Catalecticant intersections and confinement of decompositions of forms (Q820956)

The authors propose a new method for finding the possible decompositions of a tensor, based on the notion of confinement and on catalecticant maps. The possible decompositions of tensors of sub-generic rank lie into special sub-varieties, called contact loci, of the variety \(X\) of rank-\(1\) tensors. However, for tensors of generic rank, such sub-varieties are meaningless and new methods are needed. For a symmetric tensor \(f \in \mathrm{Sym}^d\mathbb C^{n+1}\), the image of the catalecticant map \(C_f: \mathrm{Sym}^{h}(\mathbb C^{n+1})^\vee \rightarrow \mathrm{Sym}^{d-h}\mathbb C^{n+1}\) lies into the span \(\Lambda\) of the \((d-h)\)-th powers of linear forms in which \(f\) decomposes, thus the sub-variety confining the possible decompositions of \(f\) is the intersection \(Y:=\Lambda \cap X^n_{d-h}\) of \(\Lambda\) with the \((d-h)\)-th Veronese variety \(X^n_{d-h}\). The authors extend the catalecticant method and apply it to study the simultaneous Waring decompositions of a given number of forms, or equivalently the decompositions of vectors of forms. The main working setting is the so-called perfect case: given \(X^n_{a_1,\ldots, a_r}=\{(l^{a_1}, \ldots , l^{a_r}) \ | \ l \in \mathbb P^n\}\) the variety of rank-\(1\) polynomial vectors in \(\mathbb P(\mathrm{Sym}^{a_1}\mathbb C^{n+1} \oplus \ldots \oplus \mathrm{Sym}^{a_r}\mathbb C^{n+1})\simeq \mathbb P^{N-1}\), \(f\) is assumed to belong to the \(k\)-th secant variety \(S^k(X^n_{a_1, \ldots , a_r})\) of expected dimension \(N=k(\dim X^n_{a_1,\ldots, a_r})+1\). The set of parameters defining the setting is denoted by \((n,r;a_1,\ldots, a_r)\). In Sections \(3\) and \(4\) the authors analyze the perfect case with parameters \((2,3;3,3,3)\): through a catalecticant approach they show that the confinement for a general \(f \in (\mathrm{Sym}^3\mathbb C^3)^{\oplus 3}\) is a sextic elliptic normal curve (Proposition 4.1), and as a consequence such \(f\) admits two simultaneous Waring decompositions with \(k=6\) summands (Theorem 4.4). In Sections \(5\) and \(6\) the case \((2,4;4,4,4,4)\) is studied: the catalecticant approach for a general \(f \in (\mathrm{Sym}^4\mathbb C^3)^{\oplus 4}\) leads to a general threefold section as confinement (Proposition 6.1). As a consequence, it is computationally obtained that a such general \(f\) admits \(18\) simultaneous Waring decompositions with \(k=10\) summands (Theorem* 6.5), but this number is theoretically proved to be just a lower bound (Theorem 6.9).