Skip to content
Licensed Unlicensed Requires Authentication Published online by De Gruyter April 20, 2023

Geometric arcs and fundamental groups of rigid spaces

  • Piotr Achinger ORCID logo , Marcin Lara ORCID logo and Alex Youcis ORCID logo

Abstract

We develop the notion of a geometric covering of a rigid space 𝑋, which yields a larger class of covering spaces than that studied previously by de Jong. Geometric coverings are closed under disjoint unions and are étale local on 𝑋. If 𝑋 is connected, its geometric coverings form a tame infinite Galois category and hence are classified by a topological group. The definition is based on the property of lifting of “geometric arcs” and is meant to be an analogue of the notion developed for schemes by Bhatt and Scholze.

Award Identifier / Grant number: 802787

Award Identifier / Grant number: 444845124

Funding statement: This work is a part of the project KAPIBARA supported by the funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 802787). During the final revisions and editing of the document, the second named author was supported by the Priority Research Area SciMat budget under the program Excellence Initiative – Research University at the Jagiellonian University in Kraków, and later, in Frankfurt, he was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) TRR 326 Geometry and Arithmetic of Uniformized Structures, project number 444845124. In addition, the final named author was partially supported by a JSPS fellowship during the final editing of this document.

A Curve-connectedness of rigid varieties

Definition A.1

A rigid 𝐾-space is curve-connected if, for every two classical points x 0 , x 1 of 𝑋, there exists a sequence of morphisms C i X ( i = 0 , , n ) over 𝐾 where each C i is a connected affinoid rigid 𝐾-curve and we have x 0 im ( C 0 X ) , x 1 im ( C n X ) , and im ( C i X ) im ( C i 1 X ) for i = 1 , , n .

Proposition A.2

Let 𝑋 be a connected rigid 𝐾-space. Then 𝑋 is curve-connected.

By [13, Theorem 3.3.6], possibly after replacing 𝐾 with K 1 / p n if 𝐾 has characteristic p > 0 , one can assume that the curves C i in Definition A.1 are smooth over 𝐾.

Remark A.3

(a) The analogous result was proven by de Jong in [16, Theorem 6.1.1] for quasi-compact rigid 𝐾-spaces in the case when 𝐾 is discretely valued. Our proof follows his, but requires some nontrivial alterations due to the fact that O K is non-Noetherian. It is also worth pointing out that Berkovich obtained similar results for Berkovich spaces (see [7, Theorem 4.1.1]) which, in the language of adic spaces, are partially proper over Spa ( K ) .

(b) The referee has suggested to us a shorter, simpler proof using more advanced machinery. See Remark A.7.

Proof

Idea of the proof. If 𝑋 is a connected affine variety, one can use the Bertini theorem to find a connected hypersurface passing through two given points. In the situation at hand, if X = Spa ( A ) is affinoid, we can construct a suitable connected hyperplane section of its reduction X ~ = Spec ( A ~ ) and lift it to the formal model Spf ( A ) . The main difficulty, resolved carefully below, is to ensure that the resulting hypersurface remains connected on the generic fiber.

Step 1 [We may assume that 𝑋 is affinoid]. Let 𝑈 be the union of all connected quasi-compact opens containing x 0 . Then 𝑈 is (open and) closed: if y U and 𝑉 is a connected affinoid neighborhood of 𝑦, then V U = ; otherwise, V U is quasi-compact and connected. Since 𝑋 is connected, x 1 U , i.e. x 0 and x 1 admit a connected quasi-compact neighborhood 𝑊. Write W = i = 1 n X i with X i affinoid, which we may assume are connected and X i X i 1 (in particular, the intersection has classical points), and it is enough to show the result for the X i .

Step 2 [We may assume that 𝑋 is geometrically connected and geometrically normal]. In the proof, we can freely pass to a finite extension 𝐿 of 𝐾 because every connected component of X L surjects onto 𝑋. By [13, § 3.2], we may assume that 𝑋 is geometrically connected. By [13], affinoid 𝐾-algebras are excellent and we have the normalization map X n X , a finite surjective morphism. Therefore, by [13, Theorem 3.3.6], after replacing 𝐾 with K 1 / p n (if char K = p > 0 ), we may assume that 𝑋 is geometrically normal.

