On The Computation Of Electromagnetic Dyadic Green'S Function In Inhomogeneous Forward And Inverse Scattering Problems
Feng Han,Na Liu,Hai Liu,Qing Huo Liu
DOI: https://doi.org/10.1109/PIERS.2016.7735491
2016-01-01
Abstract:The dyadic Greenu0027s function (DGF) method is widely used for the calculation of electromagnetic forward and inverse scattering in unbounded media, layered media, anisotropic media, etc. The forward scattering volume integral equations are usually solved by iteratively updating the total electric fields inside the inversion domain. In this process, the Greenu0027s functions for source and field points both inside the inversion domain are used to operate on the equivalent current in each iteration step to update the scattered fields which are also sampled inside the inversion domain. In most previous studies, the magnetic vector potential A values are first calculated by using the Greenu0027s function G AJ instead of the direct computation of electric fields E through the Greenu0027s function G EJ . Those A values for the sample points inside the inversion domain are later converted to electric fields E by divergence and gradient operations. The reason for this indirect derivation is to avoid the singularity of order 1/|r − r′| 3 in the Greenu0027s function G EJ . In contrast, G AJ has a singularity of order 1/|r − r′|, thus is easy to treat in the integral when the first order basis functions, e.g., the rooftop basis functions, are used [1]. In this paper, we compare the computational efficiency and precision of the forward inhomogeneous object scattering problem solved by two types of Greenu0027s functions. The singularity of G EJ is avoided by using Discrete Dipole Approximation (DDA) [2]. Both the electric field E and the magnetic vector potential A are expanded using rooftop basis functions. We validate these numerical results by comparing them with a commercial software simulation results, and find that the solved total electric fields inside the inversion domain through the Greenu0027s function G EJ and G AJ agree well in some situations but show large discrepancies in other situations. Two typical computation configurations that cause these large discrepancies are long but thin inverse domain, i.e., the discretized cell has obvious different scales in three directions, and low operating frequency, i.e., the wavelength is much larger than the scatterer inside the inversion domain. The reasons for these results are analyzed.