|M.Sc Student||Boaz Blankrot|
|Subject||A Fast Source-Model Technique for Many-Scatterer|
Problems in Electromagnetics
|Department||Department of Electrical Engineering||Supervisor||Full Professor Leviatan Yehuda|
|Full Thesis text|
An accurate and efficient analysis of electromagnetic scattering by many scatterers of different shapes and sizes remains a problem of great interest. Many-body scattering problems naturally arise in many practical applications such as wave propagation through haze and fog, metamaterial analysis, and radio-frequency design for antennas, stealth, electromagnetic interference, and packaging.
The Source-Model Technique (SMT) has been utilized extensively to solve problems of electromagnetic scattering. When SMT is applied to scattering from perfect electric conducting bodies, the scatterer is replaced by discrete sources of unknown amplitude located beneath its surface, and it is then required that these sources, together with the incident field, hold the boundary conditions at a set of matching points on the original surface. Although this technique has many advantages over other integral equation methods, it does not scale to a large number of scatterers.
In this thesis, we develop a new SMT-based method which scales to many-scatterer problems by incorporating a fast multipole formulation into the SMT. The Fast Multipole Method (FMM), introduced in 1987, divides interactions into far and near ones while grouping the far interactions, allowing an implicit representation of the problem and a reduced run time. We incorporate a fast multipole formulation into SMT and develop the Fast Multipole Source-Model Technique (FMSMT). This technique enjoys both the advantages of the SMT and the benefits of efficiently representing near and far interactions. This makes FMSMT especially attractive for many-scatterer problems.
We show that FMSMT achieves significantly better run time over SMT in a variety of two-dimensional and three-dimensional many-body scattering problems. In all of these cases, the improvement is already notable for a minimal number of unknowns and grows with the size of the problem.