Invariant aspects of elimination (Q1901245): Difference between revisions

This paper is a continuation of the author's work on various aspects of elimination and resultant. For the earlier relevant articles see Adv. Math. 37, 212-238 (1980; Zbl 0527.13005) and 90, No. 2, 117-263 (1991; Zbl 0747.13007). There is a subsequent article dealing with similar issues [Adv. Math. 126, No. 2, 119-250 (1997; see the following review)]. These papers are rather long and thus difficult to go over completely in this short review. I will just describe the basic problem that is being tackled. Let \(\kappa\) be a scheme and \(E\) a rank \(n\) locally free sheaf over \(\kappa\). Let \(r\geq 1\) be an integer and let \(d_1,\ldots,d_r\) be positive integers. Let \[ A=\text{Sym}_\kappa\Bigl(\bigoplus_{i=1}^r \text{Sym}^{d_i}(E)^*\Bigr) \] and \(S=\text{Spec} A\). Let \(C=A\otimes_\kappa\text{Sym}_\kappa(E)\). One has the canonical section \[ {\mathcal O}_\kappa\to\text{Sym}^{d_i}(E)^*\otimes\text{Sym}^{d_i}(E) \] and thus one has a natural map \(f_i:C(-d_i)\to C\), a \(C\)-module homomorphism. Essentially \(f_i\) corresponds to the `universal homogeneous polynomial' of degree \(d_i\). Let \(B=C/(f_1,\ldots, f_r)\) and \(X=\text{Proj} B\). We have the natural projective morphism from \(X\to S\) and let \(T\) be the image of \(X\). The basic issue of elimination theory from the view point of invariant theory is to describe \(T\), in terms of the representations of \(\Aut_\kappa(E)\). The author does this in great detail and in full generality. In particular no Noetherian hypothesis is made and in author's words \textit{sans contorsions perpétuelles}. The stress here is to keep track of the \(\Aut(E)\) structure throughout. In the case when \(r=n\) above, we have the classical case of resultant and the author describes methods to see which points are singular etc. Needless to say, these get connected to Jacobian ideals, Grothendieck's residue and utilises some of the classical isomorphisms like Kronecker isomorphism (if \(E,F\) are vector bundles of rank \(n,m\), then \(\det(E\otimes F)=(\det E)^m\otimes(\det F)^n\), the equality being natural). In conclusion, this paper is a very exhaustive treatment of the subject, where references in this generality are hard to come by.