Using a deep neural network for the design of an optical photonic crystal half-subtractor

In this paper, a new structure, to our knowledge, for an optical half-subtractor using periodic photonic crystal structures is presented. The difference between this structure and previous ones is that a deep neural network (DNN) is used to optimize the proposed structure. A network composed of silicon rods in air is utilized to achieve the desired structure. The waveguides of this half-subtractor are designed using linear defects and a very small number of point defects. The variable parameters in this simulation are the defect rods, which are optimized by altering them. This structure, due to its simplicity in design and small size, is suitable for use in optical integrated circuits. Another advantage of this structure is the equal power output for the high states. In this study, the plane wave expansion (PWE) method was used to calculate the band structure, and the finite difference time domain (FDTD) method was used to calculate the light emission and output of light power.