Volumes of the basins of attraction for mechanically stable disk packings

 Workshop on Topology: Identifying Order in Complex Systems Topic: Volumes of the basins of attraction for mechanically stable disk packings Speaker: Corey O'Hern Affiliation: Yale University Date: Saturday, April 9 Time/Room: 11:30am - 12:30pm/David Rittenhouse Laboratory Room A4, University of Pennsylvania

Experimental and computational model systems composed of frictionless particles in a fixed geometry have a finite number of distinct mechanically stable (MS) packings. The frequency of occurrence for each MS packing is highly variable and depends strongly on the preparation protocol. Despite intense work, it is extremely difficult to predict a priori the MS packing probabilities. We describe a novel computational method for calculating the volume and other geometrical properties of the `basin of attraction' for each MS packing. The basin of attraction for a MS packing contains all initial conditions in configuration space that map to that MS packing using a given preparation protocol. We find that the basin is a highly complex structure. For a compressive-quench-from-zero-density protocol, we show the existence of a small core volume of the basin around each MS packing for which all points map to that MS packing. However, in contrast to previous studies for supercooled liquids, glasses, and over-compressed jammed systems, we find that the MS packing probabilities are very weakly correlated with this core volume. Instead, MS packing probabilities obtained from compression protocols that use initially dilute configurations and do not allow particle overlaps (i.e. those relevant to granular media) are determined by complex geometric features of the basin of attraction that are distant from the MS packing. In particular, we find that the shape of the average basin profile function $S(l)$, which gives the probability for a point on a hyper-spherical shell a distance $l$ from a given MS packing to map back to that packing, can be described by a Gamma-distribution with a peak that increases as the system size increases and as the quench rate decreases. We find a simple model which predicts $S(l)$ for the extreme cases of very slow and fast quench rates.