@Liping: unfortunately the only source I have is in Italian and it is published by my university (Politecnico di Milano). However, the rationale is described above as an aswer to the question of "wei".

@Greg: The problem arises due to an error in the inputs. If you open the Help of the GUI, you will read for Electric Permittivity:

"Electric permittivity

...Complex values of the electric permittivity (indicating power losses associated to the material) are allowed (the imaginary part has to be negative, according to the temporal convention exp(j*omega*t))..."

This means that the right input in your case should be:
eps = [1 -100-100*j 1]

@Alexander: unfortunately I have no documentation that could be useful to explain the whole procedure.
Concerning the boundary conditions: they are the usual ones for EM fields, i.e the conservation of the total tangential H component and of the total normal B component at each interface between 2 layers.

@Alexander: The primary field is the radiation emitted by the EM source, in this case, the magnetic dipole; the secondary field originates from the interation of the primary field with the multilayered structure, i.e. consists in the reflected and transmitted components of the field. Coming to your first question (if I have understood it correctly), the method can be applied to most of the radiation patterns that are known in a closed-form solution (both concerning E and H, therefore it is possible to calculate the full EM verctor) . The tough part may be to mathematically decompose the primary field into several elementary waves. Afterwards, the propagation of simpler waves is relatively easy.

@wei: the radiation is calculated as follows:
- first the exact solution of the magnetic dipole radiation is decomposed into several more simple wave functions (plane waves or cylindrical waves)
- each simple wave interacts with the multilayered structure according to the customary EM boundary conditions
- the final radiation pattern is obtained by summing up all the contributions of the simple waves.

@Liping: unfortunately the only source I have is in Italian and it is published by my university (Politecnico di Milano). However, the rationale is described above as an aswer to the question of "wei".