Physical reservoir computing using ferroelectric Hf1−x Zr x O2 devices for efficient temporal data processing: recent progress of materials, devices, and operating modes

作者：Kasidit Toprasertpong, Rikuo Suzuki, Shin-Yi Min, Eishin Nako, Ryosho Nakane, Mitsuru Takenaka, Shinichi Takagi · 发表于：Japanese Journal of Applied Physics · 年份：2025 · DOI：10.35848/1347-4065/adfae1 · 被引用次数：3 · 研究领域：Neural Networks and Reservoir Computing、Ferroelectric and Negative Capacitance Devices、Advanced Memory and Neural Computing

Abstract We review our recent progress in research on reservoir computing using Hf 1− x Zr x O 2 -based ferroelectric devices, focusing on exploring materials, devices, and operating modes. To overcome the difficulty of ferroelectric field-effect transistors (FeFET) which suffer from being in an off state below the threshold voltage, we introduce two effective approaches using multiple devices with complementary operations: a combination of inverting inputs and a combination of n -channel and p -channel FeFETs. With these combinations, reservoir computing with dynamic reservoir states is obtained, resulting in significant performance improvement. We also discuss the utilization of voltage-mode ferroelectric reservoir computing instead of current-mode one, exhibiting more energy-efficient reservoir computing operation. Furthermore, we explore the potential of anti-ferroelectric-like properties which possess rich polarization dynamics regardless of the input history. This is because the polarization states in anti-ferroelectric-like materials can be modulated without having to apply an electric field with the opposite polarity, different from simple ferroelectric materials.