Explicit error estimates for spline approximation of arbitrary smoothness in isogeometric analysis (Q2309514)

Isogeometric analysis [\textit{T. J. R. Hughes} et al., Comput. Methods Appl. Mech. Eng. 194, No. 39--41, 4135--4195 (2005; Zbl 1151.74419)] is a generalization of the classical finite element analysis, and it advocates for the use of splines as finite elements. This has led to a renewed interest in using splines for approximation, and numerical evidence has shown the benefits of their smoothness -- smooth spline spaces demonstrate better approximation behaviour per degree of freedom than less smooth spline spaces. The authors of the present manuscript provide a priori error estimates that, in particular, help to describe the afore-mentioned numerical evidence. The error estimates are for \(L^2\) and Ritz projections onto spline spaces of arbitrary smoothness defined on non-uniform meshes, and they feature constants that depend explicitly on the spline space parameters -- mesh size \(h\), polynomial degree \(p\) and smoothness \(k\). The key to their results is the description of the considered Sobolev spaces and spline spaces in terms of integral operators what allows them to derive explicit constants in spline approximation from those in polynomial approximation. The error estimates are provided for univariate spline spaces, reduced univariate spline spaces (i.e., spline spaces with zero end-point even or odd derivatives) and multivariate tensor-product spline spaces. The multivariate results cover approximation on the spline parametric domain as well as mapped single-patch and \(C^0\) multi-patch geometries; the role of the geometric map is explicitized in the constants. Finally, the error estimates improve upon the ones presented earlier in [\textit{L. Beirão da Veiga} et al., Numer. Math. 118, No. 2, 271--305 (2011; Zbl 1222.41010); \textit{S. Takacs} and \textit{T. Takacs}, Math. Models Methods Appl. Sci. 26, No. 7, 1411--1445 (2016; Zbl 1339.41012)] and [\textit{E. Sande} et al., Math. Models Methods Appl. Sci. 29, No. 6, 1175--1205 (2019; Zbl 1428.41010)].