报告主题：Reservoir Computing Based on Oxide Semiconductor/HZO Ferroelectric Devices
主 讲 人：Sungjun Kim 教授（Dongguk University）
主 持 人：周 晔 教授
时 间：2026年7月13日（一）11：00
地 点：致知楼706
嘉宾简介：
Prof. Kim received his B.S. degree from Hanyang University in 2011 and his M.S. and Ph.D. degrees in Electrical and Computer Engineering from Seoul National University in 2013 and 2017, respectively. His research interests include oxide semiconductor devices, resistive random-access memory (RRAM), ferroelectric tunnel junctions (FTJs), ferroelectric field-effect transistors (FeFETs), Flash/2T DRAM technologies, selector devices, and hardware-based AI accelerators. He has extensive expertise in device-array integration and analog computing-in-memory. Prior to joining academia, he worked as a senior researcher in the Next-Generation Memory Team at Samsung Electronics, where he contributed to the process development of phase-change RAM (PRAM) and ovonic threshold switch (OTS) selector devices for future non-volatile memory technologies. He also served as an Assistant Professor at Chungbuk National University before joining Dongguk University.
报告摘要：
Reservoir computing (RC) has emerged as an efficient neuromorphic computing framework for processing temporal information while significantly reducing training complexity. In this seminar, I will present our recent progress on hardware-based reservoir computing based on IGZO/HZO ferroelectric thin-film transistors (FeTFTs). By integrating the volatile photoconductive dynamics of the IGZO channel with the non-volatile ferroelectric memory of HZO, a single device simultaneously performs optical reservoir processing and electrical readout. This architecture enables compact and energy-efficient in-sensor computing without requiring complex recurrent network training. The presentation will introduce hardware-based optical reservoir computing, followed by multi-wavelength reservoir operation using visible-light stimulation to improve reservoir state separability. I will also discuss wide multi-wavelength optical reservoir computing for enhanced temporal feature representation and demonstrate multisensory information processing for image recognition tasks. Finally, I will highlight the scalability and manufacturability of the IGZO/HZO platform based on CMOS-compatible oxide semiconductor and hafnia ferroelectric technologies, emphasizing its potential for future edge AI, neuromorphic vision, and intelligent sensing systems.
