Volume 13, Issue 1

High-order finite difference approximations of the solution and the flux to model interface problems in one-dimension are constructed and analyzed. Explicit formulas based on new Marchuk integral identities that give O ( h 2 ), O ( h 4 ),… accuracy are derived. Numerical integration procedures using Lobatto quadratures for computing three-point schemes of any prescribed order of accuracy are developed. A rigorous rate of convergence analysis is presented. Numerical experiments confirm the theoretical results.

For the 2D eddy currents equations, we design an adaptive edge finite element method (AEFEM) that guarantees an error reduction of the global discretization error in the H (curl)-norm and thus establishes convergence of the adaptive scheme. The error reduction property relies on a residual-type a posteriori error estimator and is proved for discretizations based on the lowest order edge elements of Nédélec's first family. The main ingredients of the proof are the reliability and the strict discrete local efficiency of the estimator as well as the Galerkin orthogonality of the edge element approximation.

In [2] we introduced a new type of mixed finite element approximations for two- and three-dimensional problems on distorted polygonal and polyhedral meshes that consist of cells having different forms. Additional degrees of freedom that arise in the process are excluded by a special condition that is natural for the mixed finite element approximations considered. This paper is devoted to the error analysis of the respective finite element solutions. We show that under certain assumptions on the regularity of the exact solution the convergence rate for the new approximations is the same as for the Raviart–Thomas finite element approximations of the lowest order.

The paper presents a new class of memory gradient methods with inexact line searches for unconstrained minimization problems. The methods use more previous iterative information than other methods to generate a search direction and use inexact line searches to select a step-size at each iteration. It is proved that the new methods have global convergence under weak mild conditions. The convergence rate of these methods is also investigated under some special cases. Some numerical experiments show that these new algorithms converge more stably than other line search methods and are effective in solving large scale unconstrained minimization problems.

The techniques to derive residual based error estimators for finite element discretisations of variational equations can be extended directly to variational inequalities by employing a suitable adaptation of Nitsche's idea (c.f. [8]). This strategy is presented here for elliptic variational inequalities. Its application is demonstrated at the obstacle problem, where numerical results show that the proposed approach to a posteriori error control gives useful error bounds.