Waring loci and the Strassen conjecture (Q2357481)

A Waring decomposition of a homogeneous polynomial \(F\) of degree \(d\) is a minimal presentation of \(F\) as a sum \(L_1^d+\dots +L_k^d\) where the \(L_i\)'s are linear forms. In many relevant cases, e.g. for monomials, the Waring decomposition is not unique. The authors define the \textit{Waring locus} of \(F\) as the locus \(W_F\) of linear forms \(L\) that appear in a minimal Waring decomposition of \(F\). The complement \(\mathcal F_F\) of the Waring locus is called the \textit{forbidden locus} of \(F\). Several properties concerning Waring decompositions are linked to the geometry of the Waring locus, which is a quasi-projective subvariety of the dual projective space. The authors use the apolarity construction to compute the Waring locus in some relevant examples. The most interesting cases that they are able to treat are monomials (which, usually, have a high Waring rank, hence a big Waring locus), binary forms, and smooth plane cubics. Then, the authors apply their analysis to prove special cases of the symmetric Strassen conjecture, which predicts that a sum of two forms \(F=F_1+F_2\) in which the \(F_i\)'s are defined in different sets of variables, has Waring rank equal to rank\((F_1)+\) rank\((F_2)\). Strassen's conjecture is quite natural, since it essentially predicts that a tensor which splits in blocks has rank equal to the sum of the ranks of the blocks. Up to now, however, Strassen's conjecture is proven only for a restricted number of special cases, even for symmetric tensors, i.e. for forms. The authors increase the number of cases in which the symmetric Strassen's conjecture is settled. Indeed, they prove that the conjecture holds in several cases in which the Waring loci \(W_{F_i}\) can be computed. In particular, for instance, they prove that the conjecture holds when both \(F_1,F_2\) are monomials, in which all the exponents are greater than \(1\).