Ultralow Power Wearable Organic Ferroelectric Device for Optoelectronic Neuromorphic Computing

Qingxuan Li, Tianyu Wang, Yuqing Fang, Xuemeng Hu, Chengkang Tang, Xiaohan Wu, Hao Zhu, Ji Li, Qingqing Sun, David Wei Zhang,

Nano Letters · 2022 · 114 citations · 45 references

Concepts

TL;DR

In order to imitate brain‑inspired biological information processing systems, various neuromorphic computing devices have been proposed, most of which were prepared on rigid substrates and have energy consumption levels several orders of magnitude higher than those of biological synapses (∼10 fJ per spike). Herein, a new type of wearable organic ferroelectric artificial synapse is proposed, which has two modulation modes (optical and electrical modulation). The device exploits the high photosensitivity of organic semiconductors and the ultrafast polarization switching of ferroelectric materials to achieve a 30 ns operation speed and 0.0675 aJ per synaptic event, and under combined photoelectric modulation it realizes associative learning. The synapse shows good plasticity under varying bending strains, enabling ultralow‑power artificial intelligence systems and future wearable devices.

Abstract

In order to imitate brain-inspired biological information processing systems, various neuromorphic computing devices have been proposed, most of which were prepared on rigid substrates and have energy consumption levels several orders of magnitude higher than those of biological synapses (∼10 fJ per spike). Herein, a new type of wearable organic ferroelectric artificial synapse is proposed, which has two modulation modes (optical and electrical modulation). Because of the high photosensitivity of organic semiconductors and the ultrafast polarization switching of ferroelectric materials, the synaptic device has an ultrafast operation speed of 30 ns and an ultralow power consumption of 0.0675 aJ per synaptic event. Under combined photoelectric modulation, the artificial synapse realizes associative learning. The proposed artificial synapse with ultralow power consumption demonstrates good synaptic plasticity under different bending strains. This provides new avenues for the construction of ultralow power artificial intelligence system and the development of future wearable devices.

References

45