betaFeedback

Reading v1View latest version

WorksMathematics

The p-adic section conjecture

Abstract

We prove the local p-adic section conjecture for smooth proper geometrically connected curves of genus at least two, over every finite extension of Qp\mathbb Q_p. Combined with established finite-descent theorems, this also yields the global section conjecture for smooth proper geometrically connected curves of genus at least two over number fields when their finite-cover descent locus equals their rational points; this includes X0(N)X_0(N) and X1(N)X_1(N) of genus at least two over Q\mathbb Q.

Introduction

Let pp be a prime, let k/Qpk/\mathbb{Q}_p be a finite extension, and let X/kX/k be a smooth proper geometrically connected curve of genus at least two. Choose an algebraic closure k‾\overline{k} and a geometric base point of XX. The full arithmetic étale fundamental group fits into

1⟶ΔX⟶ΠX⟶Gk⟶1,ΔX=π1(Xk‾),ΠX=π1(X),Gk=Gal⁡(k‾/k).(1)1 \longrightarrow\Delta_X \longrightarrow\Pi_X \longrightarrow G_k \longrightarrow1,\qquad\Delta_X=\pi_1(X_{\overline{k}}),\quad\Pi_X=\pi_1(X),\quad G_k=\operatorname{Gal}(\overline{k}/k). \tag*{(1)}

A rational point a∈X(k)a \in X(k) induces a continuous group-homomorphic section sa:Gk→ΠXs_a : G_k \to\Pi_X. Its class under conjugation by ΔX\Delta_X is independent of the path used to identify base points. The local pp-adic section conjecture asserts that this construction accounts for all sections, and distinguishes all rational points.

A section whose class comes from a rational point is called geometric. The conjecture thus asks whether rational points can be recovered from the splittings of the arithmetic fundamental group. It is a local analogue of the section conjecture in Grothendieck’s letter to Faltings of 27 June 1983, formulated over fields finitely generated over the prime field [9], equation (7) and the subsequent properness correction. The local question retains the full arithmetic and geometric groups in (1.1), while bringing the geometry of pp-adic valuations and reduction into the problem.

Theorem 1.1. For every prime pp, every finite extension k/Qpk/\mathbb{Q}_p, and every smooth proper geometrically connected curve X/kX/k of genus at least two, the map

X(k)⟶{s:Gk→π1(X) a continuous group homomorphismsuch that (π1(X)→Gk)∘s=id⁡Gk}/π1(Xk‾)X(k) \longrightarrow\left\{ \begin{array}{c} s : G_k \to\pi_1(X) \text{ a continuous group homomorphism} \\ \text{such that }(\pi_1(X) \to G_k) \circ s = \operatorname{id}_{G_k} \end{array} \right\} \bigg/ \pi_1(X_{\overline{k}})

is bijective. The quotient is by conjugation by the full geometric étale fundamental group.

Earlier results and the lifting problem

The birational analogue replaces the arithmetic étale fundamental group by the absolute Galois group of the curve’s function field. For a Galois extension with a chosen valuation, the decomposition group consists of the automorphisms preserving its valuation ring. Koenigsmann proves that every birational section over a finite extension of Qp\mathbb{Q}_p is contained in the decomposition group of a rational point [13], preprint version, Proposition 2.4(2). Pop subsequently obtains a birational criterion from sections of an elementary abelian quotient that lift to the corresponding elementary metabelian quotient [21], Theorems A and A′. These results make the passage from an étale section to a birational section a natural objective. Such a lift is additional information: a section of the quotient group in (1.1) does not by itself supply a splitting of the absolute Galois sequence.

Pop–Stix localize every étale section at a valuation extending the pp-adic valuation [22], preprint version, Theorem 26. This reduces the problem to understanding which valuations can support a section. Pajwani develops this valuative approach through étale homotopy and fixed points in Berkovich spaces [19], preprint version, Theorem 1.4 and Corollary 1.5. The passage from a fixed valuation to a rational point remains the geometric issue addressed here.

Recent results also constrain the possible nongeometric sections. Bresciani introduces toric sections and proves the section conjecture for them; they map injectively to the set of étale sections [7], Definition 3.7 and Theorems 1.4 and 5.10. For an arbitrary nongeometric étale section, his Theorem 1.5 produces a finite étale neighbourhood whose index is divisible by pp. Here a neighbourhood is a connected finite étale cover to which the section lifts, and the index of a curve is the greatest common divisor of the degrees of its closed points. In a different direction, Betts–Stix prove finiteness of the local image of global Selmer sections for smooth proper geometrically connected curves of genus at least two over number fields containing no CM subfield [4], preprint version, Theorem A. Their Selmer sections already arise from points at every completion; the finiteness assertion concerns the local images of this globally constrained set of sections.

Our lifting step is a statement about completions of the universal étale function field. It compares the absolute and étale decomposition groups at every rank-one valuation extending the pp-adic norm. This supplies the birational lift at precisely the valuations that must be excluded from the Pop–Stix localization.

The completion statement

Put K=k(X)K=k(X) and choose a separable closure KsepK^{\mathrm{sep}} containing kˉ\bar{k}. Let K~⊂Ksep\widetilde{K}\subset K^{\mathrm{sep}} be the compositum of the function fields of all connected finite étale covers of XX, with all their KK-embeddings in KsepK^{\mathrm{sep}}. Then Gal⁡(K~/K)=ΠX\operatorname{Gal}(\widetilde{K}/K)=\Pi_X after the chosen base point identification. A real-valued rank-one norm on KK means a nonarchimedean field absolute value with values in R≥0\mathbb{R}_{\geq0}; in particular its support is zero. We normalize an extension of the pp-adic norm by ∣p∣=p−1|p|=p^{-1}.

Theorem 1.2. Let pp be a prime, let k/Qpk/\mathbb{Q}_p be finite, and let X/kX/k be a smooth proper geometrically connected curve of genus at least two. Put K=k(X)K=k(X), choose a separable closure KsepK^{\mathrm{sep}}, and let K~⊂Ksep\widetilde{K}\subset K^{\mathrm{sep}} be the compositum of the function fields of all connected finite étale covers of XX, with all their KK-embeddings. For every real-valued rank-one norm on KK extending the normalized pp-adic norm on kk, and every prolongation to KsepK^{\mathrm{sep}}, set

E=Ksep^,H=K~^⊂E,E=\widehat{K^{\mathrm{sep}}},\qquad H=\widehat{\widetilde{K}}\subset E,

where the completions use that fixed prolongation. Then the specified subfield KsepK^{\mathrm{sep}} of EE is contained in HH.

The statement concerns the chosen copies of these fields inside one completion. Write Dν(L/K)D_\nu(L/K) for the decomposition group of a valuation ν\nu on a Galois extension L/KL/K. Corollary 7.1 uses Theorem 1.2 to identify the decomposition group in Ksep/KK^{\mathrm{sep}}/K with the one in K~/K\widetilde{K}/K. Indeed, an automorphism in the kernel fixes a dense subfield of HH and hence fixes KsepK^{\mathrm{sep}}.

A section localized at a rank-one valuation therefore lifts to a birational section. Koenigsmann’s theorem places this lifted section also at a rational point. Its fixed field would then carry two inequivalent henselian rank-one valuations, one extending the pp-adic valuation and the other trivial on constants. The contradiction is an explicit instance of the classical two-valuation principle associated with F. K. Schmidt [24]; Section 7 gives the needed approximation and square-root argument, including p=2p=2.

Rank-two valuations are handled by their rank-one coarsenings. The remaining case is a closed-point valuation refined by the pp-adic valuation on its residue field, called type 2h by Pop–Stix. The surjection of the section onto GkG_k forces that closed point to be rational, and their type-2h analysis then identifies the section [22], preprint version, Section 6.1. Section 7 completes this argument and proves injectivity of the point-to-section map by Kummer theory on the Jacobian.

Two stages of collar roots

The geometric content lies in Theorem 1.2. A finite extension of kˉ(X)\bar{k}(X) is the function field of a finite curve cover, which may have branch points. Near such a point, with local coordinate tt, a branch has the form t=zet=z^e. If another cover supplies a root of tt of order divisible by all these branch indices, the pulled-back cover splits over an annulus around the branch point. One can then replace the part over the inner disk by a trivial cover and glue to obtain an étale cover. The roots must exist on the entire inverse image of the annulus: a root on only one sheet does not trivialize the full cover needed for gluing. We use this operation twice, with different sources of collar roots.

This gluing and algebraization method belongs to the formal and rigid analytic patching tradition of Harbater and Liu [12, 15]. Poineau treats the corresponding Berkovich framework [20], Section 1 and Appendix A. Section 3 proves the precise replacement and field-embedding statements used here. In particular, evaluating the resulting exterior analytic map retains the specified valued field embedding; the map itself need not descend to kˉ\bar{k}.

Fix an odd prime ℓ≠p\ell\ne p. In the first stage, coordinates on the node annuli of a stable model acquire ℓ\ell-th roots on suitable finite étale covers. The construction uses the lifting theorem of Amini–Baker–Brugallé–Rabinoff, which realizes tame coverings of metrized complexes as curve covers [1], Theorem 7.7. A metrized complex records the skeleton together with the residue curve at each vertex and the marked points corresponding to edge ends. Small collars around arbitrary branch points need not themselves be stable node annuli. The resolution of nonsingularities theorem of Mochizuki–Tsujimura supplies a finite étale cover whose stable model maps to a prescribed semistable model [17], Theorem A and Proposition 2.4(iv). This extends Tamagawa’s result for closed points on a stable model [28], Theorem 0.2(v), and Lepage’s corresponding resolution theorem for Mumford curves [14], Theorem 2.7. The general theorem of Mochizuki–Tsujimura allows us to move these collars into the stable-node construction. Patching then places in HH every finite Galois extension whose horizontal ramification indices divide ℓ\ell. Here horizontal indices refer to ramification at closed points over kˉ\bar{k}. Section 4 proves this first stage.

The second stage produces roots of arbitrary positive order mm while keeping the auxiliary cover’s horizontal indices dividing ℓ\ell. We construct a cyclic cover of the projective line whose skeleton has two vertices and ℓ\ell parallel edges. A cycle using two distinct edges gives a character of the torus uniformizing its Jacobian. Along a chosen annulus that character has slope one, so it is a constant times the annular coordinate times an analytic unit close to one. Figure 1 shows the geometry. This uses the nonarchimedean uniformization theory of Raynaud and Bosch–Lütkebohmert [23, 6] and the compatibility of classical and tropical Abel–Jacobi maps proved by Baker–Rabinoff [3], Proposition 6.1 and Corollary 6.6.

Schematic construction showing parallel edges of the skeleton and the middle collar on one edge

Figure 1. The construction in Proposition 5.2, shown schematically. All edges are oriented by increasing v(T)v(T); the cycle ej−ej′e_j-e_{j'} gives a character unit uju_j of slope one on the annulus corresponding to eje_j. Division by mm on the Jacobian supplies a root of uju_j on the whole inverse annulus. Trimming its skeletal interval to D<v(T)<a−DD<v(T)<a-D makes hjh_j small enough for 1+hj1+h_j to have an mm-th root, and hence gives a root of TT.

Pullback by multiplication by mm on the Jacobian supplies a root of the character on the whole inverse annulus. On a sufficiently long middle collar, the remaining unit has an mm-th root by the pp-adic logarithm and exponential. This yields a root of the coordinate itself. The auxiliary covers still have horizontal indices dividing ℓ\ell, so the first stage places their function fields in HH. Section 5 constructs these roots, and a second application of patching in Section 6 realizes arbitrary finite Galois extensions in the same completion. The common requirement of roots on every inverse-image sheet is what allows both stages to use the same patching proposition.

