講演情報
[17p-S4_202-12]Pad Size Effects on Time Constant and NARMA Performance in Ag/Ag2S Devices
〇(M2C)Kanyarak Klaichit1, Yusuke Nakaoka1, Oradee Srikimkaew2, Alif Syafiq Kamarol Zaman1, Muzhen Xu3, Yuki Usami1,3, Hirofumi Tanaka1,3 (1.LSSE, Kyushu Inst. Tech. (Kyutech), 2.FSRC, School of Science, Walailak Univ., 3.Neumorph Center, Kyutech)
キーワード:
Materials computation、CMOS integration、Low-power consumption
Brain-inspired computing has motivated research on in-material reservoir computing, but the impact of electrode geometry on temporal dynamics remains unclear. This study examines the effect of electrode pad size on the capacitance, time constant, and reservoir performance of Ag/Ag2S nanoparticle network devices. Ag nanoparticles were deposited on Pt/Ti electrodes with pad sizes of 100–1000 µm and a fixed 10 µm gap. Impedance spectroscopy revealed that larger pads increased capacitance and time constants. NARMA tasks showed higher NARMA2 accuracy for smaller pads, while excessively long time constants degraded performance. These results demonstrate that electrode pad size directly controls temporal processing in Ag/Ag2S in-material reservoirs.
