Self-Similar and Neuromorphic Properties in Core-Shell Nanowire Network Systems

Human brains process sensory signals better than any modern computer, which is why we are investigating synthetic circuit networks that mimic neurobiological behavior. Our networks are composed of randomly distributed core-shell nanowires through which electrical signals can propagate. Current flows through their wire cores and can be transmitted from wire to wire through junctions. The junctions offer resistance to the passage of current, but they do not behave as static resistors; they exhibit an adaptable resistance memory (memristance) that can be used to emulate neuron synapses in nanowire networks (NWNs).

Through computational simulations, we will outline how to utilize established emergent properties of NWNs for neuromorphic applications. NWNs are known to synchronize by distributing an input current signal through self-selected paths, optimizing the overall network conductance, and allowing the system to operate at minimum power. We will show how these paths can be manipulated through simulated perturbations that could be feasibly reproduced in a laboratory. Of particular interest is a novel emergent phenomenon in which the network intelligently self-reroutes its current paths without the presence of any imposed drivers.