ScalingStacks

2.2. Semi-stable curves and reduction graphs [01J8]

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

2.2. Semi-stable curves and reduction graphs

In this section, we assume that X\mathrm{X} is the analytic space associated to a projective curve over a field KK which is complete for a discrete valuation. The semi-stable reduction theorem of Deligne–Mumford asserts that, up to replacing the base field KK by a finite extension, the curve X\mathrm{X} has a projective model 𝔛\mathfrak{X} over K∘K^{\circ} which is regular (as a 2-dimensional scheme) and whose special fiber is reduced, with at most double points for singularities. We may also assume that the irreducible components are geometrically irreducible. We do not require, however, that 𝔛\mathfrak{X} is the minimal semi-stable model.

The reduction graph of the special fiber

In that situation, the reduction graph R⁑(𝔛)R(\mathfrak{X}) is a metrized graph defined as follows. It has for vertices the irreducible components of the special fiber, with as many edges of length log⁑|Ο€|βˆ’1\log\left|{\pi}\right|^{-1} between two vertices as the number of intersection points of the corresponding components. In an neighbourghood of a double point, 𝔛\mathfrak{X} looks like (i.e., has an Γ©tale map to) the scheme with equation x​y=Ο€xy=\pi in the affine plane 𝐀K∘2\mathbf{A}^{2}_{K^{\circ}}.

If one replaces the field KK by a finite extension Kβ€²K^{\prime}, the base change π”›βŠ—K∘(Kβ€²)∘\mathfrak{X}\otimes_{K^{\circ}}(K^{\prime})^{\circ} may no more be regular. Indeed, π”›βŠ—K∘(Kβ€²)∘\mathfrak{X}\otimes_{K^{\circ}}(K^{\prime})^{\circ} is Γ©tale locally isomorphic to x​y=(Ο€β€²)exy=(\pi^{\prime})^{e}, where Ο€β€²\pi^{\prime} is a uniformizing element of Kβ€²K^{\prime}, and ee is the ramification index. When e>1e>1, the origin is a singular point of that scheme and one needs to blow it up repeatedly in order to obtain a regular scheme, which is a semi-stable model of XKβ€²\mathrm{X}_{K^{\prime}} over (Kβ€²)∘(K^{\prime})^{\circ}. The two initial components are replaced by a chain of e+1e+1 components, the eβˆ’1e-1 intermediate ones being projective lines. In other words, eβˆ’1e-1 vertices have been added, regularly spaced along each edge. One concludes that the reduction graph has not changed, as a topological space. Its metric has not changed neither, since the ee edges that partition an original edge (of length log⁑|Ο€|βˆ’1\log\left|{\pi}\right|^{-1}) have length log⁑|Ο€β€²|βˆ’1=1e​log⁑|Ο€|βˆ’1\log\left|{\pi^{\prime}}\right|^{-1}=\frac{1}{e}\log\left|{\pi}\right|^{-1}.

We say that a function on R⁑(𝔛)R(\mathfrak{X}) is piecewise linear if, up to passing to a finite extension (which replaces each edge by ee edges of length equal to 1/e1/eth of the initial one), it is linear on each edge.

Drawing the reduction graph on the Berkovich space

Let us analyse the situation from the Berkovich viewpoint. As we have seen, the generic points of the special fiber are the reductions of canonical points of X\mathrm{X} : the vertices of the graph R⁑(𝔛)R(\mathfrak{X}) naturally live in X\mathrm{X}. The same holds for the edges, but is a bit more subtle. As we have seen, blowing-up intersection points of components in the special fiber gives rise to new components, hence to new points of X\mathrm{X}. Would we enlarge the ground field and blow-up indefinitely, the constellation of points in X\mathrm{X} that we draw converges to a graph which is isomorphic to R⁑(𝔛)R(\mathfrak{X}).

