Hierarchical processing enabled by 2D ferroelectric semiconductor transistor for low-power and high-efficiency AI vision system
Guangcheng Wu,Li Xiang,Wenqiang Wang,Chengdong Yao,Zeyi Yan,Cheng Zhang,Jiaxin Wu,Yong Liu,Biyuan Zheng,Huawei Liu,Chengwei Hu,Xingxia Sun,Chenguang Zhu,Yizhe Wang,Xiong Xiong,Yanqing Wu,Liang Gao,Dong Li,Anlian Pan,Shengman Li
DOI: https://doi.org/10.1016/j.scib.2023.12.027
2024-02-26
Abstract:The growth of data and Internet of Things challenges traditional hardware, which encounters efficiency and power issues owing to separate functional units for sensors, memory, and computation. In this study, we designed an α-phase indium selenide (α-In2Se3) transistor, which is a two-dimensional ferroelectric semiconductor as the channel material, to create artificial optic-neural and electro-neural synapses, enabling cutting-edge processing-in-sensor (PIS) and computing-in-memory (CIM) functionalities. As an optic-neural synapse for low-level sensory processing, the α-In2Se3 transistor exhibits a high photoresponsivity (2855 A/W) and detectivity (2.91 × 1014 Jones), facilitating efficient feature extraction. For high-level processing tasks as an electro-neural synapse, it offers a fast program/erase speed of 40 ns/50 µs and ultralow energy consumption of 0.37 aJ/spike. An AI vision system using α-In2Se3 transistors has been demonstrated. It achieved an impressive recognition accuracy of 92.63% within 12 epochs owing to the synergistic combination of the PIS and CIM functionalities. This study demonstrates the potential of the α-In2Se3 transistor in future vision hardware, enhancing processing, power efficiency, and AI applications.