Global consequences

The local theorem also identifies the possible localizations of global sections. Let C/FC/F be a smooth proper geometrically connected curve of genus at least two over a number field. A global section determines a point of C(Fv)C(F_v) at every finite place by Theorem 1.1, and a connected component of C(Fv)C(F_v) at every real place by the real section theorem [8], Introduction, Theorem. Together these form a modified adelic point; complex places contribute no factor. The finite-cover descent locus consists of the tuples that lift to some global twist of each torsor under a finite FF-group scheme. Proposition 8.1 shows that this locus is exactly the image of global sections under localization. Both inclusions use the descent bridge of Harari–Stix [11].

When this finite-cover locus consists exactly of the diagonal rational points, Stix’s finite-support theorem identifies every global section with a point section [25], arXiv version, Corollary 5. The resulting bijection is Theorem 8.2. Stoll’s finite-descent criteria supply the required equality when the curve has a nonconstant morphism over its number field to an abelian variety with finitely many rational points and trivial divisible Tate–Shafarevich subgroup, or when its Jacobian has these properties. They also cover the smooth projective modular curves X0(N)X_0(N) and X1(N)X_1(N) of genus at least two over Q\mathbb{Q} [26], Corollary 8.1, Theorem 8.6, and Corollary 8.8, with the corrections in [27]. The Selmer converse in the companion article [18], Theorem 1.1, gives a further arithmetic test for an elliptic target over Q\mathbb{Q}: vanishing of its full pp-power Selmer corank at one prime. Section 8 states the precise hypotheses and deductions for these cases.

The localization description also yields constraints without assuming that the finite-cover locus consists of rational points. Every ordinary adelic representative of a localization tuple survives linear descent and lies in the Brauer–Manin set, by further results of Harari–Stix [11], preprint version, Corollary 3.1 and Theorem 4.1. If FF contains no CM subfield, Betts–Stix’s theorem makes the projection of the localization image to C(Fv)C(F_v) finite at every finite place vv [4], preprint version, Theorem A. These statements constrain the localization image; the finite-cover equality remains a separate arithmetic hypothesis in the global section-conjecture criterion.

Fields, analytic points, and horizontal ramification

Fix the data and the prolonged norm of Theorem 1.2. All subsequent field inclusions refer to these choices. We have

K⊂K0:=k‾(X)⊂K~⊂Ksep⊂E,H=K~^⊂E.K \subset K_0 := \overline{k}(X) \subset\widetilde{K} \subset K^{\mathrm{sep}} \subset E,\qquad H = \widehat{\widetilde{K}} \subset E.

The inclusion of the completion HH into EE is isometric: it identifies HH with the closure of K~\widetilde{K} in EE. Since kk is complete, the restriction of the chosen norm to k‾\overline{k} is its unique prolongation. Thus Ω:=k‾^\Omega:= \widehat{\overline{k}} is a specified subfield of HH. It is algebraically closed and has additive value group Q\mathbb{Q}, with the normalization v=−log⁡p∣⋅∣v=-\log_p|\cdot|. The equality of value groups before and after completion follows because every nonzero limit has the same norm as sufficiently close approximants.

We use Berkovich analytifications over Ω\Omega. A field embedding K0↪HK_0 \hookrightarrow H determines a point of Xk‾(H)X_{\overline{k}}(H) supported at the generic point of Xk‾X_{\overline{k}}. Together with the specified embedding Ω↪H\Omega\hookrightarrow H, it determines a point of XΩ(H)X_{\Omega}(H), whose analytic image we denote by ξ\xi. Its projection to Xk‾X_{\overline{k}} is generic; its support on XΩX_{\Omega} need not be generic. In particular, ξ\xi can be an Ω\Omega-rational point. We retain the whole HH-valued point, including the embedding of its completed residue field into HH, whenever we use ξ\xi.

If F/K0F/K_0 is finite, its normalization gives a smooth proper connected curve over k‾\overline{k} with a finite map to Xk‾X_{\overline{k}}. We call the ramification indices of this map at closed points horizontal ramification indices. The corresponding discrete valuations are trivial on k‾\overline{k} and have residue characteristic zero. These indices are unrelated to possible wild ramification on a characteristic-pp reduction model.

Fix an odd prime ℓ≠p\ell\ne p. Inside KsepK^{\mathrm{sep}}, let LℓL_\ell be the compositum of all finite Galois extensions F/K0F/K_0 whose horizontal ramification indices divide ℓ\ell. The unramified extensions are included. In particular K~⊂Lℓ\widetilde{K} \subset L_\ell: the universal étale field over K0K_0 is the same field K~\widetilde{K} after constants are included. One may see this by descending any finite étale cover of Xk‾X_{\overline{k}} to a finite extension of kk and then composing with the finite étale constant-field cover of XX.

The field-theoretic goal can be displayed as follows. The top row consists of the defining inclusions; the two right vertical arrows are the containments proved in the successive stages.

K~→Lℓ→Ksep↓↓first stage↓second stageH=H=H.\begin{CD} \widetilde{K} @>>> L_\ell @>>> K^{\mathrm{sep}} \\ @VVV @VV{\text{first stage}}V @VV{\text{second stage}}V \\ H @= H @= H. \end{CD}

Every arrow is inside EE. In particular a constructed abstract field embedding is not enough by itself: for a normal extension F/K0F/K_0 we will use normality to identify its image with the specified subfield F⊂EF \subset E.

We shall also use invariance of finite étale covers of a proper variety under extension of algebraically closed fields [10], Exposé X, Corollary 1.8. Thus a finite étale cover of YΩY_\Omega, for a smooth proper curve Y/k‾Y/\overline{k}, descends up to isomorphism to a cover of YY. The analytic maps built on smaller open regions need not descend. Their evaluation at the fixed HH-valued point will be sufficient.

Replacing branch disks by trivial covers

We first isolate the analytic operation used in both stages of the proof. It replaces a branched cover by an étale cover after another cover supplies roots of the boundary coordinates. The roots must exist over the entire inverse image of each boundary annulus. The resulting analytic map is needed only outside the branch disks. This is an application of the patching method for covers developed in formal geometry by Harbater, in rigid geometry by Liu, and in the Berkovich setting by Poineau [12, 15, 20]. Here the root on every inverse collar supplies the trivialization needed to replace the whole branched cover over a disk, including all of its sheets.

Lemma 3.1. Let f:C→Xkˉf:C\to X_{\bar{k}} be a finite surjective morphism of smooth proper connected curves, and let b∈X(kˉ)b\in X(\bar{k}). Fix a rational function tbt_b which is a local parameter at bb. There is an open disk BbB_b about bb in XΩanX_{\Omega}^{\mathrm{an}}, with coordinate tbt_b, such that its full inverse image in CΩanC_{\Omega}^{\mathrm{an}} is a disjoint union of open disks. On the disk about c∈f−1(b)c\in f^{-1}(b) there is a coordinate zcz_c for which

tb=zcec.t_b=z_c^{e_c}.

where ece_c is the ramification index at cc. No condition that ece_c be prime to pp is required. For finitely many distinct points bb, the disks can be chosen disjoint and disjoint from the fixed point ξ\xi.

Proof. Choose a local analytic parameter uu at cc. The pullback of tbt_b is ueca(u)u^{e_c}a(u) with a(0)≠0a(0)\ne0. After choosing an ece_c-th root of a(0)a(0) in Ω\Omega and shrinking the disk, a(u)/a(0)a(u)/a(0) is sufficiently close to 11 for its ece_c-th root to be given by

exp⁡(ec−1log⁡(a(u)/a(0))).\exp\left(e_c^{-1}\log\left(a(u)/a(0)\right)\right).

Here it is enough to make the valuation of a(u)/a(0)−1a(u)/a(0)-1 greater than vp(ec)+1/(p−1)v_p(e_c)+1/(p-1). Thus zc=u a(u)1/ecz_c=u\,a(u)^{1/e_c} is analytic with a nonzero derivative at the center. Analytic inversion gives a smaller disk on which it is a coordinate and tb=zcect_b=z_c^{e_c}.

Do this in disjoint neighborhoods of all the finitely many points above bb. The complement of these neighborhoods in the proper analytic curve CΩanC_{\Omega}^{\mathrm{an}} is compact. Its image under ff is compact, closed, and does not contain bb. Consequently a sufficiently small disk about bb has its entire inverse image in the chosen neighborhoods. Shrink it once more so that, in every zcz_c-chart, that inverse image is the disk defined by the corresponding bound on ∣zc∣ec|z_c|^{e_c}.

The point ξ\xi does not equal any point of X(kˉ)X(\bar{k}), since its projection to XkˉX_{\bar{k}} is generic. The shrinking disks at bb form a neighborhood basis there. Hausdorffness therefore permits all disks to avoid ξ\xi and one another. We may choose their radii in ∣Ω×∣|\Omega^\times|. The argument also applies when ξ\xi is an Ω\Omega-rational point whose projection to XkˉX_{\bar{k}} is generic. □

Proposition 3.2. Let f:C→Xkˉf:C\to X_{\bar{k}} be as in Lemma 3.1, of degree dd, with branch set BB. Choose pairwise disjoint disks BbB_b as in that lemma. For each b∈Bb\in B, let mbm_b be a positive multiple of all ramification indices above bb. Suppose that g:Y→XΩg:Y\to X_{\Omega} is a finite surjective morphism from a smooth proper connected curve and that there are radii 0<rb<Rb0<r_b<R_b in ∣Ω×∣|\Omega^\times|, strictly inside BbB_b, such that g∗tbg^*t_b has an analytic mbm_b-th root on the whole inverse image of

Ab={rb<∣tb∣<Rb}.A_b=\{r_b<|t_b|<R_b\}.

Set

U=XΩan∖⋃b∈B{∣tb∣≤rb},U=X_{\Omega}^{\mathrm{an}}\setminus\bigcup_{b\in B}\{|t_b|\le r_b\},

where each indicated closed disk is taken inside BbB_b. Then there is a smooth proper connected curve VΩV_{\Omega} and a finite étale surjective map VΩ→YV_{\Omega}\to Y, together with an analytic map

(VΩan)∣U⟶CΩanover XΩan.\left(V_{\Omega}^{\mathrm{an}}\right)|_U\longrightarrow C_{\Omega}^{\mathrm{an}}\quad\text{over }X_{\Omega}^{\mathrm{an}}.

The notation on the left denotes the inverse image of UU under the composite map to XΩanX_{\Omega}^{\mathrm{an}}. Proof. The inverse image YUY_U of UU and the inverse images YbY_b of the disks {∣tb∣<Rb}\{|t_b|<R_b\} form an open cover of YanY^{\mathrm{an}}. Distinct YbY_b are disjoint, and YU∩Yb=g−1(Ab)Y_U\cap Y_b=g^{-1}(A_b).

On YUY_U, take the pullback of CΩan→XΩanC_\Omega^{\mathrm{an}}\to X_\Omega^{\mathrm{an}}. It is finite étale of degree dd, since UU contains no branch point. On each YbY_b, take the trivial degree-dd cover. These two covers are isomorphic over their overlap. Indeed, write qbmb=g∗tbq_b^{m_b}=g^*t_b for the assumed analytic root. In the local factor corresponding to c∈f−1(b)c\in f^{-1}(b), put ac=qbmb/eca_c=q_b^{m_b/e_c}. The polynomial defining that factor splits as

Zec−g∗tb=∏ζ∈μec(Z−ζac).Z^{e_c}-g^*t_b=\prod_{\zeta\in\mu_{e_c}}(Z-\zeta a_c).

Each aca_c is an analytic unit on the overlap. Distinct linear factors are comaximal, because (ζ−ζ′)ac(\zeta-\zeta')a_c is a unit. This is true also when p∣ecp\mid e_c: a nonzero constant in Ω\Omega is invertible, regardless of its absolute value. Taking all factors in Lemma 3.1 gives a trivialization of the full degree-dd cover, not just of one sheet.

