High-precision and linear weight updates by subnanosecond pulses in ferroelectric tunnel junction for neuro-inspired computing

• DOI: 10.1038/s41467-022-28303-x
• Publication Date: 2022-02-04T11:14:57Z
• Authors: Zhen Luo; Zijian Wang; Zeyu Guan; Chao Ma; Letian Zhao; Chuanchuan Liu; Haoyang Sun; He Wang; Yue Lin; Xi Jin; Yuewei Yin; Xiaoguang Li
• Keywords: Tunnel junction

The rapid development of neuro-inspired computing demands synaptic devices with ultrafast speed, low power consumption, and multiple non-volatile states, among other features. Here, a high-performance device is designed established based on Ag/PbZr0.52Ti0.48O3 (PZT, (111)-oriented)/Nb:SrTiO3 ferroelectric tunnel junction (FTJ). advantages (111)-oriented PZT (~1.2 nm) include its switching dynamics, ultrafine domains, small coercive voltage. FTJ shows high-precision (256 8 bits), reproducible (cycle-to-cycle variation, ~2.06%), linear (nonlinearity <1) symmetric weight updates, good endurance >109 cycles an ultralow write energy consumption. In particular, manipulations 150 states are realized under subnanosecond (~630 ps) pulse voltages ≤5 V, the fastest resistance at 300 ps for FTJs achieved by <13 V. Based experimental performance, convolutional neural network simulation achieves high online learning accuracy ~94.7% recognizing fashion product images, close to calculated result ~95.6% floating-point-based software. Interestingly, FTJ-based very robust input image noise, showing potential practical applications. This work represents important improvement in towards building systems.