Highest-weight vectors for the adjoint action of \(\mathrm{GL}_n\) on polynomials (Q453235)

Let \( G \subset \mathrm{GL}_n\) be the group of invertible \(n \times n\) matrices over an algebraically closed field and let \(U\subset G\) be the subgroup consisting of upper unitriangular matrices. Then \(G\) acts on its Lie algebra \(\mathfrak g\) via the adjoint action (conjugation). This also defines an action of \(G\) on the algebra \(k[\mathfrak g]\) of polynomial functions on \(\mathfrak g\); the elements fixed under this action form the invariant algebra \(k[\mathfrak g]^G\). The highest-weight vectors of any given weight \(\lambda\) make up the \(k[\mathfrak g]^G\)-module \(k[\mathfrak g]^U_{\lambda}\). The article gives an explicit construction of a basis of this \(k[\mathfrak g]^G\)-module \(k[\mathfrak g]^U_{\lambda}\) for certain special weights \(\lambda\). Its method involves differentiation of the fundamental invariants of \(k[\mathfrak g]^G\) and it applies to any possible weight, although it is proven only for a family of \(2n-3\) special weights that it yields a basis (see Theorem 1). Moreover for five of these special weights it is also proven that the elements of this basis are algebraically independent over \(k[\mathfrak g]^G\) and they generate the \(k[\mathfrak g]^G\)-algebra \(\oplus_{r \geq 0} k[\mathfrak g]^U_{r \lambda}\) (see Theorem 2). These are generalisation of some results in [\textit{A. Premet} and \textit{R. Tange}, J. Algebra 294, No. 1, 177--195 (2005; Zbl 1097.17010)]. As an application for the case \(n=3\) a complete set of generators of the \(k\)-algebra \(k[\mathfrak g]^U\) is obtained together with the relations among them. Finally, a question regarding the possible extensions of this method is raised: namely, whether in characteristic zero the \(k[\mathfrak g]^G\)-module generators of \(k[\mathfrak g]^U_{\lambda}\) can be obtained by applying one explicit highest-weight vector of weight \(\lambda\) in the the tensor algebra \(T(\mathfrak g)\) to varying tuples of fundamental invariants.