Choose these trivializations and glue the finite locally free analytic algebras. There is no further cocycle condition between different branch disks, as those disks are disjoint. The glued algebra is coherent and finite étale over YanY^{\mathrm{an}}, these being local properties. Proper analytic GAGA [20], Appendix A, Theorem A.1(ii), algebraizes the coherent algebra and its unit and multiplication; full faithfulness preserves the algebra identities. Equivalently, one uses the corresponding equivalence for finite covers of a proper curve. Étaleness is detected after analytification [29], Fact 5.1.5, so its relative spectrum is a finite étale cover of YY.

Choose any nonempty connected component and call it VΩV_\Omega. A connected component of this finite étale cover has positive locally constant degree over the connected curve YY, hence is still surjective. It is smooth and proper. Restricting the original projection from the pullback over YUY_U gives the asserted analytic map. If BB is empty, the whole pullback is already finite étale and the same component argument applies without gluing.

We record explicitly how this analytic map produces an embedding into the fixed completion HH. This step concerns the entire HH-valued point, not just its underlying analytic point.

Lemma 3.3. In Proposition 3.2, suppose that the composite VΩ→XΩV_\Omega\to X_\Omega is isomorphic, over XΩX_\Omega, to the base change of a finite morphism V→Xk‾V\to X_{\overline{k}} from a smooth proper connected curve, and assume that ξ∈U\xi\in U. If k‾(V)\overline{k}(V) admits a K0K_0-embedding into HH, then k‾(C)\overline{k}(C) admits a K0K_0-embedding into HH.

Proof. The given embedding determines an HH-valued point of VV. Together with the fixed inclusion Ω⊂H\Omega\subset H, it determines an HH-valued point of V×k‾ΩV\times_{\overline{k}}\Omega, and the chosen isomorphism transports this point to VΩV_\Omega. Its image in XΩ(H)X_\Omega(H) is exactly the fixed point induced by K0↪HK_0\hookrightarrow H and Ω↪H\Omega\hookrightarrow H. In particular its underlying analytic image is ξ∈U\xi\in U.

An HH-valued point of an algebraic variety over Ω\Omega is equivalently an HH-valued point of its analytification [29], Fact 5.1.3. More explicitly, it gives an analytic point vv together with an isometric embedding of the completed residue field H(v)\mathcal{H}(v) into HH. The analytic map over UU in Proposition 3.2 induces a compatible map on completed residue fields, and hence an HH-valued point of CΩC_\Omega over the same point of XΩ(H)X_\Omega(H). Project this point to CC. Its image on Xk‾X_{\overline{k}} is the generic point with the specified map K0→HK_0\to H. Since C→Xk‾C\to X_{\overline{k}} is finite, its support on CC is also generic. The point therefore induces the required field embedding k‾(C)↪H\overline{k}(C)\hookrightarrow H over K0K_0.

Neither the analytic map over UU nor a chosen local trivialization has been asserted to descend to k‾\overline{k}.

We shall form auxiliary covers by pullback and compositum, then pass to a finite étale upper cover. The following lemma shows that these operations preserve the bound on horizontal ramification needed for membership in LℓL_\ell.

Lemma 3.4. Fix a positive integer nn. Among finite extensions of K0K_0, the condition that every ramification index at a closed point of Xk‾X_{\overline{k}} divide nn is preserved by taking subextensions, finite composita, and Galois closures. It is also preserved by normalized pullback along a finite morphism of smooth proper curves and by composition on top with a finite étale cover. Consequently every finite cover of Xk‾X_{\overline{k}} with indices dividing ℓ\ell has a K0K_0-embedding of its function field into LℓL_\ell.

Proof. All valuations in this statement are the horizontal, discrete closed-point valuations, which are trivial on k‾\overline{k}. Their residue characteristic is zero. The completed base field at such a point is A=k‾((t))A=\overline{k}((t)). Every finite extension of AA has residue field k‾\overline{k} and is of the form k‾((u))\overline{k}((u)) for a uniformizer uu: successive subtraction of residue representatives gives the power-series expansion in its complete valuation ring. If its ramification index is ee, then t=uea(u)t=u^e a(u) for a unit a(u)∈k‾[[u]]×a(u)\in\overline{k}[[u]]^\times. The unit has an ee-th root in k‾[[u]]\overline{k}[[u]], by Hensel’s lemma in residue characteristic zero. Changing the parameter gives t=zet=z^e. Thus, inside a fixed algebraic closure, the local field is A(t1/e)A(t^{1/e}), which is cyclic over AA since k‾\overline{k} contains all roots of unity.

For e∣ne\mid n, this is a subfield of A(t1/n)A(t^{1/n}). All local factors of a finite compositum, or of a normal closure formed from conjugates, therefore remain subfields of this same cyclic extension. This proves the assertions about subextensions, composita, and Galois closures.

For pullback, write the local base map as t=sra(s)t=s^r a(s), where r≥1r\ge1 and a(s)a(s) is a unit. Extracting its ee-th root in k‾[[s]]\overline{k}[[s]] shows that every normalized local factor after pullback has ramification index e/gcd⁡(e,r)e/\gcd(e,r), which divides ee. Finally, ramification indices multiply in a tower, and an étale upper cover contributes index 1. Apply the Galois-closure assertion to a cover with indices dividing ℓ\ell and embed that closure into the fixed KsepK^{\mathrm{sep}}. Its field lies in the defining compositum LℓL_\ell.

Remark 3.5. Lemma 3.4 makes no assertion about tame reduction in residue characteristic pp, nor about composing two arbitrarily ramified covers. Its last assertion about composition requires the upper cover to be étale.

Prime-to-pp roots from stable reduction

We first treat the extensions whose horizontal ramification indices divide the fixed prime ℓ≠p\ell\ne p. The patching construction reduces this task to extracting ℓ\ell-th roots of coordinates on collars around the branch points. We shall find these roots in finite étale covers of XΩX_\Omega. There are two steps: extract roots on stable node annuli, and use resolution of nonsingularities to reach such annuli from arbitrary small disks.

Units and roots on node annuli

We use the following elementary analytic fact. The exponent in its factorization is the slope of the additive valuation of the unit along the annulus skeleton. The unit factorization is the standard annular factorization; compare [2], Proposition 2.2(1). We give the argument on the entire open annulus, since this is the domain required for gluing.

Lemma 4.1 (Units on disks and annuli). An analytic unit on an open disk over Ω\Omega has constant absolute value in ∣Ω×∣|\Omega^\times|. If A={r<∣S∣<R}A=\{r<|S|<R\} is an open annulus and u∈O(A)×u\in\mathcal{O}(A)^\times, then

u=cSd(1+h),c∈Ω×,d∈Z,h∈O(A),∣h(x)∣<1(x∈A),u=cS^d(1+h), \qquad c\in\Omega^\times,\quad d\in\mathbb{Z},\quad h\in\mathcal{O}(A),\quad|h(x)|<1\quad(x\in A),

where cSdcS^d is a term of the Laurent expansion of uu. The function 1+h1+h has an analytic ℓ\ell-th root on AA. Consequently, on any analytic cover of AA on which SS has an ℓ\ell-th root, so does uu.

Proof. On a circle ∣S∣=ρ|S|=\rho of radius ρ∈∣Ω×∣\rho\in|\Omega^\times|, the Laurent Gauss norm is multiplicative. Normalize uu and its inverse to have norm one. Their reductions are inverse units in Ω~[s,s−1]\widetilde{\Omega}[s,s^{-1}], so the reduction of uu is a monomial. Thus exactly one Laurent term dominates at that radius. Its exponent cannot change inside the radius interval. Indeed, convergence on compact subannuli reduces any proposed change to finitely many competing terms; a tie between two terms occurs at a radius in ∣Ω×∣|\Omega^\times|, since the additive value group is Q\mathbb{Q}. Such a tie contradicts the preceding reduction argument. A single term cSdcS^d therefore dominates throughout AA, giving the asserted factorization. The same argument on closed subdisks uses that the only units in Ω~[s]\widetilde{\Omega}[s] are constants and proves the disk assertion. Its constant norm belongs to ∣Ω×∣|\Omega^\times| by evaluation at a rational point of the disk.

Since ℓ\ell is a unit in the valuation ring, the binomial series for (1+h)1/ℓ(1+h)^{1/\ell} converges on every compact subannulus; the resulting functions agree on overlaps. A root of cc lies in Ω\Omega, and a root of SdS^d follows from a root of SS, also when d<0d<0.

Recall the data attached to a split stable model of a proper curve. Its skeleton is a finite metric graph. Each vertex carries the smooth normalization of the corresponding special-fibre component, with one marked point for each incident edge end. A loop has two distinct marks on the same normalization. The tube of a node is an open annulus, and the tube of a smooth special-fibre point is an open disk. The graph with these marked residue curves is its metrized complex. Stability means

2g(v)−2+val⁡(v)>02g(v) - 2 + \operatorname{val}(v) > 0

at each vertex vv, where g(v)g(v) is the genus of its residue curve.

Lemma 4.2 (Roots on stable node annuli). Let C/ΩC/\Omega be a smooth proper connected curve of genus at least two, equipped with its stable skeleton. Let AA be a node annulus and SS an annular coordinate on AA. There is a connected finite étale cover C′→CC' \to C such that SS has an analytic ℓ\ell-th root on the entire inverse image of AA. A single such cover can be chosen for any finite collection of node annuli.

Proof. Our aim is to make the expansion factors, the absolute values of the integral slopes along edges, divisible by ℓ\ell above the chosen edge. On an inverse annulus this makes the Laurent exponent of the pulled-back coordinate divisible by ℓ\ell; its constant factor and the factor 1+h1+h also have ℓ\ell-th roots by Lemma 4.1. We construct such a cover directly when the edge lies on a circuit. For a separating edge, we first pass to an unramified cover in which every edge above it lies on a circuit.

We use the lifting theorem for tame coverings of metrized complexes [1], Theorem 7.7. A covering to which it applies consists of finite residue-curve maps and a finite harmonic graph map: every fibre of every marked point is represented by edge ends, their ramification indices equal the edge expansion factors, and the component degree is the sum of these factors over each base direction. The residue maps must be tamely ramified, and the augmented graph ramification must vanish away from punctures [1], Definitions 2.19 and 2.21. We shall use no punctures. The lift is then finite étale, and has exactly the prescribed inverse images of the vertex set and skeleton [1], Definitions 4.25 and 4.31.

First suppose that the edge ee corresponding to AA is nonseparating. Choose a simple circuit containing ee. At each vertex on this circuit, let P,QP,Q be the two circuit marks on its residue curve DD. Divisibility of Pic⁡0(D)\operatorname{Pic}^0(D) [16], Theorem 8.2 supplies a degree-zero divisor EE such that P−Q+ℓEP-Q+\ell E is principal. Extract an ℓ\ell-th root of a function with this divisor and take the smooth normalization. The resulting cyclic cover of DD has degree ℓ\ell, is connected because the order at PP is prime to ℓ\ell, and is totally ramified at P,QP,Q and nowhere else. Each other marked point has ℓ\ell distinct preimages. At every vertex outside the circuit take ℓ\ell identity copies of its residue curve.

Above each circuit edge use one edge of expansion factor ℓ\ell and of length equal to the original length divided by ℓ\ell. Above each remaining edge use ℓ\ell factor-one edges, matching all the unramified marked fibres at its ends. These rules also handle loops and chords between circuit vertices. All lengths belong to Q\mathbb{Q}. At a lifted circuit vertex each base direction has total degree ℓ\ell; at an identity-copy vertex it has total degree one. The local Riemann–Hurwitz formula is

2g(D′)−2=ℓ(2g(D)−2)+2(ℓ−1).2g(D') - 2 = \ell(2g(D) - 2) + 2(\ell- 1).

