We study artificial sandpiles, using grains composed of ball bearings. By welding several balls together, we can create a variety of grain shapes. We confine the grains between two Plexiglas sheets, separated just enough for one layer of balls; thus we are essentially studying a two-dimensional system. We then rotate the container slowly, so that the pile of balls gets steeper and eventually collapses in an "avalanche." The angle reached just before an avalanche is an indication of the stability of that particular arrangement of grains.
The spread of the avalanche angles for a given collection of grains is more than 15 degrees, an amount clearly visible to the eye. Our current efforts focus on understanding how the precise arrangement of grains relates to the stability: are there structures that give rise to particularly stable or unstable piles? The work involves a great deal of image analysis: writing software to recognize grain shapes and configurations, testing how the position of certain shapes correlates with angle across multiple images, etc.