On polynomial interpolation of two variables (Q1867273): Difference between revisions

This paper is an extension of another previous one by the same authors [SIAM J. Numer. Anal. 39, 1780-1793 (2002; Zbl 1014.41002)]. In that paper they show the existence and uniqueness of solutions of a bivariate polynomial interpolation problem. The interpolation data were of Hermite type: \(\mu_l\) normal derivatives of consecutive orders on \(2m+1\) equidistant points on each circle of radius \(r_l\) (\(0<r_l<1\)), \(l=1,2,\dots,\lambda\). The interpolation space was that of bivariate polynomials of degree not greater than \(n\), assumed that \[ \mu_1+\mu_2+\cdots+\mu_l=\biggl[{n\over 2}\biggr ]+1 \] and \[ m=\biggl [{{n+1}\over 2}\biggr ]. \] In the present paper the authors consider a Birkhoff problem introducing an incidence matrix which establishes which normal derivatives (not necessarily of consecutive orders) are prescribed at each point. Assuming there are \(\mu_l\) prescribed derivatives at each point of the circle of radius \(r_l\) the authors prove a result similar in some senses to the Atkinson-Sharma theorem in [\textit{K. E. Atkinson} and \textit{A. Sharma}, SIAM J. Numer. Anal. 6, 230-235 (1969; Zbl 0182.21604)] for Birkhoff univariate interpolation. On the other hand, a factorization method (similar to the well-known Bézout's theorem for algebraic curves with common points) is used to combine several interpolation processes to make up a new interpolation process. So, for instance, the authors combine several groups of circles, where circles of different groups need not to be concentric.