ScalingStacks

1.5 Berkovich retractions [04MQ]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context Β· Original author HTML

1.5 Berkovich retractions

Let 𝒳\mathscr{X} be a good dlt model of a smooth proper KK-variety XX. We can now define a retraction for the inclusion Sk⁑(𝒳)βŠ‚Xan\Sk(\mathscr{X})\subset X^{\text{an}} as follows: for any v∈Xanv\in X^{\an}, there exists a minimal stratum YβŠ†βˆ©j∈JDjY\subseteq\cap_{j\in J}D_{j} of 𝒳k\mathscr{X}_{k} such that the center c𝒳​(v)c_{\mathscr{X}}(v) of vv is contained in YY. We then associate to vv the quasi-monomial valuation ρ𝒳​(v)\rho_{\mathscr{X}}(v) corresponding to the data (Y,w)(Y,w) with wj=1q​v​(zj)w_{j}=\frac{1}{q}v(z_{j}), where zjz_{j} is a local equation of q​DjqD_{j} at the generic point of YY, for some qβˆˆβ„•>0q\in\mathbb{N}_{>0}. This should be seen as a monomial approximation of the valuation vv at the generic point of YY, with respect to the model 𝒳\mathscr{X} (which is snc there).

Definition 1.5.1.

The above map ρ𝒳:Xan⟢Sk⁑(𝒳)\rho_{\mathscr{X}}:X^{\an}\longrightarrow\Sk(\mathscr{X}) is the Berkovich retraction associated with the model 𝒳/R\mathscr{X}/R.

The Berkovich retraction is continuous, restricts to the identity on Sk⁑(𝒳)\Sk(\mathscr{X}), and by [Thu07, Ber99] ρ𝒳\rho_{\mathscr{X}} is a strong deformation retraction, i.e. there is a homotopy between ρ𝒳\rho_{\mathscr{X}} and the identity on XanX^{\an} that fixes the points of Sk⁑(𝒳)\Sk(\mathscr{X}). It follows that XanX^{\an} and Sk⁑(𝒳)\Sk(\mathscr{X}) are homotopy equivalent.

Let YY be a stratum of 𝒳k\mathscr{X}_{k}. The formal scheme 𝒳/Y^\widehat{\mathscr{X}_{/Y}} admits a generic fiber 𝔛Y\mathfrak{X}_{Y} in the sense of Berkovich, which is a Berkovich space and can be explicitly described as the open subset of XanX^{\an}:

𝔛Y={x∈Xan|c𝒳(vx)∈Y}.\mathfrak{X}_{Y}=\{x\in X^{\an}\lvert\,c_{\mathscr{X}}(v_{x})\in Y\}.

It furthermore coincides with Οπ’³βˆ’1​(Star⁑(Ο„Y))βŠ‚Xan\rho^{-1}_{\mathscr{X}}(\Star(\tau_{Y}))\subset X^{\an}. This Berkovich space comes with a retraction:

ρY:𝔛Y⟢Star⁑(Ο„Y),\rho_{Y}:\mathfrak{X}_{Y}\longrightarrow\Star(\tau_{Y}),

which coincides with the restriction of the retraction ρ𝒳\rho_{\mathscr{X}}. Thus, the restriction of ρ𝒳\rho_{\mathscr{X}} over Star⁑(Ο„Y)\Star(\tau_{Y}) only depends on the formal completion 𝒳/Y^\widehat{\mathscr{X}_{/Y}}.

An explicit example of Berkovich retractions, which we will use as a local model in the rest of the paper, is as follows. Let 𝕋=𝔾m,Kn\mathbb{T}=\mathbb{G}^{n}_{m,K} be a torus, with character lattice MM and cocharacter lattice NN. We view the elements mm of MM as rational functions on 𝕋\mathbb{T}, so that its analytification 𝕋an\mathbb{T}^{\text{an}} comes with a continuous map:

val:\displaystyle\val: 𝕋an⟢Nℝ,\displaystyle\,\mathbb{T}^{\text{an}}\longrightarrow N_{\mathbb{R}},
vx⟼(m↦vx​(m)),\displaystyle v_{x}\longmapsto(m\mapsto v_{x}(m)),

