Wideband nested cross approximation for Helmholtz problems (Q2340896)

Compared with explicit kernel approximation, significantly better approximations can be expected from the adaptive cross approximation (ACA) method of \textit{M. Bebendorf} [Numer. Math. 86, No. 4, 565--589 (2000; Zbl 0966.65094)], due to the quasi-optimal approximation properties given by \textit{M. Bebendorf} [Hierarchical matrices. A means to efficiently solve elliptic boundary value problems. Berlin: Springer (2008; Zbl 1151.65090)]. The authors generalizes ACA (which achieves log-linear complexity only for low-frequencies) to high-frequency Helmholtz problems by constructing approximations to A with complexity \(k^2 N \log N\) using only few of the original entries of A. It is interesting to observe that the method presented here allows for a continuous and numerically stable transition from low to high wave numbers \( \kappa\) by a generalized far-field condition that fades to the usual far-field condition if the wave number becomes small. It is also concluded that the log-linear overall storage and the log-linear number of operations are required by the new technique. To validate the analysis presented some numerical experiments are given.