Residual-based a posteriori error estimates of nonconforming finite element method for elliptic problems with Dirac delta source terms (Q2519307)

The paper deals with the Poisson problem with a Dirac delta source term and a homogeneous Dirichlet boundary condition: \[ -\Delta u = \delta_{x_0} \quad\text{in }\Omega, \qquad u=0 \quad\text{on }\partial \Omega,\tag{1} \] where \(\Omega \subset\mathbb R^2\) is a bounded domain and \(x_0\) is an inner point of \(\Omega\). Two residual-based a posteriori error estimator of the nonconforming Crouzeix-Raviart finite element are derived for elliptic problem (1). Due to the singular characteristics of the solution of the problem (1), meshes adequately refined around the delta support are required to improve the quality of the approximation. Two a posteriori error estimators, \(\varepsilon_p\) and \(\eta_p\), are served as refinement indicators to guide the adaptive mesh-refining algorithms, which are based on the bulk criterion for displacement-based adaptive finite element methods and the longest-edge bisection. Main result: A reliable and efficient a posteriori estimation of the error in the piecewise \(W^{1,p}\)-seminorm for the nonconforming \(P_1\) triangular element (Crouzeix-Raviart element) is derived. The key point in the analysis is to apply a Helmholtz-type decomposition of a function in \(L^q(\Omega)^2\). The authors show that one estimator yields global upper and lower bounds of the error in the piecewise \(W^{1,p}\) seminorm, and the other gives a global upper bound in the \(L^p\) norm. Adaptive algorithms are suggested and experimentally shown to lead to optimal orders of convergence.