under the identification Nℝ=Hom⁑(M,ℝ)N_{\mathbb{R}}=\Hom(M,\mathbb{R}). The notation val\val can be understood as follows: fix an isomorphism N≃℀nN\simeq\mathbb{Z}^{n}, so that 𝕋=Spec⁑K⁑[M]≃Spec⁑K⁑[X1Β±,…,XnΒ±]\mathbb{T}=\Spec K[M]\simeq\Spec K[X_{1}^{\pm},\ldots,X_{n}^{\pm}], and vx​(m)=vx​(Xm)=βˆ‘i=1nmi​vx​(Xi)v_{x}(m)=v_{x}(X^{m})=\sum_{i=1}^{n}m_{i}v_{x}(X_{i}), so that the map val\val reads:

val:\displaystyle\val: 𝕋anβŸΆβ„n,\displaystyle\,\mathbb{T}^{\text{an}}\longrightarrow\mathbb{R}^{n},
vx⟼(vx​(Xi))i=1,…,n.\displaystyle v_{x}\longmapsto(v_{x}(X_{i}))_{i=1,\ldots,n}.

Since vx​(Xi)=βˆ’log⁑|Xi​(x)|v_{x}(X_{i})=-\log\lvert X_{i}(x)\rvert, this is the non-archimedean analog of the map (β„‚βˆ—)nβŸΆβ„n(\mathbb{C}^{*})^{n}\longrightarrow\mathbb{R}^{n} sending (z1,…,zn)(z_{1},\ldots,z_{n}) to βˆ’(log⁑|z1|,…,log⁑|zn|)-(\log\lvert z_{1}\rvert,\ldots,\log\lvert z_{n}\rvert).

The map val\val admits a continuous section ΞΆ:Nβ„βŸΆπ•‹an\zeta:N_{\mathbb{R}}\longrightarrow\mathbb{T}^{\text{an}}, sending a point n∈Nℝn\in N_{\mathbb{R}} to the Gauss point of the affinoid torus valβˆ’1⁑(n)\val^{-1}(n). More explicitly, for xβˆˆπ•‹anx\in\mathbb{T}^{\an}, the valuation ΢⁑(val⁑(x))\zeta(\val(x)) is the valuation on the function field of 𝕋\mathbb{T} defined by the following formula:

΢⁑(val⁑(x))​(βˆ‘m∈MΞ±m​zm)=minΞ±mβ‰ 0⁑(ordt⁑(Ξ±m)+vx​(zm)).\zeta(\val(x))\big(\sum_{m\in M}\alpha_{m}z^{m}\big)=\min_{\alpha_{m}\neq 0}\big(\ord_{t}(\alpha_{m})+v_{x}(z^{m})\big).

Now let 𝒳/R\mathscr{X}/R be a regular toric model of 𝕋\mathbb{T}, i.e. a regular toric RR-scheme such that 𝒳×RSpec⁑K=𝕋\mathscr{X}\times_{R}\;\Spec K=\mathbb{T}, which we assume to have reduced special fiber. Such a model is described by a regular fan Ξ£^βŠ‚N^ℝ=Nℝ×ℝβ‰₯0\hat{\Sigma}\subset\hat{N}_{\mathbb{R}}=N_{\mathbb{R}}\times\mathbb{R}_{\geq 0}, whose cones intersect Nℝ×{0}N_{\mathbb{R}}\times\{0\} only at the origin.
We consider the following open subset of 𝕋an\mathbb{T}^{\an}:

𝒳^Ξ·:={vxβˆˆπ•‹an|vxhas a center on𝒳},\widehat{\mathscr{X}}_{\eta}:=\{v_{x}\in\mathbb{T}^{\an}\lvert\,v_{x}\;\text{has a center on}\;\mathscr{X}\},

which admits a Berkovich retraction:

ρ𝒳:𝒳^η⟢Sk⁑(𝒳)\rho_{\mathscr{X}}:\widehat{\mathscr{X}}_{\eta}\longrightarrow\Sk(\mathscr{X})

defined as above. In this case, the map ρ𝒳\rho_{\mathscr{X}} can be described explicitly as follows: let Ξ£1\Sigma_{1} be the polyhedral complex obtained by intersecting the fan Ξ£^\hat{\Sigma} with Nℝ×{1}N_{\mathbb{R}}\times\{1\}. There is a natural identification between Ξ£1\Sigma_{1} and π’Ÿβ‘(𝒳k)\mathcal{D}(\mathscr{X}_{k}), sending a vertex of π’Ÿβ‘(𝒳k)\mathcal{D}(\mathscr{X}_{k}) to the primitive generator of the corresponding ray of Ξ£^\hat{\Sigma}, and then extending on each face by linearity. Moreover, it follows from [GJKM19, Theorem A.4] that ΢⁑(|Ξ£1|)=Sk⁑(𝒳)βŠ‚π•‹an\zeta(\lvert\Sigma_{1}\rvert)=\Sk(\mathscr{X})\subset\mathbb{T}^{\an}.