The two extra terms are precisely those recorded by the two circuit edges. Thus the augmented graph ramification is zero at every vertex; the identity copies contribute none. All ramification indices are prime to pp. The resulting tame covering is connected, because each vertex outside the circuit can be joined by a lifted path to a circuit vertex.

Lift these data by the cited theorem. The inverse image of AA is exactly the disjoint union of the node annuli above ee. For completeness, the inverse-skeleton equality excludes off-skeleton points mapping to the skeleton. Each off-skeleton open ball therefore maps into a single off-skeleton ball downstairs, and continuity identifies the images of their attachment points. Retractions commute on these balls as well as on the skeleton. Since a node annulus is the inverse image of its open edge under retraction, this proves the asserted equality for the whole annulus. On each annulus above AA, the pullback of SS has exponent ℓ\ell or −ℓ-\ell in Lemma 4.1. It therefore has an analytic ℓ\ell-th root on each of these disjoint annuli.

Now suppose that ee is separating. Each side of the graph obtained by cutting its interior contains a circuit or a positive-genus vertex. Otherwise that side would be a genus-zero tree with, say, NN vertices. Its total valence, including the bridge half-edge, would be 2N−12N - 1, contrary to stability, which requires at least 3N3N.

Construct a connected unramified degree-ℓ\ell covering of each side, retaining its boundary mark for the eventual gluing. If the side contains a circuit, take a connected cyclic graph covering of degree ℓ\ell, with identity residue-curve maps. Such a graph covering is obtained by a homomorphism of its graph fundamental group onto Z/ℓZ\mathbb{Z}/\ell\mathbb{Z}. If instead a positive-genus vertex is used, take a connected cyclic étale cover of degree ℓ\ell of its residue curve, provided by a nonzero element of its Jacobian’s ℓ\ell-torsion [16], Theorem 8.2 and Remark 8.4. Explicitly, for a line bundle LL of exact order ℓ\ell, a trivialization L⊗ℓ≃OL^{\otimes\ell} \simeq\mathcal{O} makes ⨁j=0ℓ−1L⊗j\bigoplus_{j=0}^{\ell-1} L^{\otimes j} an algebra whose relative spectrum is a cyclic étale cover: ℓ\ell is invertible, and the exact order of LL makes the cover connected. Use ℓ\ell identity copies at the other vertices and match all marked fibres. Paths to the distinguished vertex show connectedness of the covered side.

Each covered side has ℓ\ell preimages of the bridge mark. Match them in pairs to obtain ℓ\ell distinct bridge edges, all with factor one. Every such edge is now nonseparating: after removing one of them, the two connected covered sides are still joined by the others. The resulting covering of the original metrized complex is tame. Indeed the component covers are unramified, all factors are one, and Riemann–Hurwitz again gives zero graph ramification. Lift it to a finite étale curve cover. Its triangulation remains stable: at a vertex of local degree dd, the quantity 2g′−2+val⁡(v′)2g' - 2 + \operatorname{val}(v') is d(2g−2+val⁡(v))>0d(2g - 2 + \operatorname{val}(v)) > 0.

Apply the nonseparating construction to each of these ℓ\ell edges. A connected component of the fibre product of the resulting finite étale covers dominates each factor, so all the constructed roots persist on a common cover. On each factor-one annulus the pullback of the original SS is a unit; Lemma 4.1 supplies its root once the new annular coordinate has a root. This proves the assertion also for separating edges. Finally, the same finite fibre-product argument treats finitely many original annuli simultaneously.

Moving disk collars into the stable skeleton

The stable-edge lemma does not apply directly to a small disk around a branch point: nodes introduced there by blowups need not survive stabilization. Resolution of nonsingularities supplies an étale cover whose stable model maps to the chosen model with those nodes. The next argument requires only that model morphism.

Lemma 4.3 (Étale covers with roots on collars). Let C/k‾C/\overline{k} be a smooth proper connected curve of genus at least two. Let b1,…,bs∈C(k‾)b_1,\ldots,b_s \in C(\overline{k}) be distinct, and let BiB_i be pairwise disjoint sufficiently small coordinate disks in CΩanC_{\Omega}^{\mathrm{an}} about bib_i, with coordinates tit_i defined over k‾\overline{k} and vanishing simply at bib_i. There is a connected finite étale cover Y→CΩY \to C_{\Omega} and concentric open collars

Ai′={ri<∣ti∣<Ri}⊂Bi,ri,Ri∈∣Ω×∣,0<ri<Ri,A_i'=\{r_i<|t_i|<R_i\}\subset B_i,\qquad r_i,R_i\in|\Omega^\times|,\qquad0<r_i<R_i,

such that tit_i has an analytic ℓ\ell-th root on the entire inverse image of Ai′A_i' in YanY^{\mathrm{an}}.

Proof. All algebraic data descend to a finite extension of kk. After a further finite extension, choose a split semistable model with the bib_i as disjoint smooth sections. Shrinking the coordinate disks causes no problem, since the desired collars need only lie inside the originally specified disks. In an integral local parameter uiu_i at a section, successive blowups at its reductions introduce node annuli

∣π∣j+1<∣ui∣<∣π∣j.|\pi|^{j+1}<|u_i|<|\pi|^j.

On a sufficiently small disk the quotient ti/uit_i/u_i is an analytic unit, since both coordinates have a simple zero at bib_i. Its absolute value is constant by Lemma 4.1. Thus these annuli are also concentric annuli for tit_i, and choosing jj large puts them inside BiB_i. Perform these modifications for all the sections and denote the resulting proper semistable model by C\mathcal C. It need not be stable; the markings can now be forgotten.

The RNS theorem, applied with the set of all primes, and its model-domination consequence [17], Theorem A and Proposition 2.4(iv), give, after finite extension of the ground field, a finite étale cover whose stable model maps to C\mathcal C. Pass to split reduction and a geometric connected component, and write

f:Y0⟶CΩ,Y0⟶Cf:Y_0\longrightarrow C_{\Omega},\qquad\mathcal Y_0\longrightarrow\mathcal C

for the analytic-base-changed curve cover and the model morphism. No finiteness assertion about the model morphism is needed.

Fix one of the target node annuli Ai⊂BiA_i\subset B_i just constructed, and choose an irrational real number aa strictly between its two additive endpoint valuations. Consider the entire inverse image

f−1({x∈Ai:−log⁡p∣ti(x)∣=a}).f^{-1}\left(\{x\in A_i:-\log_p|t_i(x)|=a\}\right).

Reduction is compatible with the model morphism. A point reducing to a component generic point of Y0\mathcal Y_0 is a model vertex, hence is type 2 and gives tit_i a value in ∣Ω×∣|\Omega^\times|; its additive valuation is rational. It cannot lie above this level. A point reducing to a smooth closed point lies in an open disk tube. If it lies above the chosen level, that smooth point maps to the target node, so the entire disk tube maps into AiA_i. The pullback of tit_i is a unit there. By Lemma 4.1 its norm is constant and belongs to ∣Ω×∣|\Omega^\times|, which again excludes the irrational level. All remaining points above the level lie in source node annuli whose nodes map to the target node.

Let NiN_i be the union of all source node annuli whose nodes map to the target node defining AiA_i; this is a finite union. The complement of NiN_i in the proper analytic curve Y0anY_0^{\mathrm{an}} is compact. Intersect it with the inverse image of a closed concentric subannulus of AiA_i containing the chosen level in its interior. On this compact intersection the continuous real function −log⁡p∣ti∣-\log_p|t_i| has compact image avoiding aa. Consequently there is an open interval about aa with rational endpoints whose entire inverse image lies in NiN_i. It determines a collar Ai′⊂AiA_i'\subset A_i.

Every source node annulus in NiN_i maps into AiA_i by reduction compatibility, so the pullback of tit_i is an analytic unit on it. Apply Lemma 4.2 simultaneously to all these source annuli, for all ii. Their coordinates acquire ℓ\ell-th roots on a common connected finite étale cover of Y0Y_0. Lemma 4.1 then gives roots of the pullbacks of the tit_i on the full inverse images of the Ai′A_i'. The resulting composite cover of CΩC_{\Omega} is finite étale. □

Proposition 4.4 (The first stage). The field LℓL_{\ell} is contained in HH. Proof. Let F/K0F/K_0 be a finite Galois extension whose horizontal ramification indices divide ℓ\ell, and let C→Xk‾C \to X_{\overline{k}} be its smooth proper curve cover. If it is unramified, its function field is already contained in K~\widetilde{K}. Otherwise take disjoint branch-point disks avoiding ξ\xi, small enough for Lemma 3.1. Lemma 4.3, applied on Xk‾X_{\overline{k}}, supplies a finite étale Y→XΩY \to X_{\Omega} and collars on whose full inverse images the branch coordinates have ℓ\ell-th roots.

Apply Proposition 3.2 with these roots. It gives a connected finite étale cover VΩ→YV_{\Omega} \to Y and an analytic map to CΩC_{\Omega} over the exterior of the smaller branch disks. The composite VΩ→XΩV_{\Omega} \to X_{\Omega} is finite étale. Proper base-extension invariance of finite étale covers descends it, up to isomorphism over XΩX_{\Omega}, to a cover V→Xk‾V \to X_{\overline{k}}. Here we use [10], Exposé X, Corollary 1.8. Its function field has a K0K_0-embedding into K~⊂H\widetilde{K} \subset H. Lemma 3.3 therefore yields a K0K_0-embedding F↪HF \hookrightarrow H at the fixed HH-valued point above ξ\xi. Since F/K0F/K_0 is normal and both fields lie in EE, this says F⊂HF \subset H for the chosen copy of FF. Taking their compositum proves the proposition.

Arbitrary roots on a collar

The first stage used roots of order ℓ\ell, a prime different from pp. To treat arbitrary ramification, we now construct an auxiliary branched cover on which a coordinate has a root of any prescribed positive order. Its own horizontal ramification indices still divide ℓ\ell. The construction uses a curve whose stable skeleton has parallel edges and multiplication on its Jacobian. Throughout this section, v=−log⁡p∣⋅∣v=-\log_p|\cdot|.

We first isolate the step that must work on every sheet. For an analytic map with a specified lift to a uniformizing torus, pullback by multiplication by mm extracts mm-th roots of the lift’s character values. The period lattice controls how local roots glue.

Lemma 5.1 (Roots after division on a totally degenerate abelian variety). Let JJ be an abelian variety over Ω\Omega with analytic uniformization q:Tan→Janq:\mathbb{T}^{\mathrm{an}} \to J^{\mathrm{an}}, where T\mathbb{T} is a split torus and the kernel Λ⊂T(Ω)\Lambda\subset\mathbb{T}(\Omega) is a free lattice acting by translation. Suppose qq is a topological covering and a local analytic isomorphism. Let AA be an analytic space, let a:A→Tana:A \to\mathbb{T}^{\mathrm{an}} be analytic, and let m≥1m \ge1 be an integer. Set

B=A×q∘a,Jan,[m]Jan.B=A\times_{q\circ a,J^{\mathrm{an}},[m]}J^{\mathrm{an}}.

For every character χ\chi of T\mathbb{T}, the pullback of χ∘a\chi\circ a to BB has an analytic mm-th root.

Proof. Let π:B→A\pi:B \to A and b:B→Janb:B \to J^{\mathrm{an}} be the projections, and write aB=a∘πa_B=a\circ\pi. Locally on BB, lift bb to an analytic map tt to Tan\mathbb{T}^{\mathrm{an}}. The identity [m]b=q∘aB[m]b=q\circ a_B says that tm/aBt^m/a_B takes values in Λ\Lambda. This is a locally constant function because Λ\Lambda is discrete. Its image in Λ/Λm\Lambda/\Lambda^m is independent of the local lift: replacing tt by tρt\rho changes tm/aBt^m/a_B by ρm\rho^m, for ρ∈Λ\rho\in\Lambda. We therefore obtain an analytic-domain decomposition into finitely many clopen subsets BλˉB_{\bar\lambda}, indexed by Λ/Λm\Lambda/\Lambda^m.

