The semistable reduction problem for the space of morphisms on \(\mathbb P^n\) (Q1932515)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The semistable reduction problem for the space of morphisms on \(\mathbb P^n\) |
scientific article |
Statements
The semistable reduction problem for the space of morphisms on \(\mathbb P^n\) (English)
0 references
18 January 2013
0 references
A rational map from \(\mathbb P^n\) to itself can be described by \(n+1\) homogeneous polynomials of degree \(d\) in \(n+1\) variables, so coefficients of the polynomials define a point on \(\mathbb P^N\) for \(N+1 = (n+1)\binom {n+d}d\). Within this \(\mathbb P^N\), the set of morphisms forms an open affine variety \(\mathrm{Hom}_d^n\). Choosing a different coordinate on \(\mathbb P^n\) amounts to a conjugation by an element of \(\mathrm{PGL}(n+1)\), so the natural moduli space for dynamical systems on \(\mathbb P^n\) is the quotient of \(\mathrm{Hom}_d^n\) by \(\mathrm{PGL}(n+1)\). \textit{C. Petsche, L. Szpiro}, and \textit{M. Tepper} [J. Algebra 322, No. 9, 3345--3365 (2009; Zbl 1190.14013)] have shown that this exists as a geometric quotient in the sense of Geometric Invariant Theory, generalizing the result of \textit{J. Silverman} [Duke Math. J. 94, No. 1, 41--77 (1998; Zbl 0966.14031)] for \(n=1\). There is also a previous work by the author [Acta Arith. 146, No. 1, 13--31 (2011; Zbl 1285.37020)], explicitly describing stable and semistable loci and giving a bound on on the stabilizer group of a point in the moduli space. In this article, the author restates semistable reduction theorem for a closely-related moduli space given by the quotient of \(\mathrm{Hom}_d^n\) by \(\mathrm{SL}(n+1)\), and further proves some refinements. The semistable reduction theorem is a standard result in GIT, stating that the semistability on the generic fiber implies that on the special fiber, possibly after taking a finite cover. In the current context, this means that given a complete curve \(C\) and a semistable map \(\varphi\) on \(\mathbb P^n_{K(C)}\), there exist an abstract curve \(D\) mapping finite-to-one onto \(C\) and a self-map \(\Phi\) on a \(\mathbb P^n\)-bundle on \(D\) such that \(\Phi\) is equivalent to the pullback of \(\varphi\) to \(D\) upon some coordinate changes and the reduction of \(\Phi\) at each point of \(D\) is semistable. In particular, one can apply this theorem to a curve \(C\) contained inside the semistable part \(M_d^{n,ss}\) of the moduli space, since each point of \(C\) represents a semistable self-map of \(\mathbb P^n\). The author then answers several natural questions arising from this setup. First, he proves that for any \(n\) and \(d\), there exists a curve in \(M_d^{n,ss}\) such that no trivial bundle on a finite cover satisfies the semistable reduction. As a result, such a curve does not have a complete curve in \(\mathrm{Hom}_d^{n,ss}\) as a finite cover. Secondly, the author shows that an infinitely many non-isomorphic bundle classes can occur for some rational curves in \(M_d^{n,ss}\). Thirdly, the author proves that if the trivial bundle on \(D\) satisfies semistable reduction, the degree of the cover \(D\to C\) is bounded in terms of the sizes of the stabilizer groups for the points on \(D\). The proofs involve standard GIT arguments together with explicit analyses of the closures of \(\mathrm{PGL}(n+1)\)-orbits of some polynomial families.
0 references
GIT
0 references
semistable reduction
0 references
moduli space
0 references
dynamical system
0 references