A proof of the set-theoretic version of the salmon conjecture (Q713410)

It is a much studied problem to determine equations defining higher secant varieties of Segre varieties (i.e., Segre embeddings of multiprojective spaces \({\mathbb P}^{n_1} \times \dots \times {\mathbb P}^{n_t}\)); actually those varieties parameterize \((n_1+1)\times \dots \times (n_t+1)\)-tensors of given border rank, hence to know their defining equations in \({\mathbb P}^{N}\), \(N=\prod_{i=1}^t (n_i+1) -1\), would give a way to compute the border rank of any given tensor. In particular, a conjecture has been stated for \(4\times 4 \times 4\)-tensors which gives for the 4-secant variety of the Segre variety \({\mathbb P}^3 \times {\mathbb P}^3 \times {\mathbb P}^{3}\) that its ideal is generated in degrees \(5\), \(6\) and \(9\), where the generating polynomials are given, respectively, by the Strassen equations, the Landsberg-Manivel ones and the symmetization conditions. This conjecture, stated by E. S. Allman and J. A. Rhodes, it is called ``the salmon conjecture'', since a freshly smoked salmon was jokingly offered by its authors as the prize for resolving it. In this paper the conjecture is proved set-theoretically, i.e. it is proved that the prescribed equations of degrees \(5\), \(6\) and \(9\) have the \(4\)-th secant variety of the Segre \({\mathbb P}^3 \times {\mathbb P}^3 \times {\mathbb P}^{3}\) as their common zero locus. In this settings, set-theoretic results are more relevant with respect to what used to be in classical Algebraic Geometry, since they give a set of equation which are an algorithm to value the border rank of any given \(4\times 4 \times 4\)-tensor, even if we don't know whether such set is ``optimal'' for the variety in question.