Varieties of completely decomposable forms and their secants (Q2253008)

The author considers polynomials over the complex field, and studies their decomposition as a sum of \textit{splitting forms}, i.e. forms that can be written as a product of linear forms \(L_1^{d_1}\cdots L_m^{d_m}\) (notice that the linear forms are not fixed, in the summands of the decomposition). The main question addressed is: for which \(s:= s(n,d)\) one can decompose a generic polynomial of degree \(d\) in \(n+1\) variables as a sum of \(s\) splitting forms? Since the problem is invariant up to scaling, it can be rephrased in term of projective geometry. Splitting forms determine a subvariety \(\mathrm{Split}_d(\mathbb P^n)\) of the projective space \(\mathbb P^N\) of forms of degree \(d\) in \(n+1\) variables. Thus, the question can be translated in the study of the dimension of secant varieties \(\sigma_k\) to the variety \(\mathrm{Split}_d(\mathbb P^n)\), with the target of finding the minimum \(s\) such that \(\sigma_s\) is equal to the whole space \(\mathbb P^N\). The author determines upper and lower bounds on \(k\) which imply that \(\sigma_k\) has the expected dimension. As these bounds match for \(n=2\), the author completely solves the problem of finding \(s(2,d)\), for polynomials in \(3\) variables. The proofs are obtained by a degeneration argument, using induction on the degree \(d\). For higher values of \(n\), the authors lists examples of secant varieties \(\sigma_k\) which do not reach the expected dimension. Indeed for \(n>2\) the upper and lower bounds provided in the paper do not match. For \(n=3\), the difference grows just linearly with \(d\), thus only few cases of defective secant varieties are allowed.