[Paper Review] Espaces de Berkovich, polytopes, squelettes et théorie des modèles

Antoine Ducros|arXiv (Cornell University)|Mar 29, 2012

Algebraic Geometry and Number Theory12 references4 citations

TL;DR

This paper provides new model-theoretic proofs for two foundational results in Berkovich geometry: (1) the image of a non-Archimedean analytic space under invertible functions is a polytope in the multiplicative vector space $({\mathbb{R}}^*_{+})^n$, and (2) the preimage of the skeleton under a morphism to $\mathbb{G}_{m,k}^{n,\text{an}}$ admits a canonical piecewise-linear structure. The proofs avoid de Jong's alterations and instead use elimination of quantifiers and properties of stable domination from model theory of valued fields.

ABSTRACT

Let $X$ be an analytic space over a non-Archimedean, complete field $k$ and let $(f_1,..., f_n)$ be a family of invertible functions on $X$. Let $ϕ$ the morphism $X o G_m^n$ induced by the $f_i$'s, and let $t$ be the map $X o (R^*_+)^n$ induced by the norms of the $f_i$'s. Let us recall two results. 1) The compact set $t(X)$ is a polytope of the $R$-vector space $(R^*_+)^n$ (we use the multiplicative notation) ; this is due to Berkovich in the locally algebraic case, and has been extended to the general case by the author. 2) If moreover $X$ is Hausdorff and $n$-dimensional, then the pre-image under $ϕ$ of the skeleton $S_n$ of $G_m^n$ has a piecewise-linear structure making $ϕ^{-1}(S_n) o S_n$ a piecewise immersion ; this is due to the author. In this article, we improve 1) and 2), and give new proofs of both of them. Our proofs are based upon the model theory of algebraically closed, non-trivially valued fields. Let us quickly explain what we mean by improving 1) and 2). - Concerning 1), we also prove that if $x\in X$, there exists a compact analytic neighborhood $U$ of $x$, such that for every compact analytic neighborhood $V$ of $x$ in $X$, the germs of polytopes $(t(U),t(x))$ and $(t(V),t(x))$ coincide. - Concerning 2), we prove that the piecewise linear structure on $ϕ^{-1}(S_n)$ is canonical, that is, doesn't depend on the map we choose to write it as a pre-image of the skeleton; we thus answer a question which was asked to us by Temkin. Moreover, we prove that the pre-image of the skeleton 'stabilizes after a finite, separable ground field extension', and that if $ϕ_1,..., ϕ_m$ are finitely many morphisms from $X o G_m^n$, the union $\bigcup ϕ_j(S_n)$ also inherits a canonical piecewise-linear structure.

Motivation & Objective

• To reprove two key results in Berkovich geometry—polytopality of images under invertible functions and canonical piecewise-linear structures on preimages of skeleta—using model theory.
• To eliminate reliance on de Jong's alterations, which were previously used in earlier proofs of these results.
• To establish that the piecewise-linear structure on $\varphi^{-1}(S_n)$ is canonical, independent of the morphism choice, resolving a question posed by Temkin.
• To show that the preimage of the skeleton stabilizes after a finite separable field extension.
• To extend the canonical piecewise-linear structure to finite unions of such preimages.

Proposed method

• Employ elimination of quantifiers in the theory of algebraically closed, non-trivially valued fields to handle the polytopality result.
• Use advanced model-theoretic tools, particularly properties of stable domination introduced by Hrushovski and Loeser, to analyze the structure of preimages of skeleta.
• Apply a formalization of the tropicalization map via multiplicative notation on $({\mathbb{R}}^*_{+})^n$, treating it as a real vector space.
• Use a compactness argument to show that the germ of the image polytope at a point is well-defined and independent of the neighborhood choice.
• Construct canonical piecewise-linear structures by leveraging invariance under base change and finite separable extensions.
• Prove that the union of finitely many preimages of skeleta inherits a canonical piecewise-linear structure via model-theoretic finiteness properties.

Experimental results

Research questions

• RQ1 Can the polytopality of the image of a Berkovich space under invertible functions be proven without de Jong's alterations?
• RQ2 Is the piecewise-linear structure on $\varphi^{-1}(S_n)$ canonical, independent of the morphism used to define it?
• RQ3 Does the preimage of the skeleton stabilize under finite separable field extensions?
• RQ4 Can the canonical piecewise-linear structure be extended to finite unions of such preimages?
• RQ5 What model-theoretic tools can be used to formalize and generalize tropicalization and skeleton structures in non-Archimedean geometry?

Key findings

• The image $|\mathbf{f}|(X)$ of a Berkovich space $X$ under a family of invertible functions $\mathbf{f} = (f_1, \dots, f_n)$ is a polytope in $({\mathbb{R}}^*_{+})^n$, and this polytope is well-defined at the germ level around any point.
• The preimage $\varphi^{-1}(S_n)$ of the standard skeleton $S_n$ in $\mathbb{G}_{m,k}^{n,\text{an}}$ admits a canonical piecewise-linear structure, independent of the morphism $\varphi$.
• The canonical piecewise-linear structure on $\varphi^{-1}(S_n)$ is preserved under finite separable base field extensions.
• For finitely many morphisms $\varphi_1, \dots, \varphi_m: X \to \mathbb{G}_{m,k}^{n,\text{an}}$, the union $\bigcup \varphi_j^{-1}(S_n)$ inherits a canonical piecewise-linear structure.
• The proofs avoid de Jong's alterations and instead rely on model-theoretic techniques, including elimination of quantifiers and properties of stable domination.
• The results generalize and strengthen earlier results by Berkovich and the author, providing a more intrinsic and canonical framework for tropicalization and skeleton theory in Berkovich geometry.