A posteriori error estimates for a finite element approximation of transmission problems with sign changing coefficients (Q548331)

The authors consider a Galerkin approximation for a Dirichlet problem attached to a second order divergence type equation with sign changing coefficients. The lack of the coerciveness is the key point in this analysis. They prove the well-posedness of the discrete approximation and carry out an a posteriori error analysis provided that the contrast, i.e., the ratio between the min of positive ``part'' of the coefficient and the max of the absolute value of the negative ``part'', is large enough. Two numerical examples (a polynomial and a singular solution) are considered in order to illustrate the theoretical predictions.