DK12: Efficient Solvers for KKT Systems

KKT systems are special indefinite linear systems of equations with a natural block 2-by-2 structure.  Of particular interest in this project are KKT systems that result from the discretization of  constrained optimization problems in function spaces, like optimal design problems or optimal control problems, and from the discretization of mixed variational problems for systems of partial differential equations.

The focus in this project is the construction and analysis of efficient iterative methods for solving such systems, based on symmetric indefinite preconditioners. For these methods better properties are anticipated than for the better-understood block triangular preconditioners, if applied to problems with a (1,1) block which is only semi-definite, a property frequently occuring in KKT systems.

Supervisor: Prof. Walter Zulehner