Abstract
In this study, we investigate the performance of the variants of the induced dimension reduction (IDR) in large-scale electromagnetic scattering problems. We also focus on the tolerance to the so-call spurious convergence and convergence property of the IDR variant methods. Comparative numerical experiments, using IDR(s), Bi-IDR(s), and GIDR(s, L), reveal that GIDR(s, L) with L = 2 or 4 and an s value of around 15 shows the best performance with respect to the balance between the convergence property and the tolerance to the spurious convergence for problems with several levels of geometry complexity.
