New preconditioners based on symmetric-triangular decomposition for saddle point problems (Q644879)

Let \[ {\mathcal A} = \begin{pmatrix} A & B \cr B^T & 0 \cr \end{pmatrix} \] be a matrix with a block structure that is typical for saddle-point problems. There are block triangular matrices \(\mathcal T\) and symmetric positive definite matrices \(\mathcal S\) such that \({\mathcal TA} = {\mathcal S} = {\mathcal LL}^T\). The matrices \(\mathcal T\) are called symmetric-triangular preconditioners and their spectral properties are studied.