Memory loss in a particle swelling model*

Liquid suspensions of particles that are sheared back and forth repetitively will self-organize and ``adapt'' to that shearing so long as its amplitude is below some critical value. This organization creates a memory in the system that can later be ``read out'' by observing how many particles are perturbed by a given shear. We use a model that allows us to efficiently study this behavior. In place of shear, particles swell to a given amplitude, and overlapping particles are repelled. This process repeats while the system is monitored. By training the system on progressively larger memories, the system ``forgets'' memories at smaller amplitudes. We present a study of this ``forgetting'' as the system approaches the critical swelling amplitude.