The Sparse Localized Schur node () is an attribute to a Schur Solver and to a Krylov Preconditioner added under a Schur Solver node. This node is added by default, together with a Direct solver subnode. The Sparse Localized Schur node (in contrast to the alternative Localized Schur node) filters out small matrix elements and can therefore use all solvers, which you can add by right-clicking the Sparse Localized Schur node if you want to use another solver than the default Direct solver.

Use the Drop tolerance field to tune the maximum allowed sizes of dropped (neglected) elements (default: 0.001). A smaller drop tolerance means that the sparse localized Schur solver drops fewer elements and so it becomes more accurate (if it is set to 0, the Sparse Localized Schur node behaves like the Localized Schur node). This leads to fewer iterations in the Schur solver, but memory requirements and preconditioning time increase. A larger drop tolerance means that the filtering drops more elements and so memory use and solution time decrease. In this case, however, the Schur solver becomes less accurate, which leads to more iterations, or, if the drop tolerance is too high, to no convergence at all. Often it is most efficient to use as high a drop tolerance as possible; that is, choose it so that the Schur solver barely converges.

Use the Termination technique list to select how to terminate the solver. Select:

Further options that apply to the selection (as indicated at each bullet) made in the Termination technique list are: