▎ 摘　　要

NOVELTY - The device has a main portion which comprises a contact metal electrode (1), patterned graphene array (2), dielectric layer film (3) and a substrate (4). The patterned graphene array is in contact with the dielectric layer film formed on the substrate. The contact metal electrode is in contact with the patterned graphene array and is used together with the substrate to generate an electric field, and the control device outputs an optical signal as the input weight of the synapse. The contact metal electrode is preferably gold, titanium, palladium and nickel metal electrodes. The dielectric layer film is a high-quality oxide film grown on the substrate and the contact metal electrode is preferably hafnium dioxide (HfO2) and silicon dioxide (SiO2). The thickness of the dielectric layer film is preferably 300 nm. The material of the substrate is preferably silicon. USE - Optical nerve unit device used in field of neural networks. ADVANTAGE - The graphical graphene has high carrier density and supports the propagation of surface plasma excimer, providing an uncomparable advantage, namely the wavelength in the range of infrared to terahertz at room temperature can be dynamically tuned by using the gate voltage. The graphene SPP filter can easily and stably process the input light signal, by changing the Fermi level of the graphene, it can simply adjust the output light power. The system based on graphene optical nerve is simpler than the traditional RRAM, does not need standard SRAM or comparison circuit when it works, similar to the human retina, the graphene array can pre-process and extract important features of the input image. graphene is highly compatible with the mature silicon-based devices, making it possible to produce optoelectronic networks and multi-pixel CMOS readout circuits at DETAILED DESCRIPTION - An INDEPENDENT CLAIM is included for a construction method for deep learning optical neural network. DESCRIPTION OF DRAWING(S) - The drawing shows a side view of the optical nerve unit device used in field of neural networks. 1Contact metal electrode 2Patterned graphene array 3Dielectric layer film 4Substrate