Proposition 1.5.2 ([NXY19, Example 3.5]).

The equality:

𝒳^Ξ·=valβˆ’1⁑(|Ξ£1|)\widehat{\mathscr{X}}_{\eta}=\val^{-1}(\lvert\Sigma_{1}\rvert)

holds, and ρ𝒳=val|𝒳^Ξ·\rho_{\mathscr{X}}=\val_{|\widehat{\mathscr{X}}_{\eta}}.

Proof.

We start by proving the first equality. Let xβˆˆπ•‹anx\in\mathbb{T}^{\an}, we know from LemmaΒ 1.5.3 below that xx has a center on 𝒳\mathscr{X} if and only if ΢⁑(val⁑(x))\zeta(\val(x)) has a center on 𝒳\mathscr{X}. Thus, it is enough to prove that for n∈Nℝn\in N_{\mathbb{R}} and y=΢⁑(n)y=\zeta(n), yy has a center on 𝒳\mathscr{X} if and only if n∈|Ξ£1|n\in\lvert\Sigma_{1}\rvert.
The elements y∈΢⁑(Nℝ)y\in\zeta(N_{\mathbb{R}}) are precisely the valuations invariant under the torus action, hence if yy has a center on 𝒳\mathscr{X}, it must be the closure of a torus orbit YβŠ‚π’³kY\subset\mathscr{X}_{k}. By [KKMSD73, Theorem 6], there exists a cone ΟƒβˆˆΞ£^\sigma\in\hat{\Sigma} such that the generic point of YY is contained in the associated toric affine chart 𝒳σ=Spec⁑R⁑[ΟƒΛ‡βˆ©M^]\mathscr{X}_{\sigma}=\Spec R[\check{\sigma}\cap\hat{M}]. In particular, for any monomial zmz^{m} that is regular on 𝒳σ\mathscr{X}_{\sigma}, we have vy​(zm)β‰₯0v_{y}(z^{m})\geq 0. In other words, writing y=΢⁑(n)y=\zeta(n), we have ⟨n,m⟩β‰₯0\langle n,m\rangle\geq 0 for all mβˆˆΟƒΛ‡m\in\check{\sigma}, so that nβˆˆΟƒn\in\sigma. Since vy​(t)=1v_{y}(t)=1, y∈΢⁑(|Ξ£1|)y\in\zeta(\lvert\Sigma_{1}\rvert).
By the same argument, if n∈|Ξ£1|n\in\lvert\Sigma_{1}\rvert, there exists a cone Οƒ\sigma such that nβˆˆΟƒn\in\sigma, which means that v΢⁑(n)v_{\zeta(n)} has positive value on each monomial mβˆˆΟƒΛ‡m\in\check{\sigma}, and thus has a center on 𝒳σ\mathscr{X}_{\sigma} and in particular on 𝒳\mathscr{X}.

To prove the second equality, since ρ𝒳\rho_{\mathscr{X}} is the identity on ΢⁑(|Ξ£1|)=Sk⁑(𝒳)\zeta(\lvert\Sigma_{1}\rvert)=\Sk(\mathscr{X}), we merely have to prove that ρ𝒳=Οπ’³βˆ˜val\rho_{\mathscr{X}}=\rho_{\mathscr{X}}\circ\val. However this follows directly from the definition of ρ𝒳\rho_{\mathscr{X}}, and the fact that c𝒳​(x)∈c𝒳​(΢​(val⁑(x)))Β―c_{{\mathscr{X}}}(x)\in\overline{c_{{\mathscr{X}}}(\zeta(\val(x)))} for xβˆˆπ’³^Ξ·x\in\widehat{\mathscr{X}}_{\eta} by LemmaΒ 1.5.3. Indeed, ρ𝒳​(x)\rho_{\mathscr{X}}(x) only depends on the values vx​(z)v_{x}(z), where zz is a local equation for a component of 𝒳k\mathscr{X}_{k} at c𝒳​(x)c_{\mathscr{X}}(x). Since 𝒳\mathscr{X} is a toric model, these local equations can be taken to be monomials, so that the result follows from the fact that xx and ΢⁑(val⁑(x))\zeta(\val(x)) take the same values on monomials. ∎

Lemma 1.5.3.

