Advancing optical neural networks: photonic crystal perceptrons for ultra-fast AI computing

This paper presents what we believe to be a novel approach to designing an artificial neuron, specifically a basic perceptron, using two-dimensional photonic crystals for the first time. By utilizing a square lattice of silicon rods in an air background, the proposed structure demonstrates potential for enhanced processing speed, reduced energy consumption, and scalability for more complex AI applications. Various design examples, including threshold detection, linear classification, logical AND gates, and Iris flower classification, are presented to validate the effectiveness of the proposed photonic crystal neuron. The results show that the proposed structure can efficiently solve several computational problems, paving the way for the development of next-generation optical processors for AI applications.