According to Berkovich [12], a far more precise result holds. Let us consider a neighborhood π”˜\mathfrak{U} of a singular point of the special fiber, pretending it is isomorphic to the locus defined by the equation x​yβˆ’Ο€xy-\pi in 𝐀2\mathbf{A}^{2} ; so π”˜=Spec⁑(Kβˆ˜β€‹[x,y]/(x​yβˆ’Ο€))\mathfrak{U}=\operatorname{Spec}(K^{\circ}[x,y]/(xy-\pi)). Its generic fibre is the affinoid space U\mathrm{U} defined by the inequality |x​y|=|Ο€|\left|{xy}\right|=\left|{\pi}\right| in the unit polydisk B2=ℳ⁑(K⁑⟨x,y⟩CLOSE\mathrm{B}^{2}=\mathscr{M}(K\langle x,y\rangle. The affinoid algebra of U\mathrm{U} is the quotient

Kβ€‹βŸ¨x,y⟩/(x​yβˆ’Ο€)K\langle x,y\rangle/(xy-\pi)

whose elements ff are (non-uniquely) represented by a series

βˆ‘m,n=0∞am,n​xm​yn,\sum_{m,n=0}^{\infty}a_{m,n}x^{m}y^{n},

with am,nβ†’0a_{m,n}\rightarrow 0 when m+nβ†’βˆžm+n\rightarrow\infty. However, observing that xx is invertible in this algebra, with inverse Ο€βˆ’1​y\pi^{-1}y, so that y=π​xβˆ’1y=\pi x^{-1}, we can replace each product xm​ynx^{m}y^{n} by Ο€n​xmβˆ’n\pi^{n}x^{m-n}, leading to an expression of the form

f=βˆ‘nβˆˆπ™an​xn,f=\sum_{n\in{\mathbf{Z}}}a_{n}x^{n},

where |an|β†’0\left|{a_{n}}\right|\rightarrow 0 when nβ†’+∞n\rightarrow+\infty and |an|β€‹Ο€βˆ’nβ†’0\left|{a_{n}}\right|\pi^{-n}\rightarrow 0 when nβ†’βˆ’βˆžn\rightarrow-\infty. Such an expression is now unique, and is called the Laurent expansion of ff.

It leads to a natural family (Ξ³r)r∈[0,log⁑|Ο€|βˆ’1](\gamma_{r})_{r\in[0,\log\left|{\pi}\right|^{-1}]} of multiplicative seminorms on the algebra π’ͺ⁑(U)\mathscr{O}(\mathrm{U}), parametrized by the unit interval in 𝐑{\mathbf{R}}. Namely, for each real number r∈[0,log⁑|Ο€|βˆ’1]r\in[0,\log\left|{\pi}\right|^{-1}], we can set

Ξ³r​(f)=maxnβˆˆπ™β‘|an|​eβˆ’r​n,f=βˆ‘nβˆˆπ™an​xn∈π’ͺ⁑(U).\gamma_{r}(f)=\max_{n\in{\mathbf{Z}}}\left|{a_{n}}\right|e^{-rn},\qquad f=\sum_{n\in{\mathbf{Z}}}a_{n}x^{n}\in\mathscr{O}(\mathrm{U}).

Obviously, Ξ³r\gamma_{r} is a norm on π’ͺ⁑(U)\mathscr{O}(\mathrm{U}) which extends the absolute value of KK ; its multiplicativity is proved analogously that of the Gauß norm. It is easy to check that the map [0,log⁑|Ο€|βˆ’1]β†’U[0,\log\left|{\pi}\right|^{-1}]\rightarrow\mathrm{U} defined by r↦γrr\mapsto\gamma_{r} is continuous (this amounts to the fact that the maps r↦γr​(f)r\mapsto\gamma_{r}(f) are continuous), hence defines an parametrized path in the topological space U\mathrm{U}.

Let S⁑(π”˜)S(\mathfrak{U}) be its image (with the induced distance) ; Berkovich calls it the skeleton of the formal scheme obtained by completing π”˜\mathfrak{U} along its special fibre. A point uu in U\mathrm{U} has two coordinates (x⁑(u),y⁑(u))(x(u),y(u)) in the completed residue field ℋ⁑(u)\mathscr{H}(u) which are elements of absolute value ≀1\leq 1 satisfying x⁑(u)​y​(u)=Ο€x(u)y(u)=\pi. In particular,

r⁑(u)=log⁑|x⁑(u)|βˆ’1∈[0,log⁑|Ο€|βˆ’1].r(u)=\log\left|{x(u)}\right|^{-1}\in[0,\log\left|{\pi}\right|^{-1}].

The map ρ:u↦γr⁑(u)\rho\colon u\mapsto\gamma_{r(u)} is a continuous from U\mathrm{U} to S⁑(π”˜)S(\mathfrak{U}).

