The minimal generating space of a polynomial (Q792386)

Let L be the space of all \({\mathbb{R}}\)-linear mappings on \({\mathbb{R}}^ n\) and \({\mathbb{R}}[H]\) the subalgebra of \({\mathbb{R}}[X_ 1,...,X_ n]\) generated by \(H\subseteq L\), where \(X_ 1,...,X_ n\) are the projections \({\mathbb{R}}^ n\to {\mathbb{R}}.\) Then \({\mathbb{R}}[L]={\mathbb{R}}[X_ 1,...,X_ n].\) For each polynomial function \(P\in {\mathbb{R}}[L]\) let \(G_ P\) denote the subspace of L which is generated by P and the partial derivatives of P. - If G is a subspace of L it is proved that \(P\in {\mathbb{R}}[G]\), and G is minimal with respect to this property if and only if \(G=G_ P\). Therefore the minimal number of projections in L necessary to generate P is equal to \(\dim(G_ P)\). By a geometrical characterization of \(G_ P\) a generating set for \(G_ P\) is determined, which gives a simple way for the evaluation of \(G_ P.\) An isomorphism \(T: L\to L\) is said to separate two polynomial functions \(P(X_ 1,...,X_ n)\) and \(Q(X_ 1,...,X_ n)\) if P and Q can be written as functions \(P^+(Y_ 1,...,Y_ k)\) and \(Q^+(Y_{k+1},...,Y_ n)\) respectively, where \(Y_ i=T X_ i .\) By the help of the spaces \(G_ P\) and \(G_ Q\) all pairs (P,Q) are characterized for which there exists such a separating isomorphism.