Sharp Bernstein inequalities on simplex

From MaRDI portal





The Bernstein inequality is one of the most important standard tools in approximation theory. It is so relevant for error estimates which are at the core of convergence and other theorems that it can be considered as one of the principal work horses of numerical analysis.\N\NIt bounds \(p\)-norms of a large ``zoo of different forms, with or without weights, on all sorts of domains, from the derivatives of functions or partial derivatives of that function, by the the same \(p\)-norm of the function itself. That derivative is typically multiplied by some specific multiplier \(\psi\).\N\NIn this excellent paper, the two authors address some new, very specific situations. They use mostly simplices in \(d\) dimensions as the aforementioned underlying domain, but employ special weights on the left-hand side of the inequality and certain forms of factors.\N\NThere are novelties in many directions in this article: a Bernstein inequality with the weight \(1/\sqrt{1-x}\) which has been established in two dimensional simplices (i.e., triangles) is extended to any dimension. The mentioned factors \(\psi\) on the lower side of the inequality is generalised to functions other than Euclidean distances as Kroó used before. In these extensions, second-order linear differential operators are used whose eigenfunctions are orthogonal polynomials.\N\NThose (spectral) operators can be expressed in many symmetric shapes each of which provides different, new Bernstein inequalities of the same form, albeit with different factors than the mentioned \(1/\sqrt{1-x}\) multiplier. All derivatives and all dimensions can appear now in these Bernstein inequalities, and several of them are optimal. Different \(L^p\)-norms are admitted too.











This page was built for publication: Sharp Bernstein inequalities on simplex

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6918071)