Emergence of winner-takes-all connectivity paths in random nanowire networks

Neuromorphic systems are circuit interconnects designed to emulate the neural processing of a biological brain. Artificial synapses can be reproduced by controlling emergent complex phenomena observed in self-assembly (network) materials exhibiting non-volatile memory in their contact points. When integrated with the proper control systems, these networks can mimic certain brain-functions, e.g. data recognition, and associative memory. We are developing an innovative computational platform designed to model the neuromorphic properties of smart network devices. We demonstrate that networks made of random nanowires are promising architectures for neuromorphic applications due to their connectivity and neurosynaptic-like behaviours. We observed a self-similar scaling of the conductance of networks and the wire-wire junctions that comprise them. These junctions connect by means of a "winner-takes-all" conducting path that spans the entire network, corresponding to the lowest-energy connectivity path. The memory stored in this conductance state is encoded in specific connectivity pathways, similar to that found in biological neuron systems. These results are expected to have important implications for development of neuromorphic devices and reservoir computing.