Algebraic notes on testing sets for lower and upper grids (Q6953032)

In a broad sense, Lagrange interpolation consists of the following. Let \(k\) be a field, \(\mathcal{A}\) be an arbitrary set and \(\mathcal{P}\) be a finite-dimensional vector subspace of the functional space \(k^{\mathcal{A}}\). Then, given a subset \(\mathcal{N}\) of elements of \(\mathcal{A}\), called nodes, and some data in each node (a mapping \(y: \mathcal{N} \rightarrow k\)), we may ask for a function \(f \in \mathcal{P}\) that agrees with \(y \in \mathcal{N}\). We say that the pair \((\mathcal{P}, \mathcal{N})\) is \textit{poised for interpolation} whenever any data \(y \in k^{\mathcal{N}}\) can be interpolated by a unique function of \(\mathcal{P}\) (see, for example, [\textit{M. Gasca} and \textit{T. Sauer}, Adv. Comput. Math. 12, No. 4, 377--410 (2000; Zbl 0943.41001)]). \N\NIn the paper under review, the following case is considered. The set \(\mathcal{A}\) is the \(n\)-dimensional affine space \(k^n\), the space \(\mathcal{P}\) consists of polynomial functions generated by polynomials in \(k[ x_1, \cdots, x_n]\) and the nodes for interpolation are selected from a cartesian product \(\mathcal{X} = \mathcal{X}_1 \times \cdots \times \mathcal{X}_n\), where \(\mathcal{X}_i = \{ x_{i0}, x_{i1}, \ldots, x_{ij}, \ldots \}\) are sets of independent variables \(x_{ij}\), which can later be replaced by elements of \(k\).\N\NMore precisely, let \(\mathcal{C}\) be a \textit{grid} (i.e., a finite subset of \(\mathbb{N}_0^n\)) and define \(\mathcal{X_C} := \{ (x_{1 \tau_1} , \ldots , x_{n \tau_n}) : (\tau_1, \cdots , \tau_n) \in \mathcal{C} \}\). Let \(\mathcal{B}\) be a finite set of monomials \(x^{\alpha} = x^{\alpha_1}_1 x^{\alpha_2}_2 \cdots x^{\alpha_n}_n\), where \(\alpha = (\alpha_1, \cdots, \alpha_n) \in \mathbb{N}_0^n\). Such a set can be identified with a grid via the correspondence \(x^\alpha \mapsto \alpha\). Evaluating a monomial \(x^{\alpha} \in \mathcal{B}\) at a node \((x_{1 \tau_1} , \ldots , x_{n \tau_n}) \in \mathcal{X_\mathcal{C}}\) produces an element \(x^{\alpha_1}_{1\tau_1} \cdots x^{\alpha_n}_{n \tau_n}\) of the field of fractions \(\mathbb{K} = k (\{ x_{ij} \})\) and we can define the evaluation map\N\[\N\mathrm{Ev}_{\mathcal{B,C}} : \mathfrak{P}_\mathcal{B} \longrightarrow \mathbb{K}^{\mathcal{X_C}},\N\]\Nwhere \(\mathfrak{P}_\mathcal{B}\) is the \(\mathbb{K}\)-subspace of \(\mathbb{K}[x_1, \ldots, x_n]\) generated by \(\mathcal{B}\). We say that \(\mathcal{B}\) is a \(\sigma\)-lower (resp., \(\sigma\)-upper) grid if it is a union of intervals of \(\mathbb{N}^n_0\) with common origin \(\sigma\) (resp., endpoint \(\sigma\)). When \(\mathcal{B} = \mathcal{C}\) and \(\mathcal{B}\) is a \(\sigma\)-lower (resp., \(\sigma\)-upper) grid, the author provides (Theorem 9) explicit formulas for \(\det \mathbb{V}[\mathcal{B}]\), the determinant of \(\mathrm{Ev}_{\mathcal{B,B}}\), and proves (Corollary 10) that if \(\mathcal{B} \subseteq \mathcal{B}'\) are \(\sigma\)-lower grids, then\N\[\N\det \mathbb{V}[\mathcal{B}] \mid \det \mathbb{V}[\mathcal{B}'].\N\]\NAdditionally, the author proves (Theorem 5) that the pairs \((\mathfrak{P}_\mathcal{B}, \mathcal{X_B})\) are poised for interpolation for any grid \(\mathcal{B}\).\N\NThe paper concludes (Section 5) with a more detailed study of the case in which \(k\) is a finite field, and relevant examples for coding theory (e.g., generalized Reed-Muller codes) are given. The author points out that this paper was inspired by the notion of \textit{non-cubic} subsets of a finite dimensional \(\mathbb{F}_2\)-vector space used by \textit{R. Livné} [Contemp. Math. 67, 247--261 (1987; Zbl 0621.14019)] in a criterion to prove that the semi-simplifications of two 2-adic Galois representations are isomorphic. Therefore, the results of this paper could also be useful in representation theory.