Choose a representative λ∈Λ\lambda\in\Lambda for each coset. On BλˉB_{\bar\lambda}, adjust each local lift by a locally constant period so that tm=λaBt^m=\lambda a_B. Two adjusted lifts on an overlap differ by a period ρ\rho with ρm=1\rho^m=1. The lattice Λ\Lambda is torsion free, so ρ=1\rho=1. Thus the adjusted lifts glue to a single analytic map tλˉ:Bλˉ→Tant_{\bar\lambda}:B_{\bar\lambda} \to\mathbb{T}^{\mathrm{an}}. Choose dλ∈Ω×d_{\lambda}\in\Omega^{\times} with dλm=χ(λ)d_{\lambda}^m=\chi(\lambda). The function χ(tλˉ)/dλ\chi(t_{\bar\lambda})/d_{\lambda} is an mm-th root of χ(aB)\chi(a_B). These functions glue across the disjoint clopen subsets. No assumption that mm is prime to pp is used.

To obtain a root of an annular coordinate from the preceding lemma, we arrange that the pullback of a torus character has slope one. It then differs from the coordinate by a constant and an analytic unit close to one. Division on the Jacobian supplies a root of the character on the whole inverse annulus; a sufficiently long annulus leaves room for a middle collar on which the remaining unit also admits a root.

Proposition 5.2 (An auxiliary branched cover). Let m≥1m \ge1 be an integer, choose an integer

D>vp(m)+1p−1,D > v_p(m) + \frac{1}{p-1},

and let α∈k‾×\alpha\in\overline{k}^{\times} satisfy a=v(α)>2Da = v(\alpha) > 2D. There is a smooth, proper, connected curve Q/k‾Q/\overline{k} and a finite map Q→Pk‾1Q \to\mathbb{P}^{1}_{\overline{k}} whose ramification indices divide ℓ\ell, such that the pullback of the coordinate TT has an analytic mm-th root on the entire inverse image, over Ω\Omega, of

A∘={D<v(T)<a−D}.A^{\circ} = \{D < v(T) < a-D\}.

Proof. Let ZZ be the smooth projective normalization of

yℓ=T(T−1)T−α.y^{\ell} = \frac{T(T-1)}{T-\alpha}.

The four distinct points 0,α,1,∞0,\alpha,1,\infty have orders 1,−1,1,−11,-1,1,-1 in the divisor of the displayed rational function. The cyclic cover is connected, has degree ℓ\ell, and is totally ramified at precisely these points. Riemann–Hurwitz gives

2g(Z)−2=−2ℓ+4(ℓ−1),g(Z)=ℓ−1.2g(Z)-2=-2\ell+4(\ell-1), \qquad g(Z)=\ell-1.

Here is a split semistable model of ZZ. Work first over the integers of a sufficiently large finite extension of kk containing α\alpha. The blowup of the model of the line along (T,α)(T,\alpha) has two rational components and one node, with local equation TU=αTU=\alpha. The inner component has coordinate S=T/αS=T/\alpha; the four branch sections are disjoint and smooth. The Kummer function reduces to T−1T-1 on the outer component and to −S/(S−1)-S/(S-1) on the inner component. Each of these degree-ℓ\ell covers of a rational component has just two totally ramified points and is rational. Near the node the Kummer function is

T−11−U,\frac{T-1}{1-U},

a unit whose reduction at the node is −1-1. Since ℓ\ell is invertible, normalization is étale there, with ℓ\ell split copies of the node after enlarging the finite constant field. Away from the node and branch sections the same unit criterion gives étaleness. At a branch section the function or its inverse is a simple relative parameter, so the normalization is smooth there. Thus there are exactly two rational components, joined by ℓ\ell nodes, each of thickness v(α)=av(\alpha)=a. They are stable because ℓ≥3\ell\ge3.

Consequently the skeleton Γ\Gamma consists of two vertices and ℓ\ell parallel edges e1,…,eℓe_1,\ldots,e_\ell, each of length aa. The node annuli AjA_j are disjoint copies of {∣α∣<∣T∣<1}\{|\alpha|<|T|<1\} with coordinate TT. This last assertion can also be checked directly: on that annulus both TT and α/T\alpha/T have absolute value less than one, and the two factors in −(1−T)/(1−α/T)-(1-T)/(1-\alpha/T) have convergent ℓ\ell-th roots. Hence the inverse image is exactly ℓ\ell copies of the annulus.

We next apply nonarchimedean uniformization of abelian varieties, developed by Raynaud and Bosch–Lütkebohmert [23, 6, 5]. Let JJ be the Jacobian of ZZ, and choose an Abel–Jacobi map i:Z→Ji:Z\to J using a point over k‾\overline{k}. Total degeneration gives a toric analytic uniformization

q:Tan⟶JΩan,ker⁡(q)=Λ⊂T(Ω).q:\mathbb{T}^{\mathrm{an}}\longrightarrow J^{\mathrm{an}}_{\Omega}, \qquad\ker(q)=\Lambda\subset\mathbb{T}(\Omega).

Its character lattice is M=H1(Γ,Z)M=H_1(\Gamma,\mathbb{Z}); its tropical target is Hom⁡(M,R)\operatorname{Hom}(M,\mathbb{R}). The period lattice is free, and qq is a topological covering and local analytic isomorphism. These are the Jacobian uniformization conventions of [3], Sections 4.1–4.2 and 4.5.2. The compatibility of the classical and tropical Abel–Jacobi maps [3], Proposition 6.1 and Corollary 6.6 identifies the latter with integration on Γ\Gamma using the edge-length pairing.

The open Berkovich annulus AjA_j strongly deformation retracts to its open skeletal interval [2], Section 2.3, and is locally path connected. It is thus simply connected, so i∣Aji|_{A_j} lifts to an analytic map ı^j:Aj→Tan\hat{\imath}_j:A_j\to \mathbb{T}^{\mathrm{an}}: first lift continuously along the covering qq, then use its analytic local inverses. Orient all edges by increasing v(T)v(T). For j′≠jj'\ne j, the cycle cj=ej−ej′c_j=e_j-e_{j'} is an element of MM. If χj\chi_j is its character, then uj=χj∘ı^ju_j=\chi_j\circ\hat{\imath}_j is an analytic unit. The derivative of v(uj)v(u_j) along eje_j, with respect to v(T)v(T), is exactly one: the integral of the harmonic form represented by cjc_j over a segment of eje_j is the length of that segment. Lifts of the tropical map differ only by a constant period, which does not change this derivative.

By Lemma 4.1, an analytic unit on an open annulus has a unique dominant Laurent monomial at every interior Gauss radius; its exponent is constant and is the slope of its log-norm. It follows that

uj=cj′T(1+hj),hj=∑n≠0hj,nTn,cj′∈Ω×.u_j=c'_jT(1+h_j),\qquad h_j=\sum_{n\ne0}h_{j,n}T^n,\qquad c'_j\in\Omega^\times.

Here cj′Tc'_jT is the actual coefficient term of degree one. Dominance at every radius r∈(∣α∣,1)r\in(\lvert\alpha\rvert,1) gives ∣hj,n∣rn<1\lvert h_{j,n}\rvert r^n<1. Taking limits at the endpoints yields, for n>0n>0,

∣hj,n∣≤1,∣hj,−n∣≤∣α∣n.\lvert h_{j,n}\rvert\le1,\qquad\lvert h_{j,-n}\rvert\le\lvert\alpha\rvert^n.

Writing s=v(T)s=v(T), the ultrametric inequality consequently gives

∣hj∣≤p−min⁡{s,a−s}<p−Don Aj∩{D<s<a−D}.\lvert h_j\rvert\le p^{-\min\{s,a-s\}}<p^{-D}\qquad\text{on }A_j\cap\{D<s<a-D\}.

Figure 1 displays the cycle that supplies the slope-one character and the trimming that gives this bound.

On every closed subannulus this estimate has a strict uniform margin. The logarithm and exponential therefore converge locally uniformly and

bj=exp⁡(m−1log⁡(1+hj))b_j=\exp(m^{-1}\log(1+h_j))

is analytic there, with bjm=1+hjb_j^m=1+h_j. Indeed, since D>1/(p−1)D>1/(p-1), the logarithm satisfies ∣log⁡(1+hj)∣≤∣hj∣\lvert\log(1+h_j)\rvert\le\lvert h_j\rvert, and the valuation of the exponent is greater than D−vp(m)>1/(p−1)D-v_p(m)>1/(p-1).

For all ℓ\ell annuli at once, form the same algebraic finite étale cover

Z′=Z×i,J,[m]J⟶Z.Z'=Z\times_{i,J,[m]}J\longrightarrow Z.

It is finite étale for every positive integer mm, since the ground field has characteristic zero. Lemma 5.1 applied to a=ı^ja=\hat{\imath}_j gives an analytic mm-th root of uju_j on the whole inverse image of AjA_j.

Dividing it by bjb_j and an mm-th root of cj′c'_j gives an analytic mm-th root of TT above the middle collar. The annuli AjA_j are disjoint, so this gives a root on its entire inverse image in Z′Z'. Choose any connected component QQ of Z′Z'. Its map to ZZ is finite étale and surjective, so it is a smooth proper connected curve with the required property. Its composite with Z→P1Z \to\mathbb{P}^1 has only ramification indices 11 or ℓ\ell. All algebraic maps used to define QQ are over kˉ\bar{k}; the analytic roots and torus lifts are only required over Ω\Omega.

From collar roots to the completed universal tower

The first stage, Proposition 4.4, places the field LℓL_\ell inside HH. We now use the auxiliary covers from Proposition 5.2 to realize every finite Galois extension of K0K_0 in that same completion. The covers used to create the collars may be branched; their horizontal ramification indices divide ℓ\ell, so the first stage is sufficient to accommodate them.

Proof of Theorem 1.2. Let F/K0F/K_0 be any finite Galois extension inside KsepK^{\mathrm{sep}}, and let C→XkˉC \to X_{\bar{k}} be its smooth proper curve cover. Its branch set BB is finite. If BB is empty, then F⊂K~⊂HF \subset\widetilde{K} \subset H by definition of the universal étale field. Suppose henceforth that BB is nonempty.

For each b∈Bb \in B, choose a rational local parameter tbt_b at bb and disjoint disks BbB_b as in Lemma 3.1; these disks avoid ξ\xi. Let mbm_b be the least common multiple of the ramification indices of CC at the points above bb. Choose βb∈kˉ×\beta_b \in\bar{k}^{\times} whose absolute value is smaller than the radius of BbB_b in the coordinate tbt_b.

Apply Proposition 5.2 with m=mbm=m_b. In its notation choose a positive integer Db>vp(mb)+1/(p−1)D_b > v_p(m_b) + 1/(p-1) and αb∈kˉ×\alpha_b \in\bar{k}^{\times} with ab=−log⁡p∣αb∣>2Dba_b=-\log_p|\alpha_b|>2D_b. It gives a smooth proper connected curve cover of the TT-line whose horizontal indices divide ℓ\ell and on whose full inverse image of

Db<−log⁡p∣T∣<ab−DbD_b < -\log_p |T| < a_b-D_b

the coordinate TT has an analytic mbm_b-th root. Pull this cover back by the nonconstant morphism Xkˉ→Pkˉ1X_{\bar{k}} \to\mathbb{P}^1_{\bar{k}} given by T=tb/βbT=t_b/\beta_b, and take a normalized connected component. Denote the resulting cover by Wb→XkˉW_b \to X_{\bar{k}}.

