Stable multiscale bases and local error estimation for elliptic problems (Q676133)

The paper deals with a posteriori estimators for Galerkin schemes for the solution of elliptic equations. Using the properties of stable multiscale bases (for which a brief summary is given) and, in particular, the norm equivalences for ranges of Sobolev norms and the compression of matrices in wavelet representation, the derivation of a posteriori error estimators in terms of wavelet coefficients of residuals is given. These estimators are shown to be computable, efficient and reliable. Moreover, it is proved under relatively mild hypotheses on the data that these error estimates give rise to convergent adaptive schemes without assuming the so called saturation property. It is interesting that these results are applicable for a large class of elliptic operators covering partial differential operators (also of negative order) and singular integral ones.