Let xβˆˆπ•‹anx\in\mathbb{T}^{\an}. Then xx has a center on 𝒳\mathscr{X} if and only if ΢⁑(val⁑(x))\zeta(\val(x)) has a center on 𝒳\mathscr{X}. Moreover, if this holds, we have c𝒳​(x)∈c𝒳​(΢​(val⁑(x)))Β―c_{{\mathscr{X}}}(x)\in\overline{c_{{\mathscr{X}}}(\zeta(\val(x)))}.

Proof.

Let π’³βŠ‚π’³Β―\mathscr{X}\subset\bar{\mathscr{X}} be a toric compactification of 𝒳\mathscr{X}, i.e. a proper toric RR-scheme containing 𝒳\mathscr{X} as a torus-invariant open subset. By the valuative criterion of properness, any valuation of 𝕋an\mathbb{T}^{\an} has a center on 𝒳¯\bar{\mathscr{X}}. We write c𝒳¯​(x)c_{\bar{\mathscr{X}}}(x) for the center of xβˆˆπ•‹anx\in\mathbb{T}^{\an}.

We start by proving that c𝒳¯​(΢​(val⁑(x))CLOSEc_{\bar{\mathscr{X}}}(\zeta(\val(x)) is the generic point of the minimal closed torus orbit ZZ in 𝒳¯\bar{\mathscr{X}} containing c𝒳¯​(x)c_{\bar{\mathscr{X}}}(x). We may work on the toric affine chart 𝒳σ=Spec⁑R⁑[ΟƒΛ‡βˆ©M^]\mathscr{X}_{\sigma}=\Spec R[\check{\sigma}\cap\hat{M}] associated with ZZ. Since the valuation ΢⁑(val⁑(x))\zeta(\val(x)) is monomial, it is enough to prove that ΢⁑(val⁑(x))​(zm)=vx​(zm)β‰₯0\zeta(\val(x))(z^{m})=v_{x}(z^{m})\geq 0 for mβˆˆΟƒΛ‡βˆ©Mm\in\check{\sigma}\cap M and that ΢​(val⁑(x))​(z)>0\zeta(\val(x))(z)>0 for zz a local equation of any torus invariant divisor containing ZZ, to have that c𝒳¯​(΢​(val⁑(x))CLOSEc_{\bar{\mathscr{X}}}(\zeta(\val(x)) lies in ZZ. Since zmz^{m} is regular on 𝒳σ\mathscr{X}_{\sigma}, the first condition holds; the local equation zz is monomial and ΢⁑(val⁑(x))​(z)=vx​(z)>0\zeta(\val(x))(z)=v_{x}(z)>0 since ZZ contains c𝒳¯​(x)c_{\bar{\mathscr{X}}}(x). Moreover, we conclude that c𝒳¯​(x)c_{\bar{\mathscr{X}}}(x) must be contained in the toric interior of ZZ by minimality of ZZ.

Now assume that vxv_{x} is centered on 𝒳\mathscr{X}, i.e. c𝒳¯​(x)βˆˆπ’³c_{\bar{\mathscr{X}}}(x)\in\mathscr{X}. Since 𝒳\mathscr{X} is torus-invariant and c𝒳¯​(x)∈Zc_{\bar{\mathscr{X}}}(x)\in Z, we have ZβŠ‚π’³Z\subset\mathscr{X}, hence its generic point c𝒳¯​(΢⁑(val⁑(x)))βˆˆπ’³c_{\bar{\mathscr{X}}}(\zeta(\val(x)))\in\mathscr{X}. This implies that ΢⁑(val⁑(x))\zeta(\val(x)) has center on 𝒳\mathscr{X}. Conversely, if ΢⁑(val⁑(x))\zeta(\val(x)) has a center on 𝒳\mathscr{X}, then ZβŠ‚π’³Z\subset\mathscr{X} and c𝒳¯​(x)∈Zc_{\bar{\mathscr{X}}}(x)\in Z as mentioned above; thus, c𝒳​(x)βˆˆπ’³c_{\mathscr{X}}(x)\in\mathscr{X}, which concludes the proof. ∎

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.