Two-grid algorithms for some linear and nonlinear elliptic systems (Q5962001)

Two-grid algorithms in the finite element methods consist in using two subspaces, one with fine grid \(\nu_{h}\) and one with a coarse grid \(\nu_{H}\) when there is a relation \(h\ll H\) between the mesh-sizes. In this paper the authors explore the two-grid idea and present new algorithms for linear and nonlinear coupled elliptic systems, respectively. For this purpose they use the information of the adjoint operator, and in the nonlinear case, a modified Newton iteration. First the new two-grid algorithm is applied to the following nonsymmetric linear system: \[ \begin{cases} -\nabla\cdot(\mathbf{K}_{i} \nabla c_{i})+\mathbf{b}_{i}\cdot \nabla u_{i}+\sum_{j=1}^{i} V_{ij}u_{j}=f_{i},\quad &\text{in} \quad \Omega\\ u_{i}=0 &\text{on} \quad \partial\Omega. (i=1,\dots,l) \end{cases} \] Systems of this form arise in some linearizations of the reaction-convection-diffusion equations. Under some assumptions it is proved that for \(H\) small enough \[ \|\mathbf{u-u}^{h}\|_{1}\leq C(h+H^{2})\parallel\mathbf{u}\parallel_{2}\quad \forall \mathbf{u}\in {H}^{1}(\Omega). \] Similar results are proved under certain assumptions for the semilinear transport systems of the form: \[ \begin{cases} -\nabla\cdot(\mathbf{K}(x) \nabla \mathbf{u})+f(x,\mathbf{u})=\mathbf{0}&\text{in} \quad \Omega\\ \mathbf{u}=\mathbf{0} &\text{on} \quad \partial\Omega. \end{cases} \] Numerical examples are also provided.