ScholarWorks@경상국립대학교

초록

Synaptic devices that simulate biological functions are of interest in neuromorphic computing, because of their low power consumption characteristics. However, achieving long-term plasticity (LTP) in electrolyte-gated transistors (EGTs) is challenging, because the electric double layer (EDL) of the electrolyte/channel disappears when the gate electrode voltage is removed. In this study, we fabricated a CNTVT-based EGTs by adjusting the polarity of the backbone. This process involves improving the polarity of the backbone by adjusting the DPP-CNTVT ratio. Furthermore, it facilitates increased binding of TFSI anions in DEME-TFSI at the electrolyte/channel interface. The CNTVT-based EGTs successfully achieved LTP and exhibited essential synaptic properties, including paired-pulse facilitation (PPF) and a high-pass filter. Furthermore, the results of driving MNIST handwritten digits based on long-term potentiation/depression (LTP/LTD) with controlled backbone polarity improved from 50.18% to 93.28%. These findings offer a simple architectural design for synaptic devices that leverage state-of-the-art neural modeling techniques. © 2024 American Chemical Society.

키워드