On stability constants and condition numbers of discretisation methods (Q1123547)

Let E, F, G be normed spaces and J: \(E\to G\) a continuous imbedding. Consider the problem \(Lu=f\) where L: \(E\to F\) is a linear mapping. The inverse stability constant \(\gamma_ 0\) is defined by \(\gamma_ 0=\inf \{\| Lu\|_ F\), \(u\in E\), \(\| Ju\|_ G=1\}\). Now discretize the problem using Stummel's approach and introduce the corresponding stability constants \(\gamma_ h\) for discrete problems. The author shows that under appropriate conditions \(\gamma_ h\to \gamma_ 0\) as \(h\to 0\). Using this abstract result he shows that the inverse stability constants of finite difference and collocation methods for boundary value problems of ordinary differential equations as well of Ritz-Galerkin methods for elliptic variational problems converge to those of their continuous counterparts as \(h\to 0\).