Lemma 3.4 shows that its horizontal indices still divide ℓ\ell. Roots pull back to every point of the normalized component, so tbt_b has an analytic mbm_b-th root on the entire inverse image in Wb,ΩanW^{\mathrm{an}}_{b,\Omega} of the collar

∣βb∣p−(ab−Db)<∣tb∣<∣βb∣p−Db|\beta_b|p^{-(a_b-D_b)} < |t_b| < |\beta_b|p^{-D_b}

inside BbB_b. Here we multiplied the pulled-back root of TT by an mbm_b-th root of βb\beta_b in Ω\Omega.

Embed the finitely many function fields kˉ(Wb)\bar{k}(W_b) into the fixed KsepK^{\mathrm{sep}} over K0K_0 and form their compositum. Let W→XkˉW \to X_{\bar{k}} be the associated smooth proper connected curve. It dominates every WbW_b; its horizontal indices divide ℓ\ell by Lemma 3.4. All collar roots therefore pull back to the full inverse images in WΩanW^{\mathrm{an}}_\Omega.

Proposition 3.2, with Y=WΩY=W_\Omega, produces a connected finite étale cover VΩ→WΩV_\Omega\to W_\Omega and an analytic map from its exterior region to CΩanC^{\mathrm{an}}_\Omega over XΩanX^{\mathrm{an}}_\Omega. Since WW is proper over the algebraically closed field kˉ\bar{k}, invariance of finite étale covers under the algebraically closed extension kˉ⊂Ω\bar{k} \subset\Omega [10] gives a connected finite étale cover V→WV \to W and an isomorphism

V×kˉΩ≃VΩV \times_{\bar{k}} \Omega\simeq V_\Omega

over WΩW_\Omega.

No descent assertion is needed for the exterior analytic map.

Although no such bound was imposed on C→XkˉC \to X_{\bar{k}}, the composite V→XkˉV \to X_{\bar{k}} still has horizontal indices dividing ℓ\ell. By Lemma 3.4, its function field embeds over K0K_0 in LℓL_\ell. This embedding can, if desired, be chosen to extend the already chosen embedding of kˉ(W)\bar{k}(W): extend that embedding to kˉ(V)\bar{k}(V) inside KsepK^{\mathrm{sep}} and then take its Galois closure, which still has indices dividing ℓ\ell. Proposition 4.4 places its image in HH.

Lemma 3.3 now gives a K0K_0-embedding F=k‾(C)↪H⊂EF=\overline{k}(C)\hookrightarrow H\subset E. Because F/K0F/K_0 is normal, this embedding has image the chosen subfield FF of EE. Indeed a primitive element of FF is sent to a root of its minimal polynomial, all of whose roots already belong to FF; the same holds for every element. The image, having degree [F:K0][F:K_0], is therefore FF itself. Consequently F⊂HF\subset H.

Every element of KsepK^{\mathrm{sep}} lies in a finite Galois extension of K0K_0 inside KsepK^{\mathrm{sep}}: this is an algebraic closure of the characteristic-zero field K0K_0. Taking the union over those extensions proves Ksep⊂HK^{\mathrm{sep}}\subset H inside EE.

From completion to point sections

We now use Theorem 1.2 to compare decomposition groups. The comparison lifts any section localized at a rank-one valuation to a birational section. Koenigsmann’s theorem then places the same section at a rational point, and the two valuations give a contradiction. The remaining valuations are horizontal refinements, where the localization theorem of Pop–Stix already identifies the section.

Corollary 7.1. Let k/Qpk/\mathbb{Q}_p be finite and let X/kX/k be a smooth proper geometrically connected curve of genus at least two. Put K=k(X)K=k(X) and let K~⊂Ksep\widetilde{K}\subset K^{\mathrm{sep}} be the extension corresponding to the full arithmetic étale fundamental group of XX. Let ww be a real-valued rank-one valuation of KK extending the pp-adic valuation, normalized by w(p)=1w(p)=1, and fix a prolongation wsepw^{\mathrm{sep}} to KsepK^{\mathrm{sep}}. Restriction induces an isomorphism of profinite groups

Dwsep(Ksep/K)→∼Dw~(K~/K),w~=wsep∣K~.D_{w^{\mathrm{sep}}}(K^{\mathrm{sep}}/K)\xrightarrow{\sim}D_{\widetilde{w}}(\widetilde{K}/K),\qquad\widetilde{w}=w^{\mathrm{sep}}|_{\widetilde{K}}.

Proof. Restriction is surjective. Indeed, lift an automorphism stabilizing w~\widetilde{w} to an automorphism of Ksep/KK^{\mathrm{sep}}/K. Its translate of wsepw^{\mathrm{sep}} is another prolongation of w~\widetilde{w}. Transitivity of Gal⁡(Ksep/K~)\operatorname{Gal}(K^{\mathrm{sep}}/\widetilde{K}) on these prolongations allows us to adjust the lift so that it stabilizes wsepw^{\mathrm{sep}}. The transitivity statement for an infinite normal extension follows from its finite-level version by compactness.

An automorphism stabilizing a rank-one valuation preserves its normalized absolute value. To see why normalization matters, any two order-preserving embeddings of the value group into R\mathbb{R} differ by a positive scale; the value on pp fixes that scale. Such an automorphism therefore extends as an isometry to E=Ksep^E=\widehat{K^{\mathrm{sep}}}. If its restriction to K~\widetilde{K} is the identity, it fixes H=K~^⊂EH=\widehat{\widetilde{K}}\subset E pointwise. Theorem 1.2 gives Ksep⊂HK^{\mathrm{sep}}\subset H, so the automorphism is the identity.

Both decomposition groups are closed subgroups of their Galois groups. For example, stabilization of a valuation ring is the intersection, over its algebraic elements, of conditions on finite Galois orbits, each of which is clopen. Restriction is consequently a continuous bijection between compact Hausdorff groups, and its inverse is continuous.

Lemma 7.2. With kk, XX, KK, and K~\widetilde{K} as in Corollary 7.1, let ww be any rank-one valuation of KK extending the pp-adic valuation and let w~\widetilde{w} be any prolongation to K~\widetilde{K}. The projection

Dw~(K~/K)⟶GkD_{\widetilde{w}}(\widetilde{K}/K)\longrightarrow G_k

has no continuous group-homomorphic section.

Proof. Normalize w(p)=1w(p)=1 and prolong the valuation to KsepK^{\mathrm{sep}}. Suppose that a section exists. Corollary 7.1 lifts it to a continuous section

s^:Gk⟶GK=Gal⁡(Ksep/K)\widehat{s}:G_k\longrightarrow G_K=\operatorname{Gal}(K^{\mathrm{sep}}/K)

whose image stabilizes this prolongation. The birational section theorem of Koenigsmann places the entire image of s^\widehat{s} also in the decomposition group of a prolongation of ord⁡a\operatorname{ord}_a for some a∈X(k)a\in X(k) [13], preprint version, Proposition 2.4(2), Definition 1.4, and Observation 1.3. Here ord⁡a\operatorname{ord}_a is the closed-point valuation of KK, trivial on kk. Write Σ=s^(Gk)\Sigma=\widehat{s}(G_k) and M=(Ksep)ΣM = (K^{\mathrm{sep}})^\Sigma. The subgroup Σ\Sigma is compact and hence closed, so GM=ΣG_M = \Sigma. Let v1v_1 and v2v_2 be the restrictions to MM of the two selected prolongations. Each is henselian: its absolute Galois group stabilizes a prolongation, and the transitive action on all prolongations therefore has a single point. Both valuations are nontrivial and have rank one. Indeed, value groups in an algebraic valued extension differ by a torsion quotient, which preserves their convex-subgroup chains. Moreover,

v1(p)=1,v2∣k×=0,v_1(p)=1,\qquad v_2|_{k^\times}=0,

so these valuations are inequivalent. The resulting contradiction is an instance of the classical uniqueness principle for henselian valuations associated with F. K. Schmidt [24]. We give explicitly the two-valuation approximation needed here.

Choose a nonsquare d∈k×d \in k^\times. It remains a nonsquare in MM: otherwise GM=ΣG_M=\Sigma would fix a square root of dd in kˉ\bar{k}, contrary to the surjection Σ→Gk\Sigma\to G_k. Let t∈Kt \in K be a uniformizer at aa. Choose a positive integer NN with N>v1(t)N>v_1(t), and put z=pN/tz=p^N/t. Then

v1(z)>0,v2(z)=−v2(t)<0.v_1(z)>0,\qquad v_2(z)=-v_2(t)<0.

For positive integers rr, set

qr=d+(1−d)zr1+zr.q_r=d+(1-d)\frac{z^r}{1+z^r}.

The denominator is nonzero because v1(zr)>0v_1(z^r)>0. At the first valuation qrq_r tends to dd, and at the second it tends to 11. More precisely,

v1(qr/d−1)=v1((1−d)/d)+rv1(z)⟶+∞,v_1(q_r/d-1)=v_1((1-d)/d)+rv_1(z)\longrightarrow+\infty,

whereas

v2(qr−1)=v2(d−1)−rv2(z)⟶+∞.v_2(q_r-1)=v_2(d-1)-rv_2(z)\longrightarrow+\infty.

For sufficiently large rr, Hensel’s lemma applied at 11 gives a square root of qr/dq_r/d using v1v_1, and a square root of qrq_r using v2v_2. The sufficient inequalities are

v1(qr/d−1)>2v1(2),v2(qr−1)>2v2(2).v_1(q_r/d-1)>2v_1(2),\qquad v_2(q_r-1)>2v_2(2).

Their quotient makes dd a square in MM, a contradiction. This argument includes p=2p=2 and arbitrary nondiscrete rank-one value groups.

Proof of the surjectivity in Theorem 1.1. Let s:Gk→ΠXs:G_k\to\Pi_X be a continuous group-homomorphic section. By the valuative localization theorem of Pop–Stix [22], preprint version, Theorem 26, there is a valuation ww of KK extending the pp-adic valuation, together with a prolongation w~\widetilde{w} to K~\widetilde{K}, such that

s(Gk)⊆Dw~(K~/K).s(G_k)\subseteq D_{\widetilde{w}}(\widetilde{K}/K).

We classify the possibilities for ww.

The valuation transcendence inequality gives

ratrank⁡(Γw/Γk)≤1.\operatorname{ratrank}(\Gamma_w/\Gamma_k)\leq1.

Since Γk\Gamma_k has rational rank one and ordered rank is at most rational rank, ww has rank at most two. Rank one is excluded by Lemma 7.2. If ww has rank two, let uu be its rank-one coarsening. The nonzero proper convex subgroup of the rank-two value group has a unique extension to the value group of w~\widetilde{w}: it consists of those values some positive multiple of which lies in the original subgroup. This follows because an algebraic valued extension has torsion quotient of value groups. Every automorphism stabilizing w~\widetilde{w} preserves this subgroup, so s(Gk)s(G_k) stabilizes the resulting prolongation of uu. If uu is nontrivial on kk, it can be normalized to extend the pp-adic valuation and is excluded by the same lemma.

It remains that uu is trivial on kk. Properness gives it a center on XX. This center cannot be generic because uu is nontrivial, so it is a closed point aa. The valuation ring of uu dominates the discrete valuation ring OX,a\mathcal{O}_{X,a}. Every element of K×K^\times is a power of a uniformizer times a unit of this local ring; hence uu is equivalent to ord⁡a\operatorname{ord}_a. Its residue field is the finite extension k(a)/kk(a)/k. The second valuation in the composition ww is the unique prolongation of the pp-adic valuation to k(a)k(a). Thus ww is of type 2h2h.

The chosen horizontal prolongation to KkˉK\bar{k} selects a geometric point above aa. Its stabilizer acts on constants through Gk(a)⊆GkG_{k(a)} \subseteq G_k, with the embedding determined by that point. Since s(Gk)s(G_k) projects onto GkG_k, necessarily k(a)=kk(a)=k. For the full étale tower the horizontal inertia is trivial, and the horizontal decomposition group is the point section at aa. Equivalently, the type-2h2h conclusion of [22], preprint version, Section 6.1, gives s=sas=s_a after the corresponding geometric choice of base point. This proves surjectivity.

