| Popis: |
Random networks offer fertile ground for achieving complexity and criticality, both crucial for an unconventional computing paradigm inspired by biological brains' features. In this work, we focus on characterizing and modeling different electrical transport regimes of self-assemblies of silver nanowires (AgNWs). As percolation plays an essential role in such a scenario, we explore a broad range of areal density coverage. Close-to-percolation realizations (usually used to demonstrate neuromorphic computing capabilities) have large pristine resistance and require an electrical activation. Up to now, highly conductive over-percolated systems (commonly used in electrode fabrication technology) have not been thoroughly considered for hardware-based neuromorphic applications, though biological systems exhibit such an extremely high degree of interconnections. Here, we show that high current densities in over-percolated low-resistance AgNW networks induce a fuse-type process, allowing a switching operation. Such electro-fusing discriminates between weak and robust NW-to-NW links and enhances the role of filamentary junctions. Their reversible resistive switching enable different conductive paths exhibiting linear I-V features. We experimentally study both percolation regimes and propose a model comprising two types of junctions that can describe, through numerical simulations, the overall behavior and observed phenomenology. These findings unveil a potential interplay of functionalities of neuromorphic systems and transparent electrodes. |
