IGZO‐Based First Spike Timing Tactile Encoders and Coupling‐Enhanced Transistor Synapses for Efficient Spiking Neural Networks

作者：Dan Cai, Jinyong Wang, Tianchen Zhao, Miao Shen, Yunbo Liu, Tieyi Zhang, Fangjie Zhang, Yang Wang, Yadong Jiang, Deen Gu · 发表于：Advanced Science · 年份：2025 · DOI：10.1002/advs.202511168 · 被引用次数：1 · 研究领域：Advanced Memory and Neural Computing、Neural Networks and Reservoir Computing、Ferroelectric and Negative Capacitance Devices

Spike encoding is the fundamental prerequisite for the hardware implementation of event-driven spiking neural networks (SNNs). However, compact device-level realization of first-spike-timing (FST) encoding remains challenging, while high-performance synaptic devices are urgently needed for efficient network training. Here, a light-accelerated SNN hardware framework is proposed that integrates sensing, temporal encoding, and synaptic learning. A PDMS/MWCNTs film with IGZO dual-TFTs (PDTFT) enables precise subthreshold modulation to restore the neuron "resting state," achieving millisecond-scale FST tactile encoding. Meanwhile, a GaOx/IGZO heterojunction introduced as a light-electric coupling synapse (LECTS), where light supplements carriers and electrical bias modulates the barrier, overcoming the intrinsic lack of long-term memory in IGZO and enabling stronger plasticity beyond single stimuli. Combining PDTFT and LECTS, autonomous-vehicle status detection (98.4% accuracy) are demonstrated and smart robotic navigation (98.2% accuracy) with a 90.9% reduction in training time under supervised SNN learning. These results demonstrate a compact and highly-efficient strategy for neuromorphic intelligence systems.