Synaptic solar-blind UV PD based on STO/AlXGa1−XN heterostructure for neuromorphic computing

Abstract

The rapid advancements in artificial intelligence and high-performance computing have emphasized the need for efficient optoelectronic artificial synapses, essential elements in neuromorphic computing. This study proposes a solar-blind ultraviolet (UV) photodetector (PD) based on the SrTiO3/AlXGa1−XN heterostructure to serve as an optoelectronic synapse. Under 265 nm illumination, the device demonstrates a remarkably low dark current of 1.08 × 10−11 A and an impressive peak responsivity of 36.43 A/W at −15 V. Notably, the UV PD functions as an optoelectronic synapse that emulates a biological neuron, simulating the fundamental operations of various biological synapses. Moreover, the research extends to the promising field of neuromorphic computing. The photoelectric artificial synapse device achieved an exceptional 97.91% accuracy rate in the challenging MNIST handwritten digit recognition task, further validating its potential in neural computing applications.