Multivariate ``needle'' polynomials with application to norming sets and cubature formulas (Q343239)

The author provides a comprehensive study of a certain class of the so-called needle polynomials that attain value \(1\) at a given point and decrease exponentially as we move away from this point which have found many applications (e.g. [\textit{A. Kroó, D. S. Lubinsky}, Can. J. Math. 65, No. 3, 600--620 (2013; Zbl 1267.42029)]). Comparing with the univariate case where there is an essential difference between the sharpness of the corresponding needles at the inner and boundary points, it is observed that this phenomenon becomes more intricate in the multivariate case, studied here. It is shown that the sharpness of multivariate needle polynomials at the boundary of convex bodies is closely related to the geometry of the boundary of the domain at the corresponding points. In particular, it is shown that the existence of \(\alpha\)-needle polynomials with \( 1 <\alpha <1\) yields that the domain cannot be of higher than \(C^{2\alpha}\) smoothness in any 2-dimensional cross section. It is observed that homogeneous logarithmic needle polynomials are possible only at vertex points of convex bodies. Applications of needle polynomials related to the distribution of optimal meshes (see [\textit{K. Jetter} et al., Math. Comput. 68, No. 226, 733--747 (1999; Zbl 1042.41003)] and [\textit{J.-P. Calvi} and {N. Levenberg}, J. Approx. Theory 152, No. 1, 82--100 (2008; Zbl 1145.41001)]) are also exhibited.