Hybrid WS2-Based Memristor With Tunable Conductance Modulation for Neuromorphic and Nociceptive Learning

Abstract

Here, a high-performance memristive device that integrates a layered WS2 switching medium with a TiOx-rich interface and a BaTiO3 (BTO) dielectric layer is reported. This hybrid structure exploits the defect tunability of WS2 to regulate oxygen vacancy dynamics, while BaTiO3 enhances electric-field stabilization and TiOx acts as a redox-controlling barrier. The device exhibits analog multilevel switching at low voltages (+/- 0.5 V), a wide memory window (>10), stable retention beyond 10(4) s, pulse endurance exceeding 10(5) cycles, and ultralow switching energy (approximate to 54.7 pJ per event). Uniform switching is achieved, with cycle-to-cycle variation of 3.6% and 2.3% for Set and Reset states, respectively. Discrete 5-bit (32-level) resistance states are realized under DC sweeps, enabling high-density memory storage. A broad range of synaptic plasticity features such as long-term potentiation (LTP), long-term depression (LTD), paired-pulse facilitation (PPF), post-tetanic potentiation (PTP), spike-number-dependent plasticity (SADP) and spike-amplitude-dependent plasticity (SADP) - are successfully reproduced. Furthermore, the incremental step pulse with verify algorithm (ISPVA) algorithm enables precise 4-6-bit conductance modulation with enhances linearity, symmetry, and suppress variability. The device also mimicked nociceptor-like behaviors including no adaptation, allodynia, and hyperalgesia. When integrated into an artificial neural network (ANN)ANN, the device achieves a recognition accuracy of 97.4% on the MNIST dataset. These results establish the WS2-based hybrid memristor as a strong candidate for energy-efficient neuromorphic and adaptive sensory applications.