A note on the construction of \(L\)-fold sparse tensor product spaces (Q378147)

An anisotropic Sobolev space on the tensor product \(\Omega\) of domaines \(\Omega_i\subset \mathbb{R}^{n_i}\) with \(n_i\in \mathbb{N}\), \(i=1,2,\dots,L\), can be defined by \[ H^s(\Omega):=H^{s_1}(\Omega_1)\otimes H^{s_2}(\Omega_2)\otimes\cdots\otimes H^{s_L}(\Omega_L), \quad s=(s_1,s_1,\dots, s_L). \] In order to approximate functions in \(H^s(\Omega)\), the authors assume to have for each subdomain \(\Omega_i\) the individual multiscale analysis \[ V_0^{(i)}\subset V_1^{(i)}\subset V_2^{(i)}\subset\cdots\subset L^2(\Omega_i), \quad i=1,2,\dots,L, \] with associated complementary spaces \(W_j^{(i)}\) \((W_j^{(i)} \oplus V_{j-1}^{(i)} = V_j^{(i)}, \;\;W_j^{(i)}\cap V_{j-1}^{(i)} =\{0\} )\). From this, the spacial sparse tensor product space is defined according to \[ \widehat{V}^{\mathbf{\alpha}}_J:=\bigoplus_{\mathbf{\alpha}^T {\mathbf{j}}\,\leq J}W_{j_1}^{(1)}\otimes W_{j_2}^{(2)}\otimes\cdots\otimes W_{j_L}^{(L)} \] for an arbitrary vector \(\mathbf{\alpha} = (\alpha_1,\alpha_2,\dots,\alpha_L) > \mathbf{0}\) and \(\mathbf{j}=(j_1,j_2,\dots, j_L) \in \mathbb{N}_0^L\). For the case \(L=2\), it is shown in [Math. Comput. 82, No. 282, 975--994 (2013; Zbl 1267.41012)] that the most efficient construction of sparse tensor product spaces is obtained if the spatial dimension of the underlying domains or the polynomial exactness (and thus the approximation power) of the ansatz spaces differ. In this paper these results are extended to arbitrary \(L\)-fold sparse tensor product spaces.