Some examples of forms of high rank (Q314649): Difference between revisions
From MaRDI portal
Changed an Item |
Added link to MaRDI item. |
||
links / mardi / name | links / mardi / name | ||
Revision as of 02:25, 30 January 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Some examples of forms of high rank |
scientific article |
Statements
Some examples of forms of high rank (English)
0 references
16 September 2016
0 references
In the present paper, the authors provide new examples of forms with greater Waring rank than previously known examples. Let us recall that for a complex homogeneous form \(F\) of degree \(d\), the Waring rank, denoted by \(r(F)\), is the least \(r\) such that there exist linear forms \(l_{1},\dots, l_{r}\) and \(c_{1}, \dots,c_{r} \in \mathbb{C}\) satisfying \(F = c_{1}l_{1}^{d} +\dots + c_{r}l_{r}^{d}\). The famous Alexander-Hirschowitz theorem states that a general form \(F\) of degree \(d > 1\) in \(n\) variables has rank \(r(F)\) equal to \[ \bigg\lceil \frac{1}{n} { n+d-1 \choose n-1 } \bigg\rceil \] except some cases. The above value is called the generic rank and it is denoted by \(r_{\mathrm{gen}}(n,d)\). It is an open problem what is the maximum Waring rank \(r_{\max}(d,n)\) of forms of degree \(d\) in \(n\) variables for each \((n,d)\). Of course \(r_{\max}(n,d) \geq r_{\mathrm{gen}}(d,n)\) and, on the other hand, by the recent result due to \textit{G. Blekherman} and \textit{Z. Teitler} [Math. Ann. 362, No. 3--4, 1021--1031 (2015; Zbl 1326.15034)] one has \(r_{\max}(n,d) \leq 2r_{\mathrm{gen}}(n,d)\). One might ask whether the difference between the maximum Waring rank and the generic rank is unbounded, but it is even not known whether this difference is positive, i.e., the maximum Waring rank is strictly greater than the generic rank. The aim of the paper is to give a lower bound for Waring rank and some new examples of forms in \(3\), \(4\), and \(5\) variables whose Waring ranks are strictly greater than previously known examples. The first theorem can be formulated as follows. Theorem 1. Let \(d\geq 3\) be odd. There exist forms of degree \(d\) in \(n=3\) variables of rank strictly greater than \(((d+1)/2)^{2}\), the maximum rank of a monomial \(r_{\max}(3,d) > ((d+1)/2)^{2}\). It is worth pointing out that previously the greatest Waring rank of a form in \(3\) variables was obtained by monomials and this rank is equal to \(((d+1)/2)^{2}\) with \(d\) odd. Another result is devoted to forms in \(4\) variables. Theorem 2. Let \(d \geq 3\) odd. There exist forms of degree \(d\) in \(n=4\) variables of rank strictly greater than the generic rank: \(r_{\max}(4,d) > r_{\mathrm{gen}}(4,d)\). It is known by a result due to \textit{A. De Paris} [Int. J. Algebra Comput. 25, No. 4, 607--631 (2015; Zbl 1322.11105)] that for forms of degree \(d = 5\) and \(n=3\) variables the maximum Waring rank is either \(9\) or \(10\). The last main result of the paper shows that in fact the maximum Waring rank is \(10\). Theorem 3. There exists a ternary quintic form (i.e., \((n,d) = (3,5)\)) of rank \(10\). More explicitly, \(F = xyz^{3} + zy^{4}\) has \(r(F) = 10\).
0 references
Waring rank
0 references
apolarity
0 references
power sum decompositions
0 references