High-Speed Nanoscale Ferroelectric Tunnel Junction for Multilevel Memory and Neural Network Computing

Area of Science:

• Materials Science
• Nanotechnology
• Device Physics

Background:

• Ferroelectric tunnel junctions (FTJs) are explored for next-generation nonvolatile data storage and neuromorphic computing.
• Scaling down FTJ devices is crucial for enhancing memory density and computational efficiency.

Purpose of the Study:

• To demonstrate the scaling down capability of ferroelectric tunnel junctions (FTJs).
• To investigate the potential of FTJs for high-density information storage and neural network computing.

Main Methods:

• Fabrication of high-performance 50 nm-diameter Au/Ti/PbZr_0.52Ti_0.48O₃/Nb:SrTiO₃ FTJs.
• Characterization of FTJ resistance states, switching speed, and synaptic plasticity emulation.
• Convolutional neural network (CNN) simulations using experimental FTJ results.

Main Results:

• Achieved FTJs with eight distinct resistance states (3 bits) and a large ON/OFF ratio (>10³), switchable in 10 ns.
• Emulated fundamental synaptic functions like long-term potentiation/depression and spike-timing-dependent plasticity.
• CNN simulations achieved high recognition accuracy (∼93.8%) on fashion images, demonstrating robustness to noise.

Conclusions:

• The developed nanoscale FTJs show great potential for high-density nonvolatile memory applications.
• FTJ-based artificial synapses can effectively emulate biological functions for efficient neural network computing.
• FTJ technology offers a promising pathway for advanced information storage and AI hardware.