Least-squares mixed finite element methods for non-selfadjoint elliptic problems. I: Error estimates (Q1923319)

The present work develops optimal error estimates for least squares mixed finite element methods for non-selfadjoint elliptic problems. As a representative problem the authors consider the second-order non-selfadjoint elliptic boundary value problem of the form \[ -\text{div} (A\;\text{grad} u)-{\mathbf b}\cdot \text{grad} u+cu=f\quad\text{in} \Omega, \] with the Dirichlet boundary condition \(u=0\) on \(\Gamma_{D}\) and the Neumann boundary condition \( {\mathbf n}\cdot (-A\;\text{grad} u)=0 \) on \( \Gamma_{N}\). Here \(\Omega \subset {\mathbb{R}}^n,\;n=2,3 \), is a bounded domain with Lipschitz boundary \(\Gamma=\Gamma_D\cup\Gamma_N,A\) is a symmetric, positive definite matrix of coefficients and \({\mathbf n} \) is the unit outward normal to \(\Gamma\). The authors develop an error analysis in \( L^2(\Omega) \) and \( H^{-1}(\Omega) \). The paper includes numerical results for a test problem.