High-performance optoelectronic synaptic transistors play a crucial role in developing and emulating artificial visual systems. However, due to the predominant use of single-structure material modulation in optimizing optoelectronic synapses, their energy consumption significantly trails behind that of electronic synapses by several orders of magnitude. Herein, polymer dielectric layers and optimized contact strategies are adopted to realize the ultralow consumption optoelectronic synapses. Integration of polyimide dielectric significantly enhances photogenerated charge carrier dissociation, leading to substantial improvements in photoresponsivity (1.5 × 106 A·W-1), photodetectivity (6.9 × 1012 Jones), and external quantum efficiency (4.0 × 108%). Additionally, optimized contact properties augment their appeal for ultralow energy consumption in optoelectronic synapse applications. Excitatory postsynaptic current is triggered at an incredibly low voltage of 5 μV and boosts an impressively low energy consumption of 0.05 aJ, ranking among the best-reported results in this field. Next, we demonstrate an integrated system combining the MoS2 optoelectronic synapses with a recurrent neural network enabling 100% accurate recognition of optical signals, particularly in scenarios with aJ-leveled energy consumption. Finally, an image encryption system has been developed, in which images are encrypted by photoelectronic conversion of synapse arrays with random voltage settings and decrypted according to the recurrent neural network-based accuracy. More importantly, once partially damaged images are encrypted, through the decryption image inpainting can be realized due to the high accuracy. The proposed innovative approach holds promise for advancing artificial intelligence applications with improved energy efficiency, information security, and computational capabilities.
Read full abstract