On Bernstein- and Marcinkiewicz-type inequalities on multivariate \(C^\alpha\)-domains (Q6651973)

Bernstein, Markov and Marcinkiewicz-type inequalities are well-known in one dimension and have been extended for some time into multiple dimensions. They give upper bounds (the desired aim is to have bounds with best possible [``sharp''] constants) to derivatives of uni- or multivariate polynomials of some maximum totald degree \(n\), possibly including non-negative weight functions.\N\NThe norms used within these bounds can be just uniform norms, but in this excellent paper they are general \(L^p\)-norms in the case of Bernstein-Markov inequalities and sample points and partitions in the case of Bernstein-Marcinkiewicz inequalities. The number of elements of the partitions are bounded by the degree of the polynomials and the dimension of the space.\N\NIn the former situation of Bernstein-Markov, the estimates are valid for all positive finite \(p\) (it should be pointed out that \(0<p<1\) is explicitly included), and for the latter case of Bernstein-Marcinkiewicz, \(p>d-1\) is required. In that case the upper bounds include a very simple factor of \(2\) for the \(L^p\) norm of the polynomial and \(4\) for the sum of sample points times the size of the partion sets.\N\NRoutinely, the ambient sets are unit-intervals (in the case of one dimension) or unit-balls (in the case of more than one dimension), but the present authors provide entirely new bounds based on general compact sets. This now requires the use of normal-derivatives and tangent-derivatives on the boundaries of those compact sets. Depending on the smoothness \(\alpha\) of the boundary of the domain, the bounds include either a factor of \(n\) or \(n^{2/\alpha}\), where \(n\) is the total degree of the polynomials to which the inequality applies. The former case requires a certain weight-function applied to the left-hand side.