Let us compute the image of Ξ³r\gamma_{r} by this map. By definition of Ξ³r\gamma_{r}, one has

|x⁑(Ξ³r)|=Ξ³r​(x)=eβˆ’r,\left|{x(\gamma_{r})}\right|=\gamma_{r}(x)=e^{-r},

hence r⁑(Ξ³r)=rr(\gamma_{r})=r and ρ⁑(Ξ³r)=Ξ³r\rho(\gamma_{r})=\gamma_{r}. In other words, the map ρ\rho is a retraction of U\mathrm{U} onto the skeleton S⁑(π”˜)S(\mathfrak{U}).

The special fiber of π”˜\mathfrak{U} is defined by the equation x​y=0xy=0 in 𝐀K~2\mathbf{A}^{2}_{\tilde{K}}, hence has two components. One can check that the point Ξ³0\gamma_{0} reduces to the generic point of the component with equation y=0y=0, while Ξ³log⁑|Ο€|βˆ’1\gamma_{\log\left|{\pi}\right|^{-1}} reduces to the generic point of the component with equation x=0x=0.

These constructions have to be done around each singular point of the special fiber of 𝔛\mathfrak{X}, locally for the Γ©tale topology of 𝔛\mathfrak{X}. Berkovich proves that they can be glued, so that the graph R⁑(𝔛)R(\mathfrak{X}) is again canonically interpreted as an actual metrized graph drawn on the analytic space X\mathrm{X} ; we write ΞΉ:R⁑(𝔛)β†ͺX\iota\colon R(\mathfrak{X})\hookrightarrow\mathrm{X} for the canonical embedding. The map ΞΉ\iota admits a continuous retraction ρ:Xβ†’R⁑(𝔛)\rho\colon\mathrm{X}\rightarrow R(\mathfrak{X}).

Although we will not use this fact, we must mention that the retraction ρ\rho is a deformation retraction. (For any t∈[0,1]t\in[0,1] and any x∈Xx\in\mathrm{X}, ρt​(x)\rho_{t}(x) is the semi-norm x1βˆ’t​ρ​(x)tx^{1-t}\rho(x)^{t}.)

Metrized line bundles and the reduction graph

A construction of S. Zhang [57], building on prior results of Chinburg–Rumely [20], furnishes continuous metrics on divisors from continuous functions on the reduction graph R⁑(𝔛)R(\mathfrak{X}). It works as follows. First of all, if P∈X⁑(K)P\in\mathrm{X}(K) is a rational point, there is a unique morphism Ξ΅P:Spec⁑Kβˆ˜β†’π”›\varepsilon_{P}\colon\operatorname{Spec}K^{\circ}\rightarrow\mathfrak{X} which extends the point PP viewed as a morphism from Spec⁑K\operatorname{Spec}K to XX. The image of this section is a divisor DPD_{P} on 𝔛\mathfrak{X} and the line bundle π’ͺ⁑(DP)\mathscr{O}(D_{P}) on 𝔛\mathfrak{X} defines a smooth metric on π’ͺ⁑(P)\mathscr{O}(P) ; we write π’ͺ⁑(P)¯𝔛\overline{\mathscr{O}(P)}_{\mathfrak{X}} for the corresponding metrized line bundle. We also define ΞΌP\mu_{P} as the Dirac measure at the vertex of the graph corresponding to the (unique) irreducible component of the special fiber by which DPD_{P} passes through. The construction and the notation is extended by additivity for divisors which are sums of rational points. More generally, if PP is only a closed point of XX, we do this construction after the finite extension K⁑(P)/KK(P)/K, so that PP becomes a sum of rational points, using for model the minimal resolution of π”›βŠ—K​(P)∘\mathfrak{X}\otimes K(P)^{\circ} described earlier.

If ff is any continuous function on R⁑(𝔛)R(\mathfrak{X}) and DD a divisor on X\mathrm{X}, the metrized line bundle π’ͺ​(D+f)𝔛\mathscr{O}(D+f)_{\mathfrak{X}} is deduced from π’ͺ⁑(D)¯𝔛\overline{\mathscr{O}(D)}_{\mathfrak{X}} by multiplying the metric by eβˆ’fe^{-f}. When ff is piecewise linear, this metrized line bundle is smooth. To prove that, we may extend the scalars and assume that DD is a sum of rational points βˆ‘nj​Pj\sum n_{j}P_{j} and that ff is linear on each edge corresponding to an intersection point of components of the special fiber. Letting (Vi)(V_{i}) being the family of these components, and writing viv_{i} for the vertex of R⁑(𝔛)R(\mathfrak{X}) corresponding to ViV_{i}, the divisor

βˆ‘jnj​DPj+βˆ‘if⁑(vi)​Vi\sum_{j}n_{j}D_{P_{j}}+\sum_{i}f(v_{i})V_{i} (2.2.1)

defines the metrized line bundle π’ͺ​(D+f)𝔛\mathscr{O}(D+f)_{\mathfrak{X}}.

