Electromagnetic Modeling Of Nonlocal Dispersion Model Based On Parallel FDTD Algorithm And Its Application

The Finite-Difference Time-Domain(FDTD)algorithm is a time iterative numerical algorithm.The advantage of FDTD is that wideband results can be obtained in only one single calculation.FDTD is the core algorithms of electromagnetic simulation technology and is widely used in military,scientific research and consumer electronics.It has a wide range of applications in aircraft stealth and guidance,optical devices and biosensing,microwave devices and antennas,electromagnetic compatibility and signal integrity.In the past years since it is proposed,the FDTD algorithm has made great progress.However,with the development of science and technology,more stringent requirements have been placed on electromagnetic simulation technology and numerical algorithms.Nowadays,The FDTD algorithm has shortcomings in grid generation,computational efficiency,dispersion material model and computation platform design,which is difficult to meet the needs of contemporary electromagnetic simulation calculation.In order to solve these problems,this dissertation has mainly completed the following aspects:1.The FDTD algorithm first needs to solve the mesh generation problem in engineering applications.For simple structures,the Yee grid can be generated based on the geometric definition.However,for the complex structures,the mesh generation algorithm is needed.This dissertation proposes a new mesh generation algorithm based on ray tracing,which is highly tolerant and suitable for all kinds of complex structures.In order to further optimize the mesh generation algorithm,the combination of ray tracing technology and bounding volume hierarchies data structure greatly improves the mesh generation efficiency,and the grid generation time is reduced by two orders of magnitude compared with the existing mesh generator.In addition,the grid generation algorithm proposed in this dissertation can be directly combined with the Computer-Aided Design(CAD)kernel,and the geometric model output from the CAD design software can be used in the FDTD algorithm directly,which saves the geometric modeling time in engineering simulation.2.As a full-wave numerical algorithm,FDTD requires a lot of computational resources and computation time when processing large-scale simulations.In order to make full use of computing resources and reduce simulation time,this dissertation proposes a hybrid parallel algorithm of FDTD based on Message Passing Interface(MPI)and Open Multi-Processing(OpenMP).All key techniques such as dispersion material model,Fourier transform and total field scattering field are processed in parallel.The simulation results show that the hybrid parallel algorithm has the same memory usage as the non-parallel algorithm,and the parallel algorithm reduces the simulation time.The advantage of the hybrid parallel algorithm is that it uses OpenMP's shared memory technology to make full use of the stand-alone computing resources,and then uses MPI's distributed memory technology to perform cluster computing,which improves the utilization efficiency of computing resources.3.In nano-optics,due to the reduction of metal size,the quantum effect becomes more and more obvious,and the nonlocal phenomenon occurs.The free electrons gas on the metal surface is governed by the hydrodynamics equation,and the simplified hydrodynamic equation is coupled with Maxwell's equation to obtain the nonlocal dispersion model.In this dissertation,the nonlocal dispersion model is used in the FDTD algorithm.The simulation results show that the extinction cross-section has a blue shift phenomenon,which is consistent with the experimental phenomenon.In addition,the reasons for the peak shift and amplitude change are explained theoretically,which is of great significance for the study of the optical properties of metal nanoparticles.4.This dissertation designs the FDTD computing platform based on ray tracing mesh generation algorithm,hybrid parallel technology,and nonlocal dispersion model.The platform is divided into two parts:the interactive interface and the computing engine.The interactive interface is responsible for parameter setting and result display,and the computing engine is responsible for high-performance parallel FDTD calculation.The FDTD computing platform is simple to operate and has high computational efficiency.It can be used for nano-optical research and electromagnetic computing in engineering.