An Efficient Implementation Of Yasuura'S Method Of Modal Expansion By Sequential Accumulation

As an efficient numerical implementation of the large-sized least-squares problem that arises in Yasuura's method, the sequential accumulation processing is introduced. We describe the numerical algorithm of the QR decomposition method with sequential accumulation to solve the large-sized least-squares problem whose Jacobian is a dense matrix or a banded one. It is shown that computer storage required at one time is significantly saved in comparison with that of the implementation without sequential accumulation. The proposed numerical algorithm enables us to apply Yasuura's method to large-sized problems such as scattering by a cluster of nanoparticles or diffraction from multilayered periodic structures.