Breakdown Lifetime Analysis of HfO2-based Ferroelectric Tunnel Junction (FTJ) Memory for In-Memory Reinforcement Learning

We clarified breakdown mechanisms of HfO 2 -based ferroelectric tunnel junction (FTJ) memory via systematic time-dependent dielectric breakdown (TDDB) measurement for realization of reliable in-memory reinforcement learning (RL) system. The defect generation in the interfacial layer SiO 2 determines...

Full description

Saved in:
Bibliographic Details
Published in2020 IEEE International Reliability Physics Symposium (IRPS) pp. 1 - 6
Main Authors Yamaguchi, Marina, Fujii, Shosuke, Ota, Kensuke, Saitoh, Masumi
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.04.2020
Subjects
cross-point
Dielectric breakdown
Dielectric measurement
Endurance
Ferroelectric HfO 2
FTJ
Hafnium compounds
In-memory reinforcement learning
Reinforcement learning
Reliability
Silicon compounds
Systematics
Online AccessGet full text

Cover

More Information
Summary:We clarified breakdown mechanisms of HfO 2 -based ferroelectric tunnel junction (FTJ) memory via systematic time-dependent dielectric breakdown (TDDB) measurement for realization of reliable in-memory reinforcement learning (RL) system. The defect generation in the interfacial layer SiO 2 determines the device breakdown. By detailed TDDB analysis, we found that the lifetime of FTJ with ~ 30nm in diameter satisfies endurance requirement for practical RL array operation. As a result, HfO 2 FTJ has high potential to achieve reliable cross-point RL system.
ISSN:1938-1891
DOI:10.1109/IRPS45951.2020.9129314