In this context, Zhang has defined a curvature operator, which associates to a metrized line bundle a distribution on the graph R⁑(𝔛)R(\mathfrak{X}), defined in such a way that

  • β€”

    for any divisor DD on X\mathrm{X}, curv⁑(π’ͺ⁑(D)¯𝔛)=ΞΌD{\operatorname{curv}}(\overline{\mathscr{O}(D)}_{\mathfrak{X}})=\mu_{D} ;

  • β€”

    for any continuous function ff, curv⁑(O¯​(f)𝔛)=βˆ’Ξ”β€‹f{\operatorname{curv}}(\overline{O}(f)_{\mathfrak{X}})=-\Delta f, where Ξ”\Delta is the Laplacian operator of the graph R⁑(𝔛)R(\mathfrak{X}),

and depending linearly on the metrized line bundle. The following lemma compares this construction with the general one on Berkovich spaces.

Lemma 2.2.2.

Let LΒ―=π’ͺ⁑(D+f)¯𝔛\overline{L}=\overline{\mathscr{O}(D+f)}_{\mathfrak{X}} be a metrized line bundle on X\mathrm{X} associated to a divisor DD on X\mathrm{X} and a continuous function ff on the graph R⁑(𝔛)R(\mathfrak{X}). If it is semi-positive, resp. admissible in the sense of [57] then it is semi-positive, resp. admissible in the sense of this article, and one has

c1​(LΒ―)=ΞΉβˆ—β€‹curv⁑(π’ͺ⁑(D+f)Β―)​log​|Ο€|βˆ’1.c_{1}(\overline{L})=\iota_{*}{\operatorname{curv}}(\overline{\mathscr{O}(D+f)})\log\left|{\pi}\right|^{-1}.

In other words, the measure c1​(LΒ―)c_{1}(\overline{L}) is supported by the graph R⁑(𝔛)R(\mathfrak{X}) where it coincides essentially with Zhang’s curvature.

DΓ©monstration.

We first assume that ff is linear on each edge of R⁑(𝒳)R(\mathscr{X}) and that DD is a sum of rational points of XX. Then, LΒ―\overline{L} corresponds to the line bundle 𝔏\mathfrak{L} on the model 𝔛\mathfrak{X} given by Equation 2.2.1. By definition, the measure c1​(𝔏)c_{1}(\mathfrak{L}) is computed as follows. It is a sum, for all components ViV_{i} of the special fiber, of deg⁑(𝔏|Vi)​log⁑|Ο€|βˆ’1\deg(\mathfrak{L}|V_{i})\log\left|{\pi}\right|^{-1} times the Dirac measure at the corresponding point viv_{i} of R⁑(𝔛)R(\mathfrak{X}). In particular, it is supported by R⁑(𝔛)R(\mathfrak{X}). Then,

deg⁑(𝔏|Vi)=βˆ‘jnj​{1ifΒ DPjΒ passes throughΒ ViΒ ;0otherwise}+βˆ‘jf⁑(Vj)​(Vi,Vj),\deg(\mathfrak{L}|V_{i})=\sum_{j}n_{j}\left\{\begin{array}[]{cc}1&\text{if $D_{P_{j}}$ passes through $V_{i}$ ;}\\ 0&\text{otherwise}\end{array}\right\}+\sum_{j}f(V_{j})(V_{i},V_{j}),

where (Vi,Vj)(V_{i},V_{j}) is the intersection number of the divisors ViV_{i} and VjV_{j}. That DPjD_{P_{j}} passes through ViV_{i} means exactly that ρ⁑(Pj)=vi\rho(P_{j})=v_{i}. Moreover, if jβ‰ ij\neq i, then (Vi,Vj)=mi,j(V_{i},V_{j})=m_{i,j} is just the number of intersection points of ViV_{i} and VjV_{j}, while

