Direct and inverse finite difference projection methods (Q1907449)

The quality of a discretization depends on the choice of the space of approximations, and vice versa, in some sense. For \(\varphi\in W^m_2 (E_n)\) the author defines the subspaces \(B_h (\Lambda, \varphi)\) consisting of all dilates/translates \(\varphi^h_\alpha (\cdot)= \varphi (\cdot/ h-\alpha)\), \(\alpha\in \Lambda\), with respect to some lattices \(\Lambda\). Then the approximability of the elements of \(W^m_2 (E_n)\) by elements of \(B_h (\Lambda, \varphi)\) can be characterized by the behaviour of \(\widehat {\varphi}\) in the points of \(2\pi \Lambda^* \setminus \{0\}\), \(\Lambda^*\) conjugated lattice of \(\Lambda\). Moreover, if \(P(D)\) is a differential operator of order \(2m\) which satisfies some positivity conditions, then it can be replaced by certain difference projection operators, where consistency occurs if and only if \(\widehat {\varphi}\) vanishes in the points of \(2\pi \Lambda^* \setminus \{0\}\) of order \(m+1\).