A convergence analysis of block accelerated over-relaxation iterative methods for weak block \(H\)-matrices to partition \(\pi \) (Q852621)

The aim of the paper is to establish the convergence of the block iteration methods such as the block succesively accelerated over-relaxation method (BAOR) and the symmetric block succesively accelerated over-relaxation method (BSAOR). The first section given a survey of the main concepts like: weak block diagonally dominant matrix to partition \(\pi\) and weak block \(H\)-matrix to partition \(\pi\), including the block diagonally dominant matrix, which are defined by \textit{D. G. Feingold} and \textit{R. S. Varga} [Pac. J. Math. 12, 1241--1250 (1962; Zbl 0109.24802)] and \textit{F. Robert} [Linear Algebra Appl. 2, 223--265 (1969; Zbl 0182.21302)]. The second section concerns the weak block diagonally dominant matrices and weak block \(H\)-matrices to partition \(\pi,\) giving a characterization for weak block \(H\)-matrices to partition \(\pi.\) The third section is devoted to the exact domains for the convergence and divergence of the block successive overrelaxation (SOR) and block symmetric SOR iterative methods. Based on these results, the main results of \textit{Z.-Z. Bai} [Parallel Comput. 25, 681--701 (1999; Zbl 0933.68055)] can be improved. The proof is given in the last section. Some applications of weak block \(H\)-matrices in the numerical analysis of generalized ultrametric matrices are also presented.