Note that, a priori, the replacement of 𝐾 with K 1 / p n is not permissible, as the curves we obtain may not be rigid 𝐾-curves if [ K 1 / p n : K ] is infinite. That said, let X = X K 1 / p . The absolute Frobenius F : X X factors through the relative Frobenius F X / K : X X and the base change map π : X X . Each of those maps is a (universal) homeomorphism, so in particular, X is connected. Suppose that C X is a map from a connected curve connecting the unique points x i X over x i X . Let C X be its base change along F X / K ; then C C is a homeomorphism, so 𝐶 is connected. This argument is inspired by [16, § 6.2].

Step 3 [Reduced Fiber Theorem]. Write X = Spa ( A ) . By the Reduced Fiber Theorem [26, 11], passing to a finite extension of 𝐾, we may assume that A is a topologically finitely presented O K -algebra such that A k is geometrically reduced and the formation of A is compatible with further finite extensions of 𝐾.

Write A ~ = A k and X ~ = Spec ( A ~ ) , and set d = dim X . We denote the irreducible components of X ~ by Z 1 , , Z r . Note that we have d = dim Z i for all 𝑖. To see this, let U ~ i X ~ be an affine open subset contained in Z i so that dim U ~ i = dim Z i and let U i Spf ( A ) be the corresponding open formal subscheme. Then the generic fiber U i , η is an affinoid subdomain of 𝑋 and hence has dimension 𝑑. We conclude by [14, Theorem A.2.1].

Step 4 [Noether normalization]. Since X ~ is geometrically reduced, Lemma A.4 below yields a finite and generically étale morphism f ~ : X ~ A k d (if 𝑘 is finite, we might need to pass to a finite extension of 𝐾). Choosing lifts of f ~ * ( x i ) to A , we lift f ~ to a map

f : Spec ( A ) Spec ( O K x 1 , , x d ) ,

which we claim is finite, and étale at the points of X ~ where f ~ is étale. Then the morphism f η : X D K d induced by K x 1 , , x d A is finite as well.

To prove this claim, set B = O x 1 , , x d and let f 0 : B / ϖ B A / ϖ A be the induced map. Since f 0 is a thickening of f ~ in the sense of [49, Tag 04EX] and A / ϖ A is of finite presentation over O K / ϖ O K , by [49, Tag 0BPG], the finiteness of f ~ implies the finiteness of f 0 . By [24, Chapter I, Proposition 4.2.3], this implies that 𝑓 is finite and therefore also finitely presented [9, Theorem 7.3/4].

Let now x X ~ Spec ( A ) be a point where f ~ is étale; we will show that 𝑓 is étale at 𝑥. Set y = f ( x ) . As 𝑓 is finitely presented and Spec ( A ) y Spec ( k ( y ) ) is étale at 𝑥 (as it agrees with f ~ ), it is enough by [49, Tag 01V9] to show that O Spec ( B ) , y O Spec ( A ) , x is flat. Note that

( O Spec ( A ) , x O Spf ( A ) , x ) = lim x D ( f ) ( A [ f 1 ] A f 1 )

and the right-hand side is flat by Gabber’s lemma [9, Lemma 8.2/2], but the map is also local by [9, Remark 7.2/1]. Thus, O Spec ( A ) , x O Spf ( A ) , x is faithfully flat, and similarly, the map O Spec ( B ) , y O Spf ( B ) , y is faithfully flat. Finally, the map O Spf ( B ) , y O Spf ( A ) , x is flat by [24, Proposition I, 5.3.11] and [49, Tag 0CF4]. Note that the compositions

O Spec ( B ) , y O Spec ( A ) , x O Spf ( A ) , x , O Spec ( B ) , y O Spf ( B ) , y O Spf ( A ) , x

are equal. By the discussion above, the second composition is flat, and thus the first. As

O Spec ( A ) , x O Spf ( A ) , x

is faithfully flat, we see O Spec ( B ) , y O Spec ( A ) , x is flat as desired.