Notice that a type-2h2h valuation can lie over a non-rational closed point. Rationality here follows from the surjection to GkG_k; no unrestricted equivalence between “not type 2h2h” and “not originating from a rational point” is used.

Proof of the injectivity in Theorem 1.1. Suppose a,b∈X(k)a,b\in X(k) have the same section class modulo the full geometric fundamental group. Let JJ be the Jacobian of XX, and use aa as base point for the Abel–Jacobi map

ia:X⟶J,x⟼[x−a].i_a:X\longrightarrow J,\qquad x\longmapsto[x-a].

Their induced sections on JJ are again geometrically conjugate. For every positive integer nn, multiplication [n]:J→J[n]:J\to J is a finite étale torsor under J[n]J[n], also when p∣np\mid n since kk has characteristic zero. Conjugate sections give isomorphic GkG_k-actions on its geometric fiber, so the existence of a fixed point is unchanged by conjugation. The section at ia(a)=0i_a(a)=0 has the fixed lift 00. Therefore the fiber over ia(b)i_a(b) has a GkG_k-fixed point, which is a kk-rational point of that fiber. Thus

ia(b)∈⋂n≥1nJ(k).i_a(b)\in\bigcap_{n\geq1}nJ(k).

This is also the usual Kummer argument: the class of the multiplication torsor pulled back along a section is unchanged by geometric conjugation.

The compact pp-adic analytic group J(k)J(k) has an open subgroup U≃Zpg(X)[k:Qp]U\simeq\mathbb{Z}_p^{g(X)[k:\mathbb{Q}_p]} by the local logarithm. If q=[J(k):U]q=[J(k):U], then qJ(k)⊆UqJ(k)\subseteq U. For each r≥0r\geq0, take n=qprn=qp^r in the displayed intersection. Any element of that intersection can then be written as pr(qy)p^r(qy) with y∈J(k)y\in J(k), and hence belongs to prUp^rU. As UU is a finite free Zp\mathbb{Z}_p-module,

⋂r≥0prU=0,\bigcap_{r\geq0}p^rU=0,

and hence ia(b)=0i_a(b)=0. If a≠ba\ne b, the linear equivalence of the degree-one divisors aa and bb would give a degree-one morphism X→P1X\to\mathbb{P}^1, forcing genus zero. Hence a=ba=b, as required.

Global consequences via finite descent

We now combine the local theorem with established descent and finite-support results. First we identify the adelic tuples obtained from global sections. We then give a criterion that makes these sections geometric and apply it to abelian targets and modular curves. Finally, we describe further restrictions on the localization tuples without assuming that criterion. In this section KK is a number field. For a field FF of characteristic zero and a smooth proper geometrically connected curve D/FD/F, write Sec⁡(D/F)\operatorname{Sec}(D/F) for the conjugacy classes of continuous splittings of

1⟶π1(DF‾)⟶π1(D)⟶GF⟶1,GF=Gal⁡(F‾/F),1\longrightarrow\pi_1(D_{\overline{F}})\longrightarrow\pi_1(D)\longrightarrow G_F\longrightarrow1,\qquad G_F=\operatorname{Gal}(\overline{F}/F),

where conjugation is by the full geometric étale fundamental group. For a smooth proper geometrically connected curve C/KC/K of genus at least two, the global section conjecture asserts that the point map

κC:C(K)⟶Sec⁡(C/K)\kappa_C : C(K) \longrightarrow\operatorname{Sec}(C/K)

is bijective. Thus the section classes considered here are the full arithmetic étale classes, without a birational lifting assumption.

Localization and descent

For a proper KK-scheme VV, use the modified adelic set

V(AK)∙=∏v∤∞V(Kv)×∏v realπ0(V(Kv)).V(\mathbb{A}_K)_\bullet= \prod_{v\nmid\infty} V(K_v) \times\prod_{v\ \mathrm{real}} \pi_0(V(K_v)).

where π0\pi_0 denotes the set of connected components. Complex places are omitted, as in the corrected convention of Stoll’s erratum [27], Section 1. For geometrically connected CC the complex component factors in the earlier convention are singletons, but their omission matters for geometrically disconnected torsors. A point of C(AK)∙C(\mathbb{A}_K)_\bullet survives a torsor Y→CY \to C under a finite KK-group scheme GG if it lifts to Yξ(AK)∙Y^\xi(\mathbb{A}_K)_\bullet for some twist YξY^\xi, where ξ∈H1(K,G)\xi\in H^1(K,G). Write C(AK)∙f-covC(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}cov}} for the points surviving every such torsor, and C(AK)∙f-abC(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}ab}} for those surviving every such torsor with GG abelian [26], Definitions 5.2 and 5.4. We identify C(K)C(K) with its diagonal image. The definitions give

C(K)⊆C(AK)∙f-cov⊆C(AK)∙f-ab.(2)C(K) \subseteq C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}cov}} \subseteq C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}ab}}. \tag*{(2)}

The local section theorems define a localization map for every such curve, without any descent equality hypothesis. Let s∈Sec⁡(C/K)s \in\operatorname{Sec}(C/K). At every finite place vv, its localization is the section of a unique point Pv∈C(Kv)P_v \in C(K_v) by Theorem 1.1. At a real place the section map

π0(C(Kv))⟶Sec⁡(C/Kv)\pi_0(C(K_v)) \longrightarrow\operatorname{Sec}(C/K_v)

is bijective by the real section theorem; the genus-at-least-two case is included in the theorem of Bresciani–Vistoli [8], Introduction, Theorem. At a complex place the local Galois group is trivial and the local condition is automatic. Thus ss is a Selmer section: it comes from a local point at every place. Properness makes these localizations a modified adelic point x(s)∈C(AK)∙x(s) \in C(\mathbb{A}_K)_\bullet. We obtain the canonical map

loc⁡:Sec⁡(C/K)⟶C(AK)∙,s⟼x(s).\operatorname{loc} : \operatorname{Sec}(C/K) \longrightarrow C(\mathbb{A}_K)_\bullet,\qquad s \longmapsto x(s).

Proposition 8.1 (The localization image). Let C/KC/K be a smooth proper geometrically connected curve of genus at least two over a number field. Then

loc⁡(Sec⁡(C/K))=C(AK)∙f-cov.(3)\operatorname{loc}(\operatorname{Sec}(C/K)) = C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}cov}}. \tag*{(3)}

Proof. Let s∈Sec⁡(C/K)s \in\operatorname{Sec}(C/K). Choose a representative point of the determined component at each real place, and any point at each complex place. Harari–Stix’s Theorem 2.1, with U=1U = 1 and the omitted set of places S=∅S = \varnothing, shows that these localizations survive every torsor under a finite KK-group scheme; their Remark 2.2(1) identifies this as finite-cover descent [11], preprint version, Theorem 2.1 and Remark 2.2(1). Consequently

x(s)∈C(AK)∙f-cov.x(s) \in C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}cov}}.

Conversely, let xx belong to the finite-cover locus and, by properness, choose an ordinary adelic representative P=(Pv)∈C(AK)P=(P_v)\in C(\mathbb{A}_K). Stoll’s cohomological description after Definition 5.2, together with Definition 5.4, identifies survival with diagonal global H1(K,G)H^1(K,G)-membership of the local evaluations for every torsor under a finite KK-group scheme GG [26]. These evaluations factor through the real components, and complex places impose no lifting condition, in the corrected convention already used above. Thus every such PP satisfies the ordinary finite-torsor condition. Harari–Stix’s Theorem 2.1 with U=1U=1 and S=∅S=\varnothing, using implication (i)⇒(iii), gives a global section whose restrictions are the point sections of PvP_v, up to conjugation by the full geometric fundamental group [11], preprint version, Theorem 2.1 and Remark 2.2(1). The finite-point and real-component uniqueness used above identify its localization with xx, proving (3).

Criteria for point sections

Proposition 8.1 determines which adelic tuples come from sections. To recover rational points from the sections themselves, we also need the finite-support theorem of Stix.

Theorem 8.2 (Finite-cover descent criterion). Let C/KC/K be a smooth proper geometrically connected curve of genus at least two over a number field. If

C(AK)∙f-cov=C(K),(4)C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}cov}} = C(K), \tag*{(4)}

then κC:C(K)→Sec⁡(C/K)\kappa_C:C(K)\to\operatorname{Sec}(C/K) is bijective.

Proof. Let s∈Sec⁡(C/K)s\in\operatorname{Sec}(C/K). By Proposition 8.1 and (4), there is a∈C(K)a\in C(K) such that x(s)x(s) is the diagonal localization of aa, hence equals the localization of the point section sas_a.

At every finite place the point determined by ss is therefore aa. At each real place choose aa itself in the determined component, and choose aa also at each complex place. The resulting ordinary adelic representative of the localization tuple is supported on {a}\{a\}. Thus the support Z(s)Z(s) in Stix’s terminology, the Zariski-closed support of this tuple, is finite over KK. Since CC is proper of genus at least two, it is hyperbolic and has no cuspidal sections. Stix’s finite-support result therefore makes ss a point section [25], arXiv version, Section 3 and Corollary 5. Finally, if two points give the same global section, their sections agree after restriction to any one finite completion. Injectivity in Theorem 1.1 then identifies the two points. This proves injectivity and also identifies the point giving ss with aa.

The equality in (4) is supplied by several published criteria. For an abelian variety A/KA/K, let

Sha⁡(A/K)=ker⁡(H1(K,A)⟶∏vH1(Kv,A)),\operatorname{Sha}(A/K)=\ker\left(H^1(K,A)\longrightarrow\prod_v H^1(K_v,A)\right),

and write Sha⁡(A/K)div\operatorname{Sha}(A/K)_{\mathrm{div}} for its maximal divisible subgroup.

Corollary 8.3 (Abelian targets). Let C/KC/K be a smooth proper geometrically connected curve of genus at least two over a number field. The map κC\kappa_C is bijective if either of the following conditions holds:

  1. there is a nonconstant KK-morphism C→AC\to A to an abelian variety with A(K)A(K) finite and Sha⁡(A/K)div=0\operatorname{Sha}(A/K)_{\mathrm{div}}=0;

  1. the Jacobian JJ of CC has J(K)J(K) finite and Sha⁡(J/K)div=0\operatorname{Sha}(J/K)_{\mathrm{div}}=0.

Proof. Stoll’s Theorem 8.6 gives C(AK)∙f-ab=C(K)C(\mathbb{A}_K)_{\bullet}^{\mathrm{f\text{-}ab}}=C(K) under the first condition, and his Corollary 8.1 gives the same equality under the second, without requiring a KK-rational base point on CC [26]. The corrected covering arguments and the repair of Proposition 8.5 used by Theorem 8.6 are in [27], Sections 2–3; the stated equalities are retained. The inclusions (2) now imply (4), so Theorem 8.2 applies. ∎

For an elliptic curve E/QE/\mathbb{Q} and a prime pp, put

sp(E)=corank⁡ZpSel⁡p∞(E/Q),s_p(E)=\operatorname{corank}_{\mathbb{Z}_p}\operatorname{Sel}_{p^\infty}(E/\mathbb{Q}),

where the Selmer group uses the local Kummer conditions at every place.

Corollary 8.4 (A Selmer-corank-zero elliptic target). Let C/QC/\mathbb{Q} be a smooth proper geometrically connected curve of genus at least two. Suppose that there is a nonconstant Q\mathbb{Q}-morphism C→EC\to E to an elliptic curve and that sp(E)=0s_p(E)=0 for at least one prime pp. Then κC\kappa_C is bijective.

