Convergence and quasi-optimality of adaptive nonconforming finite element methods for some nonsymmetric and indefinite problems (Q707578)

A standard adaptive finite element method (AFEM) consists of successive loops of the cycle: \[ SOLVE\rightarrow ESIMATE\rightarrow MARK\rightarrow REFINE \] Given a bounded polyhedral domain \(\Omega\subset \mathbb {R}^{d},\quad d=2 \,\text{or}\, 3, \) the authors consider the following nonsymmetric and indefinite elliptic boundary value problem \[ \begin{aligned} -div(\mathbf{A}\nabla u)+ \mathbf{b}\cdot\nabla u+cu&=f \qquad\text{in}\quad \Omega,\\ u&=0\qquad\text{on}\quad \partial\Omega. \end{aligned} \] It is supposed that \(\mathbf{A}:\Omega\mapsto\mathbb {R}^{d\times d}\). The authors extend an ANFEM method (nonconforming and FEM for nonsymmetric and indefinite problems) to nonsymmetric and indefinite elliptic boundary value problems. The main result of the paper is the following convergence and quasi-optimality for this ANFEM. Suppose \(\tau_{h}\) is generated by REFINE from \(\tau_{H}.\) If \(\mathbf{A}\) is piecewise constant and \(h_{0}\) small enough, then there are constants \(0<\rho<1\) and \(\beta>0\) such that \[ \mid\parallel u-u_{h}^{N} \mid\parallel_{h}^{2}+\beta J_{h}^{2}(u_{h}^{N})\leq\rho(\mid\parallel u-u_{H}^{N}\mid\parallel)_{H}^{2}+\beta J_{H}^{2}(u_{H}^{N})). \] Here \(\mid\parallel_{h}\) stands for the energy norm and \(J\) are volume terms. The last section of the paper presents numerical experiments.