Surface Integral Equation Methods And Its Parallelization Techniques Of Electrically-large 3d Dielectric Structures

With the progress in the technology of stealth and anti-stealth and the development of civil microwave engineering, many complex dielectric structures emerge. Moreover the metal shows dielectric properties at the optical frequencies. It is significant in theory and engineering to carry out the study of dielectric targets. Based on the above background, the electromagnetic scattering properties of the dielectric targets are studied in the thesis.The solution is based on surface integral equation(SIE), and the method of moment(MOM) as the dispersing scheme, the PMCHW and JMCFIE is mainly studied. As the traditional discretization of SIE leads to ill-conditioned impedance matrix with high condition number, two optimization methods from different aspects are proposed here. For the limitation of MOM for electrically large problems, the multilevel fast multipole algorithm(MLFMA) is employed to reduce the computing and storage complexity. To further strengthen the capacity of solving electrically large targets, a paralleled version of MLFMA based on MPI-OpenMP is developed for symmetric multi-processing(SMP) clusters.Begin with the fundamental electromagnetic theory, all kinds of integral equations are introduced together with the basic principles of MOM and some of its the key details. Regular and irregular targets are analysized, and the numerical results prove the reliability of the code.Then some common SIEs are deduced. Two improving schemes called left and right(L&R) diagonal matrix preconditioner and the surface integral equations using normalized field quantities are discuss in detail. Numerical results show that the both of the improving scheme proposed here reduce the condition numbers of the impedance matrix effectively and also lead to a clear improvement on the convergence of iterative solutions.Furthermore the paralleled MLFMA is added to strengthen the capacity of solving electrically large targets. Firstly the framework of MLFMA is described in detail, and the specific MLFMA expressions in the case of dielectric problems are derived. Subsequently, a detailed description of the basic theory of parallel computing, including pa-rallel hardware systems and software programming models is followed. Eventually the paralleled MPI-OpenMP-MLFMA is presented including the mixed portioning strategy of task, MPI-OpenMP hybrid parallelization scheme, multilevel aggregation-transformation-disaggregation pattern and some other details. Numerical results demonstrate the efficiency of parallel MLFMA for solving large objects.Finally, as an extensional application of the surface integral equation, preliminary study of the surface plasmon(SPs) of metallic nanostructures is followed. Some numerical experiments of the near-field enhancing are conducted, and some interesting phenomena appear.