AL2O3/ALOX Memristor with Nearly Ideal Synaptic Characteristics
- Resource Type
- Conference
- Authors
- He, Qian; Wang, Hailiang; Bai, Yongqing; Hu, Jiayang; Li, Hanxi; Ma, Weiming; Xu, Yang; Zhang, Yishu; Yu, Bin
- Source
- 2024 Conference of Science and Technology for Integrated Circuits (CSTIC) Science and Technology for Integrated Circuits (CSTIC), 2024 Conference of. :1-3 Mar, 2024
- Subject
- Bioengineering
Components, Circuits, Devices and Systems
Engineered Materials, Dielectrics and Plasmas
Engineering Profession
Photonics and Electrooptics
Water
Training
Integrated circuits
Neuromorphic engineering
Neural networks
Memristors
Modulation
- Language
This work demonstrates that a memristor based on bilayer oxides with a stack structure of Pt/Ti/Al 2 O 3 /AlO X /Ti/Pt can perform the nearly ideal synaptic characteristics for energy-efficient neuromorphic computing. The engineered oxygen content promotes the formation of the uniform oxygen vacancy distribution, leading to steadily widening conductive filaments rather than abruptly forming. This provides an effective way to meet the requirement of gradually setting/resetting of synapses, which can achieve multilevel states of conductance and long-term potentiation/depression (LTP/LTD). Training on the MNIST machine-learning dataset demonstrates that this memristor exhibits online learning accuracy exceeding 80%.