Silver-Based Self-Organized Resistive Switching Nanoparticle Networks with Neural-Like Spiking Behavior: Implications for Neuromorphic Computing

High Resolution Image Download MS PowerPoint Slide Brain-inspired computational systems with complex, dynamic behaviors are under active development. Self-organized networks of interacting nano-objects exhibit rich, neural-like spiking behaviors driven by resistive switching. The collective switching dynamics in nanoparticle networks (NPNs) position them as promising candidates for neuromorphic computing. However, the individual switching behavior and structural factors that govern spiking behavior in these networks remain to be fully understood. In this study, the relation between switching behavior and network composition and morphology is explored by the example of three types of silver-based, self-organized percolating NPNs: a monometallic silver (Ag) NPN, an alloy silver–gold (AgAu) NPN, and a composite silver–zirconium nitride (Ag/ZrN) NPN. These are compared with respect to the time scales of switching events (SEs). The SEs comprise a switch-on and a relaxation process. Transmission electron microscopy (TEM) offers valuable insights into the morphological characteristics of the NPNs. Molecular dynamics simulations provide insights into the filamentary processes occurring between Ag nanoparticles (NPs), exploring the dynamics involved in an SE. The distinct differences observed in switching time and kinetics suggest that the composition, morphology, and local environment of the NPs play a critical role in modulating resistive switching behavior, including filament formation and dissolution processes. This materials-based approach enables modification of switch-on and relaxation time scales and patterns in self-organized networks. Such tunability is advantageous for applications where these networks function as physical reservoirs for signal processing and classification (e.g., speech recognition) and in reservoir computing (RC), potentially enhancing adaptive learning capabilities and computational performance.