High-performance asymmetric electrode structured light-stimulated synaptic transistor for artificial neural networks

Photonics neuromorphic computing shows great prospects due to the advantages of low latency, low power consumption and high bandwidth. Transistors with asymmetric electrode structures are receiving increasing attention due to their low power consumption, high optical response, and simple preparation technology. However, intelligent optical synapses constructed by asymmetric electrodes are still lacking systematic research and mechanism analysis. Herein, we present an asymmetric electrode structure of the light-stimulated synaptic transistor (As-LSST) with a bulk heterojunction as the semiconductor layer. The As-LSST exhibits superior electrical properties, photosensitivity and multiple biological synaptic functions, including excitatory postsynaptic currents, paired-pulse facilitation, and long-term memory. Benefitting from the asymmetric electrode configuration, the devices can operate under a very low drain voltage of 1 x 10(-7) V, and achieve an ultra-low energy consumption of 2.14 x 10(-18) J per light stimulus event. Subsequently, As-LSST implemented the optical logic function and associative learning. Utilizing As-LSST, an artificial neural network (ANN) with ultra-high recognition rate (over 97.5%) of handwritten numbers was constructed. This work presents an easily-accessible concept for future neuromorphic computing and intelligent electronic devices.

Automated summary

Photonics neuromorphic computing has promising applications due to its advantages of low latency, low power consumption, and high bandwidth. Transistors with asymmetric electrode structures are gaining attention for their low power consumption, high optical response, and simple preparation technology. However, there is a lack of systematic research and mechanism analysis on intelligent optical synapses constructed by asymmetric electrodes.