Set evincing the ranks with respect to an embedded variety (symmetric tensor rank and tensor rank) (Q1634408): Difference between revisions

Let \(X\) be a non-degenerate projective variety in \(\mathbb P^n\). For any point \(P\in \mathbb P^n\), the \(X\)-rank \(r_X(P)\) of \(P\) is the minimal integer \(r\) such that \(P\) sits in the linear span of a finite subset \(S\subset X\) of cardinality \(r\). In this case, one says that \textit{\(S\) evinces the \(X\)-rank of \(P\)}. The author considers the inverse point of view: given a finite subset \(S\subset X\), determine for which points \(P\) the set \(S\) evinces the \(X\)-rank of \(P\). Of course \(S\) does not evinces the \(X\)-rank of \(P\) if \(P\) lies in the linear span of a proper subset of \(S\). The author says that \(S\) \textit{totally evinces the \(X\)-rank} if \(S\) evinces the rank of all the points in its span, except those contained in the span of a proper subset. The general philosophy about sets evincing the rank is that, as soon as the cardinality of \(S\) is sufficiently small and \(S\) is sufficiently general, then \(S\) totally evinces the \(X\)-rank. The authors gives a contribution in this direction. First, if \(u(X)\) denotes the maximal integer such that each subset of \(X\) with cardinality \(u(X)\) is linearly independent, then the author shows that all the subsets of cardinality \(c\leq (u(X)+1)/2\) totally evince the \(X\)-rank. Examples prove that the bound for \(c\) is sharp. Turning to the case where \(X\) is the image of \(\mathbb P^a\) under the Veronese map \(\nu\) of degree \(d\), \(d>a>2\), the author proves that if \(S=\nu(A)\) and \(h^1 (I_A(\lfloor d/2\rfloor,\mathbb P^a))=0\), then \(S\) totally evinces the \(X\)-rank.