A note on spectrum analysis of augmentation block Schur complement preconditioners (Q411532)

The author proposes and analyses a set of augmentation block Schur complement preconditioners \[ {\mathcal T}_{k,j}(W) = \begin{pmatrix} A + B^TWB & k B^T \\ 0 & j B(A + B^TWB)^{-1} B^T \end{pmatrix} \] for saddle point systems \[ {\mathcal A} \begin{pmatrix} x \\ y \end{pmatrix} = \begin{pmatrix} A & B^T \\ B & 0 \end{pmatrix} \;\begin{pmatrix} x \\ y \end{pmatrix} = \begin{pmatrix} f \\ g \end{pmatrix} \, , \] where \(A\) is a symmetric, positive semidefinite matrix and \({\mathcal A}\) is nonsingular. \(W\) is a symmetric, positive semidefinite matrix and \(A + B^TWB\) is symmetric, positive definite. The presented preconditioners are a generalization of preconditioners proposed by \textit{G. H. Golub, C. Greif} and \textit{J. M. Varah} [SIAM J.~Matrix Anal.~Appl.~27, No.~3, 779--792 (2006; Zbl 1105.65048)]. It is shown that the optimal parameters in the preconditioner \({\mathcal T}_{k,j}(W)\) are \(k = 2\) and \(j = -1\). Additionally, bounds of the eigenvalues of the preconditioned matrices \({\mathcal T}_{2,-1}(W)^{-1}{\mathcal A}\) and \({\mathcal T}_{1,1}(W)^{-1}{\mathcal A}\) as well as bounds of the real and imaginary parts of the eigenvalues of the matrix \({\mathcal T}_{1,-1}(W)^{-1}{\mathcal A}\) are derived.