Energy-Efficient Artificial Synapses Based on Oxide Tunnel Junctions

摘要

The development of artificial synapses has enabled the establishment of brain-inspired computing systems, which provides a promising approach for overcoming the inherent limitations of current computer systems. The two-terminal memristors that faithfully mimic the function of biological synapses have intensive prospects in the neural network field. Here, we propose a high-performance artificial synapse based on oxide tunnel junctions with oxygen vacancy migration. Both short-term and long-term plasticities are mimicked in one device. The oxygen vacancy migration through oxide ultrathin films is utilized to manipulate long-term plasticity. Essential synaptic functions, such as paired pulse facilitation, post-tetanic potentiation, as well as spike-timing-dependent plasticity, are successfully implemented in one device by finely modifying the shape of the pre- and postsynaptic spikes. Ultralow femtojoule energy consumption comparable to that of the human brain indicates its potential application in efficient neuromorphic computing. Oxide tunnel junctions proposed in this work provide an alternative approach for realizing energy-efficient brain-like chips.