Superconvergence of triangular mixed finite elements for optimal control problems with an integral constraint (Q606732)

An optimal control problem with quadratic cost functional, linear elliptic state equation and control constraints is considered. The admissible control set belongs to \(L^2(\Omega)\), where \(\Omega\) is a bounded 2D domain. The state and the adjoint state are approximated by Raviart-Thomas mixed finite elements of order \(k=1\) and the control is discretized by piecewise constant functions. The authors demonstrate that the error for the control is of order \(O(h^2)\) in the corresponding \(L^2\) norm, where \(h\) is the finite element parameter. Moreover, using a postprocessing technique, the same error estimate is obtained between the projection of the discrete adjoint state and the exact control. Two examples with numerical tests are presented. For one of them the state equation is nonlinear.