The higher transvectants are redundant (Q1041264)

Transvection is a fundamental tool to produce new invariants and covariants from known ones in the study of invariant theory during the 19-th century. Explicitly, if \[ \begin{aligned} A&=\sum_{0 \leq i \leq m} {m \choose i} a_i x_1^{m-i} x_2^i\\ \text{and} B&=\sum_{0 \leq i \leq n} {n \choose i} b_i x_1^{n-i} x_2^i \end{aligned} \] are binary forms of degree \(m,n\) in the variables \(x_1,x_2\), the \(r\)-th transvectant of \(A\) and \(B\), denoted by \((A,B)_r\), is defined as \[ (A,B)_r = \{ (m-r)!(n-r)!/m!n! \} \sum_{0 \leq i \leq r} (-1)^i {r \choose i} \{ \partial^r A/ \partial x_1^{r-i} \partial x_2^i \} \{ \partial^r B/ \partial x_1^i \partial x_2^{r-i} \}. \] Note that \((A,B)_0=AB\), \((A,B)_1\) is the Jacobian of \(A,B\) with respect to \(x_1,x_2\), and \((A,B)_2\) is the polarized Hessian of \(A\) and \(B\). On the other hand, the transvectant \((A,B)_r\) can be understood via the \(SL_2\) representation theory. Let \(k\) be a field with \(\mathrm{char} k =0\), and \(V\) be a \(2\)-dimensional vector space over \(k\). Thus \(SL(V)\) acts on \(V\) and on \(S_m=\mathrm{Sym}^m(V)\), the \(m\)-th symmetric power of \(V\). If \(x_1\), \(x_2\) is a basis of the dual space of \(V\), then we may identify \(S_m\) with the set of all binary forms of degree \(m\) in the variables \(x_1,x_2\). It is known that the irreducible representation of \(SL(V)\) is of the form \(S_m\) for some non-negative integer \(m\). The Clebsch-Gordan decomposition tells \(S_m \otimes S_n = \bigoplus_{0 \leq r \leq \min \{ m,n \}} S_{m+n-2r}\). The formula of the projection map \(\pi_r : S_m \otimes S_n \rightarrow S_{m+n-2r}\) leads to the \(r\)-th transvectant of \(A\) and \(B\) (see Section 1.5 of the paper under review). Let \(A\) and \(B\) be generic binary forms of degree \(m,n\) in the variables \(x_1,x_2\), i.e., \(x_1,x_2, a_0, \dots, a_m, b_0, \dots, b_n\) are algebraically independent over the ground field \(k\). Denote \(u_r=(A,B)_r\). There exists algebraic relations among the transvectants \(u_r\) and their transvectants when \(r=0,1,2,3, \dots\). For example, we have a relation when \(m=5, n=3\), i.e., \(21 \cdot 32 (u_0, u_0)_2 + 21 \cdot 16 (u_0,u_1)_1 + 315 u_1^2=256 u_0u_2\). The main result of this paper is the following Theorem. Assume \(m,n,r \geq 2\) and \(A,B\) are generic binary forms of degree \(m,n\) in the variables \(x_1,x_2\). Then \(u_0u_r\) is a linear combination of \((u_i,u_j)_{r-i-j}\) with rational coefficients where \(i,j\) runs over \(0 \leq i \leq j < r\). The proof of this result, given in Section 2 of this paper, relies on the Cauchy exact sequence of \(SL_2\)-representations (see 2.1) and Wigner's \(9-j\)-symbols from the quantum theory of angular momentum (see Section 7). This result may be interpreted in terms of the incomplete Segre embedding in algebraic geometry. Similar results were computed for the cases of \(SL_3\), the \(14\)-dimensional exceptional Lie algebra \(g_2\), and the symmetric group of degree \(5\).