Secant varieties of the varieties of reducible hypersurfaces in \(\mathbb{P}^n\) (Q2414550)

The authors consider varieties of reducible hypersurfaces in \(\mathbb{P}^{N-1}\) and the varieties of their secants. They prove a series of results and state some conjectures (supported by proved cases and numerical tests). Below there is a short overview of the work. The variety \(X_{n-1,\lambda}\) of degree \(d\) in \( \mathbb P^{N-1}\) with \(N = \binom{d+n-1}{n-1}\) is given as \( X_{n-1,\lambda}:= \{ [F] \in \mathbb P^{N-1} \ | \ F = F_1 \cdots F_r \text{ for some }0\neq F_i \in [S]_{d_i} \},\) for a partition \(\lambda=[d_1,\dots, d_r]\) of \(d\), i.e. \(d=\sum_{i=1}^rd_i\). By \(\sigma_l (X_{n-1,\lambda})\) the authors denote the \((l-1)\)-secant variety of \(X_{n-1,\lambda}\). The main object of the paper is to study the dimension of \(\sigma_l (X_{n-1,\lambda})\). The expected dimension of this variety is \[ \operatorname{exp.dim} \sigma_{l} (X_{n-1,\lambda}):=\min \left \{ \binom{d+n-1}{n-1} -1, \; l \cdot \dim X_{n-1,\lambda} + l-1 \right \}. \] If the actual dimension of \(\sigma_l (X_{n-1,\lambda})\) is less than expected the variety \(\sigma_l \) is called defective, and one of the aims of the paper is to find the defective cases and compute the defect. The authors prove the following theorem: Let \(\lambda = [d_1,\ldots,d_r]\) be a partition of \(d\) with \(r \ge 2\) parts and let \(s=d_2+\cdots+d_r\). Then, if \(d_1 < s\) (and hence \(r\geq3\)) and \(2l \le n\), then \(\sigma_{l} (X_{n-1, \lambda})\) is not defective. If \(d_1 \ge s\) and \(2l \le n\), then \(\sigma_{l} (X_{n-1, \lambda})\) is defective if and only if it does not fill its ambient space. Finally, if \(l\geq n\) and \(d_1\geq(n-1)(s-1)\), then \(\sigma_l(X_{n-1,\lambda})\) always fills its ambient space, while if \(2l \le n\), then \(\sigma_l(X_{n-1,\lambda})\) fills its ambient space if and only if either \(n = 4\), \(l = 2\), and \(\lambda \in \{[1,1], [2,1], [1,1,1]\}\) or \(n = 2 l \ge 6\) and \(\lambda = [1,1]\). The authors prove also: If \(2l \le n\), then \( \dim \sigma_{l} (X_{n-1, d}) = \dim \sigma_{l} (X_{n-1, [d-1, 1]}). \) Moreover, \(\sigma_{l} (X_{n-1, d})\) is defective if and only if it does not fill its ambient space. In the paper the authors state some conjectures. The main one says that for \(\lambda = [d_1,\ldots,d_r]\), a partition of \(d\) with \(r \ge 2\) parts, and for some integers \(a_j \) depending on \(l, n,\text{ and } \lambda\) (the definition of \(a_j\) is here omitted, as rather lenghty) we have: 1. The secant variety \(\sigma_{l} (X_{n-1,\lambda})\) fills its ambient space if and only if there is some integer \(j\) with \(s = d_2 + \cdots + d_r \le j \le d\) such that \(a_j (l, n, \lambda) \le 0\). 2. If \(\sigma_{l} (X_{n-1,\lambda})\) does not fill its ambient space, then it has dimension \begin{multline*} \dim \sigma_{l} (X_{n-1, \lambda}) = \ l \cdot \dim X_{n-1, \lambda} + l -1 \\ - \sum_{k=2}^l (-1)^k \binom{l}{k} \binom{d_1 - (k-1) (d_2 + \cdots + d_r) + n-1}{n-1} \\ - \binom{l}{2} \binom{2 d_2 - d + n-1}{n-1} - l (l -1) \binom{d_1 + d_2 - d+ n-1}{n-1}. \end{multline*} The authors prove the conjecture in some cases: Let \(\lambda=[d_1,\ldots,d_r]\) be a partition of \(d=d_1+s\) into \(r\geq 2\) parts, where \(s=d_2+\cdots+d_r\). Then the main conjecture is true in the following cases: if \(l \le \frac{n}{2}\) or \(l\geq\binom{s+n-1}{n-1}\); or if \(r = 2\) and either \(l \le \frac{n+1}{2}\), or \(\lambda = [d-1, 1]\), or \(n=3\); or if \(r\geq3\) and \(n\leq \ell\leq 1+\frac{d_1+n-1}{s}\). Another conjecture stated by the authors is as follows. For \(\lambda=[d_1,\ldots,d_r]\), a partition of \(d=d_1+s\) into \(r\geq 2\) parts, where \(s=d_2+\cdots+d_r\), if \(d_1 < d_2 + \cdots + d_r\) (and thus \(r \ge 3\)), then \(\sigma_{l} (X_{n-1, \lambda})\) is not defective or if \(d_1 \ge d_2 + \cdots + d_r\), then the secant variety \(\sigma_{l} (X_{n-1, \lambda})\) is defective if and only if it does not fill its ambient space. The authors show also the connection between the computation of the dimension of \(\sigma_{l} (X_{n-1, \lambda})\) for arbitrarily large \(l\), corresponding to improper intersections, and the Lefschetz Properties. Let us recall that for \(A = \mathbb{K}[x_1,\dots,x_n]/J\) an artinian graded \(\mathbb{K}\)-algebra \(A\) is said to have the Weak Lefschetz Property if, for each integer \(i\), multiplication by a general linear form \(L \in A\) from \([A]_i\) to \([A]_{i+1}\) has maximal rank. Then \(L\) is called a Lefschetz element of \(A\). The authors conjecture that the coordinate ring of a certain join variety has enough Lefschetz elements if \(2l > n\). The authors show that if this conjecture is true, then the main conjecture follows. The authors consider also separately the case \(r =2\) and show that main conjecture is a consequence of Fröberg's Conjecture on the Hilbert function of ideals generated by generic forms. The authors study also the variety of reducible forms. Let \(X_{n-1, d} = \bigcup_{k=1}^{\lfloor \frac{d}{2} \rfloor} X_{n-1, [d-k, k]}\) be the variety of reducible forms of degree \(d\) in \(n\) variables. They prove that for \(2l\le n\), \( \dim \sigma_{\ell} (X_{n-1, d}) = \dim \sigma_{l} (X_{n-1, [d - 1, 1]}), \) and moreover firstly, the variety \(\sigma_{l} (X_{n-1, d})\) fills its ambient space if and only if \(2l = n\) and \(d = 2\) or \(l = 2, \ n = 4\), and \(d = 3\), and secondly, if \(\sigma_{l} (X_{n-1, d})\) does not fill its ambient space, then its dimension is \( \dim \sigma_{l} (X_{n-1, d}) = \binom{d+n-1}{n-1} - \binom{d+n- l - 1}{d} + \ell (n- l) - 1, \) and \(\sigma_{l} (X_{n-1, d})\) is defective. The authors also use their results to prove some conjectures stated in the work of Abo, Ottaviani and Petersen [\textit{H. Abo} et al., Trans. Am. Math. Soc. 361, No. 2, 767--792 (2009; Zbl 1170.14036)] about defectivity of Segre Varieties.