Scaling and maximum complexity of waveform data stored in a frictional granular medium by shearing during compression

Based on DEM simulations, we reported previously that waveform data can be stored in a packed, frictional granular medium by shearing the sample as it is compressed; when the sample is decompressed the shear waveform applied earlier is recalled as measurable shear stresses (PRL 130, 268202 (2003)). Here we report a more detailed investigation of this memory effect, varying the sample size, amplitude of the compression and shear inputs, shape of the grains, and complexity of the stored waveforms. We find that the recalled signal amplitude is very nearly proportional to the compresssion and very nearly independent of the sample size, suggesting that this ability to store shear waveforms is a bulk property that could exist in various types of medium with random microstructure. The ability of a simple neural net trained independently of the granular medium to properly classify the recalled signals is used to assess the maximum complexity of waveforms that the granular medium can store.