Reservoir Computing\nSystem with Diverse Input Patterns\nin HfAlO-Based Ferroelectric Memristor

Ferroelectric memristors, particularly those based on\nhafnia, are\ngaining attention as potential candidates for neuromorphic computing.\nThese devices offer advantages over perovskite-based ferroelectric\nmemristors owing to their simpler structures, compatibility with complementary\nmetal-oxide semiconductor technology, and low-power consumption characteristics.\nAdditionally, improvements in ferroelectric memristor’s performance,\nsuch as enhancing tunneling electro resistance (TER) and polarization\nretention, can be achieved using methods like aluminum doping and\ninsulating film deposition. In this study, we implement a physical\nreservoir computing (RC) system utilizing the metal-ferroelectric-insulator–semiconductor-structured\nferroelectric memristor based on Al-doped HfO<sub>2</sub> as an artificial\nsynapse. Specifically, we ensure the universality and diversity of\nthe system by experimentally demonstrating a robust reservoir layer\ncapable of handling various types of input pulses. To utilize the\nferroelectric memristor in the reservoir layer of the RC system, we\nemploy partial polarization switching of ferroelectric materials.\nWe measure the retention loss characteristics of the device for pulse\namplitude, interval, and width, and quantify the time constant values\nby fitting them to a stretched exponential function. Additionally,\nwe validate the suitability of the fabricated device as an artificial\nsynapse by mimicking various short-term plasticity functions of biological\nsynapses. Furthermore, we experimentally demonstrate various applications\nrelated to learning and memory of the brain, such as image training\nand Pavlov’s experiment, utilizing the short-term memory characteristics\nof the fabricated device. Lastly, we evaluate the robustness of the\nRC system under various input conditions by employing the fabricated\ndevice as a reservoir layer.