(Vi,Vi)=(Vi,βˆ‘jVj)βˆ’βˆ‘jβ‰ i(Vi,Vj)=βˆ’βˆ‘jβ‰ i(Vi,Vj),(V_{i},V_{i})=(V_{i},\sum_{j}V_{j})-\sum_{j\neq i}(V_{i},V_{j})=-\sum_{j\neq i}(V_{i},V_{j}),

since the whole special fiber is numerically equivalent to zero. Consequently,

βˆ‘jf⁑(vj)​(Vi,Vj)=βˆ‘jβ‰ imi,j​(f⁑(Vj)βˆ’f⁑(Vi)).\sum_{j}f(v_{j})(V_{i},V_{j})=\sum_{j\neq i}m_{i,j}\big(f(V_{j})-f(V_{i})\big).

Observe that this is the sum, over all edges from ViV_{i}, of the derivative of ff along this edge. Comparing with the definitions given by Zhang in [57], one finds, for any function gg on R⁑(𝔛)R(\mathfrak{X})

βˆ‘ideg⁑(𝔏|Vi)​g​(vi)\displaystyle\sum_{i}\deg(\mathfrak{L}|V_{i})g(v_{i}) =βˆ‘jnj​g​(ρ⁑(Pj))+βˆ‘iβŸ¨Ξ΄β€‹f​(vi),g⟩\displaystyle=\sum_{j}n_{j}g(\rho(P_{j}))+\sum_{i}\langle\delta f(v_{i}),g\rangle
=∫R⁑(𝔛)g⁑(ΞΌD+δ​f)\displaystyle=\int_{R(\mathfrak{X})}g\,(\mu_{D}+\delta f)
=∫R⁑(𝔛)g​curv⁑(π’ͺ⁑(D+f)Β―).\displaystyle=\int_{R(\mathfrak{X})}g\,{\operatorname{curv}}(\overline{\mathscr{O}(D+f)}).

This proves the claimed formula when ff is linear on each edge of 𝔛\mathfrak{X} and DD is a sum of rational points.

By working over an appropriate finite extension of KK, it extends to the case where ff is only piecewise linear, DD being any divisor on XX.

Zhang defines π’ͺ⁑(D+f)¯𝔛\overline{\mathscr{O}(D+f)}_{\mathfrak{X}} to be semi-positive if ff is uniform limit of piecewise linear functions fnf_{n} such that curv⁑(π’ͺ⁑(D+fn)¯𝔛)β‰₯0{\operatorname{curv}}(\overline{\mathscr{O}(D+f_{n})}_{\mathfrak{X}})\geq 0. The metrized line bundle LΒ―\overline{L} is then the limit of the metrized line bundles LΒ―n\overline{L}_{n} corresponding models 𝔏n\mathfrak{L}_{n} (on appropriate models 𝔛n\mathfrak{X}_{n} of X\mathrm{X} after some extension of scalars) of π’ͺ⁑(D)\mathscr{O}(D). By the previous computation, these metrics are smooth and c1​(LΒ―n)β‰₯0c_{1}(\overline{L}_{n})\geq 0. Reversing the computation, this means that 𝔏n\mathfrak{L}_{n} is numerically effective on 𝔛n\mathfrak{X}_{n}, hence LΒ―\overline{L} is semi-positive. By definition of the measure c1​(LΒ―)c_{1}(\overline{L}), one has

c1​(LΒ―)\displaystyle c_{1}(\overline{L}) =limnc1​(LΒ―n)=limnΞΉβˆ—β€‹curv⁑(π’ͺ⁑(D+fn)Β―)\displaystyle=\lim_{n}c_{1}(\overline{L}_{n})=\lim_{n}\iota_{*}{\operatorname{curv}}(\overline{\mathscr{O}(D+f_{n})})
=ΞΉβˆ—β€‹limncurv⁑(π’ͺ⁑(D+fn)Β―)=ΞΉβˆ—β€‹curv⁑(π’ͺ⁑(D+f)Β―).\displaystyle=\iota_{*}\lim_{n}{\operatorname{curv}}(\overline{\mathscr{O}(D+f_{n})})=\iota_{*}{\operatorname{curv}}(\overline{\mathscr{O}(D+f)}).

The case of an admissible metrized line bundle follows by linearity. ∎

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