Comments on "Computation of cutoff wavenumbers of TE and TM modes in waveguides of arbitrary cross sections using a surface integral formulation" [with reply

Abstract

For the original article see ibid., vol.38, no.2, p.154-9 (1990). The authors use the method of moments and convert the E-field integral equation into a homogeneous set of equations of the form ZJ=0, where Z is a complex nonsymmetric matrix whose elements depend on the wavenumber and J is a vector containing the expansion coefficients for the surface current. They compute the determinant of the matrix as a product of eigenvalues and claim that the computation requires on the order of n/sup 2/ operations, which represents a considerable savings over conventional methods. The commenters point out that since the matrix is complex and nonsymmetric, and the authors state that they first compute eigenvalues and then sort them, it is sensible to assume that the general QR algorithm was used for which the number of operation is much higher than approximately=n/sup 2/ claimed by the authors. The authors reply that the confusion arises because some important statements were left out of their paper as a result of page limitations. They provide the missing statements and clarify their conclusion. >