High Precision Conductance Modulation in CuCrP <sub>2</sub> S <sub>6</sub> Synaptic Devices for Enhanced Neuromorphic Computing
Xin Cheng, Zhipeng Zhong, Yezhao Zhuang, Wan Wang, Qianyi Yang, Xiang Li, Wu Shi, Xiangjian Meng, Yanan Cao, Jianlu Wang, Junhao Chu, Hai Huang
Abstract
Abstract Artificial synapses are essential components for realizing neuromorphic computing at the physical level. Although numerous artificial synaptic devices have been fabricated in recent years, their performance is often limited by their resistance state modulation capabilities and stability. Developing artificial synaptic devices with a high number of intermediate states, excellent linearity, and ultralow power consumption remains a challenge. This work presents a neuromorphic synaptic device based on a van der Waals layered ionic conductor material, CuCrP 2 S 6 (CCPS). By precisely controlling the ionic conductivity of the device, it exhibits exceptional biomimetic synaptic behaviors, including long‐term potentiation (LTP) and depression (LTD) with up to 8000 intermediate states (13‐bit), an exceptional nonlinearity of <0.31, and operating energy consumption of <45 pJ per pulse. Importantly, the LTP and LTD behaviors demonstrate outstanding stability, sustaining reliable modulation over 32 cycles. A convolutional neural network (CNN) based on the device's synaptic performance achieves recognition accuracy approaching full precision simulation in image recognition tasks. Additionally, the device shows significant advantages in processing complex auditory signals, achieving a recognition accuracy of 96.4% for sound signals, highlighting its potential in complex sound recognition applications.