Proof. The corank-zero case of the companion Selmer converse [18], Theorem 1.1, gives rank⁡ZE(Q)=0\operatorname{rank}_{\mathbb{Z}} E(\mathbb{Q})=0 and finiteness of the entire group Sha⁡(E/Q)\operatorname{Sha}(E/\mathbb{Q}). The Mordell–Weil theorem therefore makes E(Q)E(\mathbb{Q}) finite, while finiteness of Sha⁡(E/Q)\operatorname{Sha}(E/\mathbb{Q}) makes its maximal divisible subgroup zero. Corollary 8.3(i) applies.

Corollary 8.5 (Modular curves). For every positive integer NN such that the smooth projective geometrically connected curve X0(N)/QX_0(N)/\mathbb{Q} has genus at least two, the map κX0(N)\kappa_{X_0(N)} is bijective. The same assertion holds for the smooth projective geometrically connected curve X1(N)/QX_1(N)/\mathbb{Q} whenever its genus is at least two.

Proof. Stoll’s Corollary 8.8 gives C(AQ)∙f-ab=C(Q)C(\mathbb{A}_{\mathbb{Q}})_{\bullet}^{\mathrm{f\text{-}ab}}=C(\mathbb{Q}) for each of these projective modular curves of positive genus [26], with the corrected conventions of [27]. Apply (8.1) and Theorem 8.2.

The general implication retains the finite-cover equality as a genuine arithmetic hypothesis; local point-theoreticity alone does not supply it. The elliptic-target specialization uses only Selmer corank zero over Q\mathbb{Q}, while the modular-curve cases come directly from Stoll’s descent result and do not depend on that specialization.

Further restrictions on localization

We return to an arbitrary smooth proper geometrically connected curve C/KC/K of genus at least two over a number field, without assuming (8.3). The localization image in Proposition 8.1 satisfies additional descent conditions and, under a restriction on KK, has finite image at each finite place.

Let ρ:C(AK)→C(AK)∙\rho:C(\mathbb{A}_K)\to C(\mathbb{A}_K)_\bullet be the natural projection. Write Dlin(C)D_{\mathrm{lin}}(C) for the ordinary adelic descent set: PP belongs to it if, for every CC-torsor T→CT\to C under a linear algebraic KK-group GG, it lifts to Tξ(AK)T^\xi(\mathbb{A}_K) for some global twist ξ∈H1(K,G)\xi\in H^1(K,G). Harari–Stix’s Theorem 4.1 sends the finite-torsor condition established in Proposition 8.1 to this linear descent condition, while their Corollary 3.1, applied to the geometric abelianization of the interpolating section, gives Brauer–Manin membership [11], preprint version, Corollary 3.1 and Theorem 4.1. Consequently

ρ−1(loc⁡(Sec⁡(C/K)))⊆Dlin(C)∩C(AK)Br.(5)\rho^{-1}\bigl(\operatorname{loc}(\operatorname{Sec}(C/K))\bigr)\subseteq D_{\mathrm{lin}}(C)\cap C(\mathbb{A}_K)^{\mathrm{Br}}. \tag*{(5)}

This applies to every ordinary adelic representative of a modified localization tuple; the lifting condition uses ordinary adeles also when TT is nonproper.

Finally, if KK contains no CM subfield, Betts–Stix’s Theorem A, applied to the sections shown to be Selmer in Section 8.1, gives

#loc⁡v(Sec⁡(C/K))<∞,loc⁡v=pr⁡v∘loc⁡,\#\operatorname{loc}_v(\operatorname{Sec}(C/K))<\infty,\qquad\operatorname{loc}_v=\operatorname{pr}_v\circ\operatorname{loc},

for every finite place vv, with image in C(Kv)C(K_v) [4], arXiv version, Definition 1.1 and Theorem A. This asserts neither injectivity of localization nor finiteness of the global section set, and supplies no effective or uniform bound.

References

  1. [1]Omid Amini, Matthew Baker, Erwan Brugallé, and Joseph Rabinoff. Lifting harmonic morphisms I: metrized complexes and Berkovich skeleta. Research in the Mathematical Sciences, 2, 2015. Article 7. Preprint version: arXiv:1303.4812v3, https://arxiv.org/abs/1303.4812v3.
  2. [2]Matthew Baker, Sam Payne, and Joseph Rabinoff. On the structure of nonarchimedean analytic curves. In Tropical and Non-Archimedean Geometry, volume 605 of Contemporary Mathematics, pages 93–121. American Mathematical Society, Providence, RI, 2013. Preprint version: arXiv:1404.0279v1, https://arxiv.org/abs/1404.0279v1.
  3. [3]Matthew Baker and Joseph Rabinoff. The skeleton of the Jacobian, the Jacobian of the skeleton, and lifting meromorphic functions from tropical to algebraic curves. International Mathematics Research Notices, 2015(16):7436–7472, 2015. Preprint version: arXiv:1308.3864v1, https://arxiv.org/abs/1308.3864v1.
  4. [4]L. Alexander Betts and Jakob Stix. Galois sections and p-adic period mappings. Annals of Mathematics, 201(1):79–166, 2025. Preprint version: arXiv:2204.13674v1, https://arxiv.org/abs/2204.13674v1.
  5. [5]Siegfried Bosch. Stable reduction and rigid analytic uniformization of abelian varieties. Groupe d’étude d’analyse ultramétrique, 11:1–4, 1983–1984. Exposé 20. https://www.numdam.org/item/GAU_1983-1984__11__A12_0/.
  6. [6]Siegfried Bosch and Werner Lütkebohmert. Stable reduction and uniformization of abelian varieties. II. Inventiones Mathematicae, 78(2):257–297, 1984. https://doi.org/10.1007/BF01388596.
  7. [7]Giulio Bresciani. The section conjecture for the toric fundamental group over p-adic fields. arXiv:2409.07923v5, 2026. Revision of September 14, 2026. https://arxiv.org/abs/2409.07923v5.
  8. [8]Giulio Bresciani and Angelo Vistoli. An elementary approach to Stix’s proof of the real section conjecture. arXiv:2012.06278v2, 2020. https://arxiv.org/abs/2012.06278v2.
  9. [9]Alexander Grothendieck. Letter to G. Faltings (translation into English). In Leila Schneps and Pierre Lochak, editors, Geometric Galois Actions, 1, volume 242 of London Mathematical Society Lecture Note Series, pages 285–293. Cambridge University Press, 1997. Letter dated June 27, 1983. https://doi.org/10.1017/CBO9780511758874.018.
  10. [10]Alexander Grothendieck, editor. Revêtements étales et groupe fondamental (SGA 1), volume 3 of Documents Mathématiques. Société Mathématique de France, Paris, 2003. Recomposed and annotated edition of the 1971 original. https://arxiv.org/abs/math/0206203v2.
  11. [11]David Harari and Jakob Stix. Descent obstruction and fundamental exact sequence. In Jakob Stix, editor, The Arithmetic of Fundamental Groups – PIA 2010, volume 2 of Contributions in Mathematical and Computational Sciences, pages 147–166. Springer, 2012. Author preprint dated February 11, 2011. https://www.imo.universite-paris-saclay.fr/~david.harari/articles/finidesc.pdf.arxiv.org/abs/1005.1302
  12. [12]David Harbater. Galois coverings of the arithmetic line. In Number Theory, New York 1984–1985, volume 1240 of Lecture Notes in Mathematics, pages 165–195. Springer, Berlin, 1987.DOI
  13. [13]Jochen Koenigsmann. On the ‘section conjecture’ in anabelian geometry. Journal für die reine und angewandte Mathematik, 588:221–235, 2005. Preprint version: arXiv:math/0305226v1, https://arxiv.org/abs/math/0305226v1.
  14. [14]Emmanuel Lepage. Resolution of nonsingularities for Mumford curves. Publications of the Research Institute for Mathematical Sciences, 49(4):861–891, 2013. https://doi.org/10.4171/PRIMS/122.
  15. [15]Qing Liu. Tout groupe fini est un groupe de Galois sur ℚₚ(T), d’après Harbater. In Recent Developments in the Inverse Galois Problem, volume 186 of Contemporary Mathematics, pages 261–265. American Mathematical Society, Providence, RI, 1995. Author version. https://www.math.u-bordeaux.fr/~qliu/articles/rigidharb.pdf.DOI
  16. [16]James S. Milne. Abelian varieties, 2022. Author version, January 2, 2022, of the chapter in Arithmetic Geometry (1986), pp. 103–150. Theorem 8.2 and Remark 8.4. https://www.jmilne.org/math/xnotes/AVs.pdf.
  17. [17]Shinichi Mochizuki and Shota Tsujimura. Resolution of nonsingularities, point-theoreticity, and metric-admissibility for p-adic hyperbolic curves. RIMS Preprint 1974, 2024. Revision of March 30, 2024. Authors’ revised preprint.DOI
  18. [18]OpenAI. The Selmer converse for elliptic curves at every prime. OpenAI Math Release preprint OAI:The-Selmer-converse-for-elliptic-curves-at-every-prime-September-24-2026, 2026.
  19. [19]Jesse Pajwani. The valuative section conjecture, étale homotopy, and Berkovich spaces. Homology, Homotopy and Applications, 27(2):261–288, 2025. Preprint version: arXiv:2304.03970v3, https://arxiv.org/abs/2304.03970v3.
  20. [20]Jérôme Poineau. Raccord sur les espaces de Berkovich. Algebra & Number Theory, 4(3):297–334, 2010. Appendix A, Theorem A.1. https://msp.org/ant/2010/4-3/ant-v4-n3-p03-s.pdf.DOI
  21. [21]Florian Pop. On the birational p-adic section conjecture. Compositio Mathematica, 146(3):621–637, 2010. https://doi.org/10.1112/S0010437X09004436.
  22. [22]Florian Pop and Jakob Stix. Arithmetic in the fundamental group of a p-adic curve: On the p-adic section conjecture for curves. Journal für die reine und angewandte Mathematik, 725:1–40, 2017. Preprint version: arXiv:1111.1354v1 (2011), https://arxiv.org/abs/1111.1354v1.
  23. [23]Michel Raynaud. Variétés abéliennes et géométrie rigide. In Actes du Congrès International des Mathématiciens (Nice, 1970), volume 1, pages 473–477. Gauthier-Villars, Paris, 1971.
  24. [24]Friedrich Karl Schmidt. Mehrfach perfekte körper. Mathematische Annalen, 108:1–25, 1933. https://doi.org/10.1007/BF01452819.
  25. [25]Jakob Stix. On the birational section conjecture with local conditions. Inventiones Mathematicae, 199(1):239–265, 2015. arXiv version: arXiv:1203.3236v3, https://arxiv.org/abs/1203.3236v3.
  26. [26]Michael Stoll. Finite descent obstructions and rational points on curves. Algebra & Number Theory, 1(4):349–391, 2007. Revised preprint: https://mathe2.uni-bayreuth.de/stoll/papers/covers-revised.pdf.arxiv.org/abs/math/0606465
  27. [27]Michael Stoll. Errata: Finite descent obstructions and rational points on curves, 2017. Online supplement, document dated March 29, 2017. https://msp.org/ant/2007/1-4/ant-v1-n4-x01-Errata-FiniteDescent.pdf.
  28. [28]Akio Tamagawa. Resolution of nonsingularities of families of curves. Publications of the Research Institute for Mathematical Sciences, 40(4):1291–1336, 2004. https://doi.org/10.2977/PRIMS/1145475448.
  29. [29]Michael Temkin. Introduction to Berkovich analytic spaces. Lecture notes, Jussieu summer school, 2010. Section 5.1, https://berkovich-2010.imj-prg.fr/lecture-notes/Temkin-lecture-notes.pdf.arxiv.org/abs/1010.2235

Paper details

vlnc:W_u7uv5ena
Published
License
LicenseRef-OpenAI-public
Usage

views · 21

downloads · 6

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents