Error estimates in \(W^{2,\infty}\)-semi-norms for discrete interpolating \(D^2\)-splines (Q2574175)

Let \(\Omega\) denote a polyhedral subset of the Euclidean vector space \({\mathbb R}^n\) and let the standard Sobolev space \(W^{m,p}(\Omega)\) be endowed with the spectrum of seminorms \(| \cdot| _{\ell,p,\Omega}\) where \(1 \leq \ell \leq m\) and \(p \in [1,+\infty]\). For \(h > 0\) and a triangulation of the closure \(\bar{\Omega}\) made up of \(n\)-dimensional simplices \(K\) of diameter \(| K| \leq h\), the main result of the present paper is to establish the existence of a multivariate discrete spline \(\sigma_h\) of minimal energy interpolating a given real-valued function \(f \in W^{2,\infty}(\Omega)\) on the set of vertices of the triangulation such that the estimate of the deviation \(| f - \sigma_h| _{\ell,\infty,\Omega} \leq C.h^{(2 - \ell)}.| f| _{2,\infty,\Omega}\) holds for \(\ell \in \{0,1,2\}\) and an appropriate constant \(C > 0\) [cf. \textit{P. G. Ciarlet}, The finite element method for elliptic problems. Classics in Applied Mathematics. 40 (Philadelphia) (2002; Zbl 0999.65129); \textit{R. Arcangéli, M. C. López de Silanes}, and \textit{J. J. Torrens}, Multidimensional minimizing splines. Kluwer Academic Publishers (Boston) (2004; Zbl 1064.41001)]. Based on this result and the Bramble-Hilbert lemma, operators \(\Pi_h\) are constructed providing a piecewise polynomial approximation \(\Pi_h (v)\) of optimal approximation order for the elements \(v \in W^{m,p}(\Omega)\) so that the estimate \(\Bigl(\sum_{\mid K \mid \leq h}| v - \Pi_h (v)| ^p_{\ell,p,K}\Bigr)^{1 \over p} \leq C.h^{(m - \ell)}.| v| _{m,p,\Omega}\) holds for \(0 \leq \ell \leq m, p \in [1,+\infty[\), and an appropriate constant \(C > 0\) [\textit{V. Girault, L. R. Scott}, Math. Comput. 71, 1043--1074 (2002; Zbl 1002.65129)]. Based on the Lax-Milgram lemma, a similar estimate of the deviation has been established for \(p = \infty\) and for projectors \({\mathcal P}_h\) linearizing the operators \(\Pi_h\) for \(h > 0\) [\textit{J. Douglas, jun., T. Dupont}, and \textit{L. Wahlbin}, Numer. Math. 23, 193--197 (1975; Zbl 0297.41022)]. Some numerical tests support the results.