Unraveling Stochastic Dynamics and Switching Mechanism in Ag Network-Based Neuromorphic Device by Impedance Spectroscopy

Neuromorphic devices are driving advancements in brain-inspired computing, and in this study, impedance spectroscopy was employed on an in-plane volatile neuromorphic device with self-formed silver (Ag) structures to gain deeper insights into its conduction mechanism. The Ag islands function as synaptic junctions, while smaller Ag nanoparticles act as signal transmission channels, effectively mimicking a neural network. Electrical stimulation emulates synaptic functionalities, and the equivalent RC circuit model confirms the structural similarity with biological synapses. Impedance analysis throughout the switching process reveals how resistive, capacitive, and diffusion components contribute to device behavior, with the resistance dropping by nearly six orders of magnitude as the device transitions from the high resistance (HRS) to the low resistance state (LRS). Interestingly, the capacitance remains steady in the picofarad (pF) range, and its low dielectric constant and negligible variation during switching strongly indicate a metallic filamentary conduction mechanism. The sporadic switching between HRS and LRS under AC and DC fields highlights the device’s volatile nature, which is promising for real-time dynamic neural networks and reservoir computing. Humidity-dependent impedance spectra further reveal significant diffusion contributions at high relative humidity (RH), and the distribution of relaxation times (DRT) analysis provides valuable electrochemical insights. The time constant for diffusion, derived from DRT, along with X-ray photoelectron spectroscopy (XPS) findings, supports the proposed mechanism. These results underline the importance of controlled atmospheres and effective encapsulation strategies for consistent device performance. Altogether, the successful equivalence of the device’s electrical behavior with that of a biological synapse marks a significant step forward in understanding and designing next-generation neuromorphic devices.

Schematic showing the device equivalence with the biological synapse. Impedance spectroscopic studies show the transition from the high resistance (HRS) to the low resistance state (LRS) with increasing voltage (V_DC)

Small 21, no. 35 (2025): 21, 2502771, DOI: https://doi.org/10.1002/smll.202502771