An S-matrix based hybrid method


Gregory A. Kriegsmann

Department of Mathematical Sciences

New Jersey Institute of Technology, Newark, NJ 07102

The use of S-Matrix theory by engineers and scientists to understand and design microwave systems has roots that are over sixty years old. The basic idea behind the method is to decompose the scattering domain into sub-domains and to characterize the scattering physics in each by an S-Matrix. These are then "connected" using simple matrix algebra to achieve an accurate approximation of the complete scattering problem. Of course the method is viable only if an accurate approximation can made to the S-Matrices in each sub-domain. These approximations were computed analytically for simple structures and were obtained experimentally for more realistic and complicated scatterers.

This theory has more recently been used, in conjunction with numerical algorithms, to produce accurate and efficient hybrid schemes for several classes of scattering problems that are not amenable to either finite difference, finite element, or integral equation methods. The novel feature here is to replace the experimental determination of the S-Matrix for a sub-domain by an accurate numerical approximation to it.

Two examples will be presented and discussed which illustrate the methodology. The first arises in microwave heating experiments in single mode, highly resonant cavities. The highly resonant character of the problem makes finite difference time domain methods inefficient, due to the time required to radiate away the "transients" present in the simulation. Similarly, finite element and integral equation methods give rise to ill-conditioned algebraic problems. The hybrid method suffers neither shortcoming. Numerical experiments will be given.

The second example arises in simple photonic structures. If the scatterers that make up this structure are sufficiently far apart in the direction of propagation, then S-Matrix theory can be used to construct a very efficient and accurate numerical method to study this problem. An example is provided to make this point.

Finally, we note that the accurate description of the scattering physics in each sub-domain requires a numerical scheme that is not only accurate, but faithfully reproduces the properties of the scattering matrix. For example, if the target in the sub-domain is lossless, then the S-Matrix is unitary. The numerical method must yield this property. If it fails, then inaccuracies occur. An example is provided to illustrate this feature.