Universal Neuromorphic Element: NbOx Memristor with Co‐Existing Volatile, Non‐Volatile, and Threshold Switching
Ungbin Byun, Hyesung Na, Sungjun Kim
Abstract
Abstract This study investigates the multifunctional resistive switching behaviors of W/NbO x /Pt devices, highlighting their potential for integrated neuromorphic computing. The device exhibits both volatile and non‐volatile resistive switching characteristics, enabling its dual use in reservoir computing and synaptic weight modulation. Gradual I–V responses in the volatile switching allow for temporal signal processing and short‐term memory functionalities, while abrupt switching in the non‐volatile switching supports long‐term potentiation and depression (PD). The application of the incremental step pulse with verify algorithm (ISPVA) enhances PD linearity and precision of synaptic weight modulation, enabling multi‐level conductance states with improved retention and endurance. Furthermore, by tuning the compliance current, the device exhibits threshold switching behavior, allowing the implementation of leaky integrate and fire (LIF) neuron circuits. These multiple characteristics allow the W/NbO x /Pt device to act as a reservoir layer, readout layer, and spiking neuron, thereby forming a compact and unified platform for on‐chip learning. This work demonstrates the feasibility of using a single resistive device architecture to implement both memory and computation, paving the way for highly integrated and energy‐efficient neuromorphic systems.