A sparse \({\mathcal H}\)-matrix arithmetic. II: Application to multi-dimensional problems (Q1976416)

\textit{W. Hackbusch} [Computing 62, No. 2, 89-108 (1999; Zbl 0927.65063)] introduced in part I the concept of hierarchical matrices \({\mathfrak H}\). These matrices are data sparse and allow an approximate matrix arithmetic of almost linear complexity. In this paper, the construction of \({\mathfrak H}\)-matrices for applications in two and three dimensions is described. A general construction of block cluster trees and the reliability of \({\mathfrak H}\)-matrix approximations in the boundary element method is presented. The complexity analysis is done in several steps. The results obtained for tensor-product meshes are utilized to construct an asymptotically optimal approximation to the minimal admissible cluster tree on the given unstructured grid. Cases with additional complexity are also discussed.