[17p-A33-1~12] FS.1 Focused Session "AI Electronics"

2:45 PM - 3:00 PM JST(5:45 AM - 6:00 AM UTC)English Presentation

[17p-A33-5] Reservoir Computing Using Dynamic Polarization and Charge Coupling of Anti-ferroelectric HZO/Si FETs

〇(D)SHINYI MIN¹, Kasidit Toprasertpong¹, Eishin Nako¹, Ryosho Nakane¹, Mitsuru Takenaka¹, Shinichi Takagi¹ (1.The Univ. of Tokyo)

Keywords:

Reservoir Computing,Anti-ferroelectric,FET

Reservoir computing (RC) offers real-time information processing with low computation cost by only training the output layer weights and is promising for edge-AI applications. A reservoir in the RC system needs to convert sequential input data into high-dimensional node states for effective learning. We have demonstrated an Hf_1-xZr_xO₂ (HZO)/Si FeFET-based RC system utilizing the ferroelectric polarization dynamics. However, the insufficient polarization dynamics of FeFETs are still limiting the performance. Meanwhile, the anti-ferroelectric (AFE) properties with double polarization switching can be easily realized by increasing the Zr content in HZO films, which allows us to use complicated response signals. In this study, we propose an effective RC system using HZO/Si AFE-FETs with rich polarization dynamics and polarization-charge coupling.