학술논문

Convergence of an adaptive $hp$ finite element strategy in one space dimension.
Document Type
Journal
Author
Dörfler, W. (D-KLRH-AN2) AMS Author Profile; Heuveline, V. (D-KLRH-AN2) AMS Author Profile
Source
Applied Numerical Mathematics. An IMACS Journal (Appl. Numer. Math.) (20070101), 57, no.~10, 1108-1124. ISSN: 0168-9274 (print).eISSN: 1873-5460.
Subject
65 Numerical analysis -- 65L Ordinary differential equations
  65L60 Finite elements, Rayleigh-Ritz, Galerkin and collocation methods
  65L70 Error bounds
Language
English
Abstract
The authors present an adaptive algorithm for one-dimensional elliptic boundary value problems solved with $hp$ finite elements. They prove that with the proposed scheme the error in the energy norm decreases monotonically at each step. Numerical experiments show that the decrease follows an exponential law, though a formal proof of optimality is still lacking. \par The method is based on a residual based a posteriori error estimator and on the use of a carefully selected set of refinement patterns. The adaptation procedure is formally extendable in higher dimension, though the cost of testing different refinement patterns could be high. No coarsening step has been considered in this work. Yet, the theoretical result leading to the monotone decrease of the error cannot be directly extended to higher dimensions since $p$-uniform equivalence between the discretisation error and the proposed residual-based estimation has been proved only in 1D.