Comparison of TE and TM Inversions in the Framework of the Gauss-Newton Method

The Gauss-Newton inversion method in conjunction with a regularized formulation of the inverse scattering problem is used to invert transverse electric (TE) and transverse magnetic (TM) data. The utilized data sets consist of experimental data provided by the Institut Fresnel as well as synthetic data. The TE inversion outperformed the TM inversion when utilizing near-field scattering data collected using only a few transmitters and receivers. However, very little difference was found between TE and TM inversions when using far-field scattering data. It is conjectured that the reason for the better performance of the near-field TE result is that the near-field TE data contains more information than the near-field TM data at each receiver point. In all cases considered herein, the TE inversion required equal or fewer iterations than the TM inversion. The per-iteration computational complexity of both TE and TM inversions is discussed in the framework of the Gauss-Newton inversion method. Actual costs are consistent with the computational complexity analysis that is given.