Apolarity for determinants and permanents of generic matrices (Q2259545)

This paper proves two theorems about the degree, generators, and number of generators of the apolarity ideal for the determinant and the permanent of a generic matrix. In fact explicit generators are given. There are also results for the Pfaffian. The main new application is a lower bound for the cactus rank of the generic determinant and generic permanent. Recall that the (\(r\)-th) cactus variety is the closure of the union of the linear spaces spanned by the rank \(r\) zero dimensional subschemes of a variety. The paper begins with a nice summary of apolarity including a lemma (1.3) which strikes the reviewer as interesting and probably related to the paper of \textit{J. Sidman} and \textit{S. Sullivant} [Can. J. Math. 61, No. 4, 930--949 (2009; Zbl 1174.13042)] on prolongations. To frame the results of this paper informally, one has to define a number of degree 2 differential operators. Firstly if \(D\) is any 2 by 2 matrix of differential operators we can form it's permanent \(P_D\) and it's determinant \(M_D\) (author's notation). A monomial is acceptable if it is square free and doesn't contain two variable from the same row or same column. A monomial (of differential operators) is `unacceptable' if it doesn't divide any term of the determinant of the generic \(n\) by \(n\) differential operator (and hence will annihilate any subdeterminant of the generic differential matrix). One defines the vector spaces of determinantal, permanental, and unacceptable differential operators by addition and scalar multiplication. An interesting preliminary result is that every acceptable polynomial (differential operator) is the sum of permanents and determinants. The two main results are theorem 2.12 and theorem 2.13. The first states that the apolarity idea of the generic determinantal matrix is generated by the vector space of unacceptable and permanent differential operators and is generated degree 2. The generic permanent matrix has the same result except that the determinant differential operators replace the permanent differential operators in the result. Finally, applying a result of \textit{K. Ranestad} and \textit{F.-O. Schreyer} [J. Algebra 346, No. 1, 340--342 (2011; Zbl 1277.13016)] this implies a lower bound for the cactus rank of either the generic determinantal or permanental matrix, namely, \(\frac{1}{2}\binom{2n}{n} 2 n\).