A posteriori estimation of the linearization error for strongly monotone nonlinear operators (Q2372914)

Let \(V\) be a reflexive Banach space. Consider a nonlinear operator \(A:V \to V'\) with (i) \(A0=0\), (ii) \(A\) is strongly monotone, (iii) \(A\) is Lipschitz continuous for bounded arguments. Further, let \(H\supset V\) be a Hilbert space and \(A_L:H \to H'\) be a linear approximation of \(A\), supposed to be coercive and bounded. Some computable upper and lower estimators are constructed for the error function \(e_L=u-u_L\) where \(u\) (resp., \(u_L\)) is the unique solution in \(V\) (resp., \(H\)) of the problem \(Ax=F\) (resp., \(A_Lx=F\)). Some numerical examples are then given to illustrate the method.