Title: Shear memory in frictional granular packs

Author (Table Talk): Don Candela, UMass Amherst

Abstract:

We have done DEM simulations which show that a dense pack of frictional grains "remembers" shear strains that are applied to it while it is steadily compressed. The memory is apparently stored in the transverse stresses of the grain-grain contacts that are formed during the compression. When, at an arbitrarily later time, the grain pack is decompressed the memory is recalled in time-reversed order as shear stresses measurable on the pack surfaces. We have found that a pack of 10^4 grains can recognizably recall up to several hundred different input waveforms. The fidelity of the memory is measured by the ability of a simple neural net, trained only on the input waveforms, to accurately classify the recalled signals. For the future we would like to test this effect in physical systems - both grain packs and other contact-forming media such as random fiber nests.