Year: 2024
Type: dissertation
Abstract: The growing interest in artificial intelligence has accelerated research in neuromorphic systems, which aim to replicate the functions of the human brain with the hope of increasing computational speed and reducing power usage. In neuromorphic systems, the development of energy-efficient and reliable synaptic devices and neuron devices is critical as synaptic devices and neuron devices are fundamental building blocks of the neuromorphic systems. One of the strong candidates of the synaptic devices is NAND Flash memory due to its technical maturity and CMOS compatibility. However, NAND Flash memory is disadvantageous in terms of reliability due to the lateral migration of charges to the spacer region. In this dissertation, a novel lateral migration-based synaptic device is proposed by deliberately modifying the characteristics of lateral migration, a trait traditionally viewed as a limitation in the field of memory technologies. By adjusting the charge injection pathway of conventional Flash memory devices, the proposed device diverges from conventional techniques in existing Flash memory and introduces a new approach to neuromorphic computing. Using the novel program/erase method, the proposed synaptic device is operable at a significantly low voltage while maintaining formidable retention and endurance characteristics, which are key requirements of synaptic devices. In addition, with the proposed synaptic device, an efficient hybrid off-chip/on-chip training method, which is robust to device-to-device variations, is presented. The results demonstrate a fully CMOS-compatible synaptic device for reliable and energy-efficient neuromorphic computing. Keywords: Neuromorphic Computing, Synapse, Neuron, Flash memory, NAND Flash, Lateral Migration Student number: 2019-21766 The text and figures of this dissertation, in part or in full, are a reprint of the material as it appears in the following journal: Min-Kyu Park, Joon Hwang, Kyung Min Lee, Sung Yun Woo, Jae-Joon Kim, and Jong-Ho Bae, Advanced Electronic Materials, Jan. 2024, Reprinted with permission, copyright © 2024 Wiley
Author: 박민규
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7020438594
