On the foundation of bases of spline spaces (Q1576467)

The problem of finding a least founded basis and a basis formed by minimally founded functions for univariate tensor product spaces is investigated. The concept of foundation is a sharper concept than that of support. It was introduced by \textit{J. M. Carnicer}, \textit{T. N. T. Goodman} and \textit{J. M. Peña} [Ann. Numer. Math. 4, No. 1-4, 161-180 (1997; Zbl 0893.41004)]. A function \(f: \Omega\to \mathbb{R}\) is minimally founded if there is no other function \(g\neq 0\) with smaller foundation than that of \(f\). The following theorem is one of the main results. Theorem: Let \(U\) be a vector space of univariate functions which has a locally linearly independent basis \((u_0,\dots, u_n)\) satisfying the B-condition. Namely, \(\text{supp }u_i= [a_i, b_i]\), and \[ \lim_{t\to a^+_i} {u_j(t)\over u_i(t)}=0\quad\text{and} \quad \lim_{t\to b^-_j} {u_i(t)\over u_j(t)}= 0, \] where \(i< j\) and \(a_i\leq a_j\), \(b_i\leq b_j\). Then the basis functions are minimally founded and this basis is the unique least founded basis of \(U\) (up to permutation and scaling). A section is devoted to study a bivariate space whose spline functions are \({\mathcal C}^{k-1}\) piecewise polynomials of degree \(k\) defined on angular regions.