Semi-stable reduction implies minimality of the resultant (Q405881)

The group \(\mathrm{PGL}_{n+1}\) acts on \(\mathbb{P}^n\) as coordinate-changes, and hence it acts on the set of morphisms from \(\mathbb{P}^n\) to \(\mathbb{P}^n\) via conjugation. The set \(M_d^n\) of morphisms which induce a map of degree \(d\) on the Picard group can be written as the affine open of \(\mathbb P^N\) where the Macaulay resultant \(\rho\) does not vanish, with \(N = \binom {n+d}d (n+1) - 1\). Since \(\rho\) is invariant under \(\mathrm{PSL}\) and \(\mathrm{PGL} = \mathrm{PSL}\) over algebraically closed fields, one can apply the geometric invariant theory in this setting with \(\mathrm{PSL}\). The semi-stable locus can be very explicitly written as \(\mathrm{Proj}\,\, \mathbb Z[x_0,\dots, x_N]^{\mathrm{SL}_{n+1}}\). Using this description, the authors improve their earlier work [\textit{L. Szpiro} et al., Publ. Mat., Barc. 58, No. 2, 295--329 (2014; Zbl 1392.37108)] to show that if a point \([a_0:\dots:a_N]\) defined over a global field is semi-stable modulo a place \(p\), then the point is minimal with respect to \(\rho\) at \(p\) (Theorem 3.3). That is, \[ \frac{\rho(a_0,\dots, a_N)}{\max(|a_0|_p, \dots, |a_N|_p)^{\deg \rho}} \] has the least \(p\)-adic valuation among the coordinate-changes (i.e. \(\mathrm{PGL}\)-action) of \([a_0:\dots : a_N]\). Thus, good/bad reduction can be checked immediately for semi-stable points, and numerical criteria for semi-stability for \(n=1\) [\textit{J. H. Silverman}, Duke Math. J. 94, No. 1, 41--77 (1998; Zbl 0966.14031)] automatically imply the minimality. As a corollary, they show that a bad reduction at a semi-stable point persists in algebraic extensions (Corollary 4.1). For number fields, they also show that each morphism has a coordinate-change over an extension such that the map is semi-stable (and thus minimal) for all \(p\) (Theorem 5.1). The question of finding globally semi-stable presentation has a counterexample over function fields [\textit{A. Levy}, Algebra Number Theory 6, No. 7, 1483--1501 (2012; Zbl 1268.14043)]. Note that [\textit{N. Bruin} and \textit{A. Molnar}, LMS J. Comput. Math. 15, 400--417 (2012; Zbl 1300.37051)] and [\textit{R. Rumely}, ``The minimal resultant locus'', preprint, \url{arXiv:1304.1201}] also give other approaches to minimal resultants in the case of \(n=1\). The authors give several examples, including one that shows that the Lattés maps on \(\mathbb P^1\) coming from an elliptic curve \(E\) are actually unstable at places where \(E\) has semi-stable reductions (Example 3.2).