Ferroelectric tunnel junctions (FTJs) based on ferroelectric switching and quantum tunneling effects with thickness down to a few unit cells have been explored for applications of two-dimensional (2D) electronic devices in data storage and neural networks. As a key performance indicator, the enhanced tunneling electrosistance (TER) ratio provides a broader dynamic range for precise modulation of synaptic weights, improving the stability and accuracy of neural networks. Herein, we report an observation of pronounced enhancement in the TER ratio by over 4 orders of magnitude through the fabrication of large-scale heterostructures combining bismuth ferrite with two-dimensional Ruddlesden-Popper oxide Bi2FeO4. The significant difference in Schottky barrier height between Bi2FeO4 and electrodes leads to a remarkable TER value of 7.8 × 106. Moreover, benefiting from the enhanced conductance contrast between the high resistance state (HRS) and low resistance state (LRS), we demonstrate image recognition and dehaze processing using artificial neural synapses based on these FTJs. These results indicate that giant barrier height modulation can be realized through 2D Ruddlesden-Popper oxides, providing a facile technique for high-density in-memory computing applications.
Zeng et al. (Mon,) studied this question.