The Cycle Structure Rcwa Simulation Software And Hardware Acceleration Method Research

Field simulation calculation for the periodic structure is very important in the material and device design, manufacturing and measurement. There are varieties of simulation methods for different structures and materials. Rigorous Coupled-wave Analysis (RCWA) is a method to get the field distribution of the media by solving the Maxwell equations strictly. It’s also called Fourier modal method (FMM), as it solve the equations by Fourier expansion with the periodicity of the media and field. The implementation of this method can be used in the design, manufacturing and measurement of grating, microelectronic structure, photonic crystals and many other periodic structures.The simulation process is very slow as involved lots of matrix operations. The convergence of the algorithm is quite different for different structure and incent wave. To be more efficient in the implementation of the simulation, there are two aspects to consider, first is improving the convergence by modify the algorithm, the second is to take advantage of new hardware to accelerate the matrix operations. At present, the major research work to improve the computational performance is about the algorithms.Considered from the practical application, in this article, the major research is about how to use the math library implemented with GPU (Graphic Processing Unit) to accelerate the simulation process of RCWA. Most of the time, the math library with GPU is more efficiency when operating on the large matrices. But the matrix size is related to the implementation of RCWA. Therefore, this article analyzes the RCWA and an improved algorithm first. And then analyzes some common libraries of matrix operations, including the version of using CPU only and the implementation with hybrid architecture of CPU and GPU. The advantages and weakness of GPU to do matrix operation was illustrated through the analysis of GPU features. And then verify this by testing several common functions used in the RCWA. Based on these tests, a matrix library was designed and implemented by combine several different matrix libraries to provide a unit interface for external calls. Finally, this matrix library was used in the RCWA algorithm implementation, and the acceleration effect of the GPU was tested in practical environment.