Memory and Training in Long-Range Interacting Granular Systems

Granular materials are known to exhibit a memory when driven repeatedly under a fixed strain. Typically, this behavior is observed in disordered systems with short-range interactions, where cyclic driving can produce reversible steady states. We present a quasi-2D disordered granular system with long range interactions mediated by magnetic dipoles sitting on a fluid-air interface. Under externally applied periodic shear cycles, the system eventually settles into a limit cycle where the configuration consistently returns to its original state, indicating a trained state.

The memory behavior we observe in this system has several distinguishing features that make it distinct from other amorphous solids. Interestingly, our system exhibits a memory of a shear amplitude that is slightly higher than the training amplitude. We argue that this shift is a consequence of a variety of factors including the strain rate and the fluid viscosity. Our results therefore uncover a new class of memory phenomena enabled by long-range interactions, offering a rich platform for exploring the physics of memory, training and self-organization in driven disordered systems.