On stable local bases for bivariate polynomial spline spaces (Q5956460)

For a regular triangulation \(\Delta\) of a polygonal set \(\Omega \subset\mathbb{R}^2\), let \(S_d^r(\Delta): =\{s\in C^r(\Omega): s\mid T\in {\mathcal P}_d\) for all triangles \(T\in \Delta\}\) denote the bivariate polynomial spline space. Here \({\mathcal P}_d\) means the set of all polynomials of degree \(\leq d\). The authors also consider superspline subspaces \(S_d^{r,\rho} (\Delta):= \{s\in S^r_d (\Delta): s\in C^{\rho_v} (v)\) for all \(v\in V\}\) with \(\rho:= \{\rho_v:v \in V\}\) where \(\rho_v\in \mathbb{Z}\cap [r,d]\) are given and \(V\) is the set of all vertices of \(\Delta\). In this interesting paper, the authors describe algorithms for constructing locally supported bases \(\{B_i:i\in I\}\) for these spaces which are stable in the following sense: \[ K_1\|c\|_\infty \leq\Bigl\|\sum_{i \in I}c_iB_i \Bigr\|_\infty\leq K_2\|c\|_\infty \] for arbitrary coefficient vectors \(c=(c_i)_{i\in I}\). The constants \(K_1\) and \(K_2\) depend only on \(d\) and the smallest angle in the triangulation \(\Delta\). If \(r\geq 1\) and \(d\geq 3r+2\), then stable locally supported bases are constructed for the spline space \(S_d^r(\Delta)\) as well as for various superspline spaces. The construction is based on Bernstein-Bézier technique. In addition, it is shown that for given \(r\geq 1\) and \(d\geq 3r+2\), there is a triangulation \(\Delta\) such that no basis of \(S_d^r(\Delta)\) is simultaneously stable and locally linearly independent.