Leaky 2T Dynamic\nRandom-Access Memory Devices Based\non Nanometer-Thick Indium–Gallium−Zinc-Oxide Films for\nReservoir Computing

This paper explores the integration of indium–gallium–zinc\noxide (IGZO)-based 2-transistor 0-capacitor dynamic random-access\nmemory (2T0C DRAM, or shortly, 2T DRAM) into reservoir computing for\nadvanced semiconductor artificial intelligence (AI) applications.\nThe short-term memory characteristics of IGZO 2T DRAM enable rapid\nread–write speeds essential for processing time-varying input\ndata. Experimental results confirm high on/off ratios and leaky retention\nbehaviors. The study also examines paired-pulse facilitation (PPF)\nphenomena, offering insights into reinforcement mechanisms for cognitive\ncomputing. Finally, the reservoir computing approach achieves notable\npattern recognition accuracy with a 4-bit pulse scheme, showcasing\nits effectiveness in complex data sets.