Light-Induced Artificial Synapse Integrated With AlGaN Quantum Well DUV Source for Neuromorphic Computing

Artificial optoelectronic synapses, bio-inspired neuromorphic devices, are gaining prominence for their pivotal role in overcoming the von Neumann bottleneck in traditional computing. However, challenges remain in terms of integrability and compatibility. Here, we present a light-induced, single-chip artificial synapse integrated with a deep ultraviolet light source using an AlGaN quantum well structure. In such a synaptic device, excitation light, photogenerated carriers, and electrodes are used as action potentials, neurotransmitters, and pre-synaptic/post-synaptic membranes, respectively. The device exhibits favorable synaptic behaviors, such as excitatory postsynaptic current, paired pulse facilitation, and photoelectric potentiation-depression. Subsequently, an artificial neural network is simulated, achieving a maximum handwriting image recognition accuracy of 94.5%. Such a device, integrating on-chip artificial synapse and light source, possesses significant integration potential, high compatibility, and demonstrates promising capabilities for application in neuromorphic visual systems.