to moveable boundaries. Though dilation occurred in all the experiments, careful observation supplemented by theoretical treatments of both the two- and three-dimensional cases demonstrated that only the balls which started in a close-packed or approximately close-packed hexagonal arrangement deformed by failure along well-defined slip planes. In the two-dimensional close-packed state, preferential dilation gave rise to linear ''ghosts'' over single gaps or complex arches above the rubber sheet. Dropping of either close-packed or approximately close-packed triangular blocks in two or three dimensions created approximately 60¿ faults, triangular horst, and trapezoidal grabens. Under certain conditions we also observed crosscutting faults, faults which died out along strike, and rotating, internally deforming fault blocks. Our experiments duplicate many of the features of sandbox models. We use the predicted behavior of an idealized arrangement of like spheres to explain the mechanical process by which the steel balls deform and compare our results directly with observations on similar experiments on sandboxes.

Though sand grains and steel balls differ in size, unit density, interparticle coefficient of friction, roughness and angularity, we conclude that both the sand and the steel balls dilate as a result of deformation and form triangular fault blocks which move as coherent units. Further, we argue that close packing determines both the appearance and angle of faulting. Sandbox models are used to visually illustrate the deformation of both soft sediments and the much harder continental crust. The implicit assumption is that sand, soft sediments, and continental crust all deform by the same mechanism. We believe that the sandbox models may represent close to true scale models of the upper brittle continental crust if the mechanical process of deformation is the same. In addition, they may provide useful analogues for the deformation of sedimentary sequences. Finally, our model of sand deformation which stresses dilation and pervasive antithetic faulting provides a physical basis to current ''pseudo-continuum'' models for determining amount of extension from fault dips and bed geometries in sedimentary basins. ¿ American Geophysical Union 1992