High order parametric polynomial approximation of quadrics in \(\mathbb{R}^d\) (Q662072)

This paper presents a method to approximate implicit quadrics in \(\mathbb R^d\) by parametric polynomial hypersurfaces with \(d-1\) parameters. This class of hypersurfaces only admits rational parametrizations, i. e., exact polynomial parametrizations do not exist. A motivating example that is included is the \(2\)-sphere, showing the clear advantage of the method when compared to Taylor's expansion approximation. The method consists in minimising the error term arising from the implicit equation of a quadric. The authors prove that this minimisation is equivalent to minimising the normal distance between the quadric and the polynomial approximant. They also provide a description of how their method applies in the case of conic sections. The main part of the paper focuses an asymptotic analysis of the method. It is shown that the distance decreases at least exponentially w.r.t. the polynomial degree. This result is confirmed by numerical examples in \(\mathbb R^3\), and strengthens the conjecture of \textit{K. Höllig} and \textit{J. Koch} [``Geometric Hermite interpolation with maximal order and smoothness'', Comput. Aided Geom. Des. 13, No. 8, 681--695 (1996; Zbl 0875.68880)] that one can achieve an approximation order of \(2n\) when using parametric polynomial approximation.