On polarizations in invariant theory (Q2498893): Difference between revisions

Let \(G\) be a reductive algebraic group over an algebraically closed field \(k\) of characteristic \(0\), and let \(V\) be a finite dimensional algebraic \(G\)-module. Given a positive integer \(n\), finding generators of the invariant algebra \(k[V^{\oplus n}]^G\) of \(V^{\oplus n}\) is the classical problem of invariant theory. Let \(\text{pol}_nk[V]^G\) be the subalgebra of \(k[V^{\oplus n}]^G\) generated by the polarizations of all the invariants \(f\in k[V]^G.\) Then there are \(G\)-modules such that \(\text{pol}_nk[V]^G=k[V^{\oplus n}]^G.\) The authors give several well known examples when this condition holds and also some examples, even for finite \(G\), on when the condition does not hold. The main purpose of the article is to analyze the relationship between \(k[V^{\oplus}]^G\) and \(\text{pol}_nk[V]^G.\) The authors prove that if \(G\) is finite, then \(k[V^{\oplus n}]^G\) is the integral closure of \(\text{pol}_nk[V]^G\) in its field of fractions, and that the natural morphism of affine varieties determined by these algebras is bijective. The authors generalize the problem: They study the more general setting of actions on arbitrary affine varieties for which they define a generalization of polarizations. They then prove that if \(G\) is finite, then the invariant algebra is integral over the subalgebra generated by generalized polarizations, and the natural dominant morphism between affine varieties determined by these algebras is injective. For connected \(G\), this result obviously does not hold. The authors define \(\text{pol}\text{ind}(V)\) as the supremum over \(n\) such that the result holds. They then prove that \(k[V^{\oplus n}]^G\) is integral over \(\text{pol}_mk[V]^G\) for every \(m\leq \text{pol}\text{ind}(V)\) The authors prove that calculating \(\text{pol}\text{ind}(V)\) is related to the description of linear subspaces of the Hilbert nullcone of \(V\), and using this, the authors are able to compute the polarization index of some \(G\)-modules \(V\). That is: If \(G\) is a finite group or a linear algebraic torus, then \(\text{pol}\text{ind}(V)=\infty.\) For \(G=\text{SL}_2\) all linear subspaces of \(V\) in the nullcone of \(V\) are described and the authors then prove that \(\text{pol}\text{ind}(V)=\infty\) if \(V\) does not contain a simple 2-dimensional submodules, and \(\text{pol}\text{ind}(V)=1\) otherwise. Finally the authors calculate the polarization index of every semisimple Lie algebra \(\mathfrak g\), and they get some nice results from this. The article is explicit to read. The authors generalize classical terms in a nice and detailed way, and the proofs are fully possible to understand.