Lemma A.4

Lemma A.4 (Generically étale Noether normalization)

Let 𝑌 be an affine scheme of finite type over a field 𝑘 which is geometrically reduced and whose irreducible components have the same dimension 𝑑. Then there exists a finite and generically étale map Y A k d .

Proof

Apply [34, Lemma 6][12] to the closure of Y A k n in P k n and a zero-dimensional subscheme of the smooth locus of 𝑌 meeting every irreducible component. ∎

Step 5. We claim that it is enough to show that

if d 2 , there is a hyperplane H D K d such that f η 1 ( H ) X is connected.

Here, by a hyperplane, we mean the zero set H = V ( h ) of a linear form h = i = 1 d a i x i a with a i , a O K with a i not all zero in 𝑘.

The argument for this is exactly as in [16]: we prove the entire theorem by induction on d = dim X 0 . Since the assertion is clear for d 1 , we assume d 2 . For the induction step, we find 𝐻 as in the claim, and since f η 1 ( H ) is curve-connected by the induction assumption, it suffices to connect a given classical point x X to a point y f η 1 ( H ) by a curve in 𝑋. Let p : D K d H be a linear projection (so p | H = id H ); then C = p 1 ( p ( f ( x ) ) ) is a connected curve (in fact, isomorphic to D L 1 for a finite extension L / K ) connecting f ( x ) with 𝐻. Let C be the connected component of f 1 ( C ) containing 𝑥. We claim that dim C = 1 ; otherwise, 𝑥 is an isolated point of f 1 ( C ) , which is impossible because the map

Spec ( A ) Spec ( K x 1 , , x d )

is open by [49, Tag 0F32][13]. The map C C is finite and hence surjective, and therefore C connects 𝑥 with a point in f 1 ( H ) .

In the following steps, we shall construct a hyperplane H ~ A k d and show that every hyperplane H D K d lifting H ~ has connected preimage in 𝑋.

Step 6 [Construction of 𝐻]. For a decomposition I J = { 1 , , r } with 𝐼, 𝐽 non-empty and disjoint, we set

Z I , J = ( i I Z i ) ( j J Z j ) ,

treated as a reduced subscheme of X ~ . We claim as in [16, § 6.4] that dim Z I , J = d 1 for every such I , J . To this end, note that X ~ Z I , J is disconnected, and hence so is U η X , where U Spf ( A ) is the open formal subscheme supported on X ~ Z I , J . But if dim Z I , J < d 1 , then by [39, Satz 2], we have O ( U η ) = A , which does not have nontrivial idempotents.

We denote by T ~ X ~ the closed subscheme where f ~ is not étale, and by S ~ X ~ the closed subscheme where f ~ is not flat. By construction, we have dim T ~ < d . By Miracle Flatness [49, Tag 00R4], S ~ is the non-Cohen–Macaulay locus of X ~ , see [49, Tag 00RE]. Since X ~ is reduced, it is S 1 , see [49, Tag 031R], and therefore we have dim S ~ < d 1 .

By the Bertini theorem [32, Théorème 6.3], a generic hyperplane (which exists after replacing 𝑘 by a finite extension if 𝑘 is finite) H ~ = V ( h ~ ) A k d , h ~ = i = 1 d a ~ i x i a ~ ( a ~ i , a ~ k ) satisfies the following properties.

  1. The intersections f ~ 1 ( H ~ ) Z i ( i = 1 , , r ) are irreducible of dimension d 1 and generically étale over H ~ .

  2. For every decomposition I J = { 1 , , r } with 𝐼 and 𝐽 non-empty and disjoint, the intersection f ~ 1 ( H ~ ) Z I , J has a component of dimension d 2 , and all such components are generically flat over H ~ .

Indeed, [32, Théorème 6.3 (1b), (4)] shows that an open set of hyperplanes will have the property that f ~ 1 ( H ~ ) Z i are irreducible of dimension d 1 . To ensure étaleness in (1), it suffices to choose H ~ not contained in f ~ ( T ~ ) . If V Z I , J is an irreducible component of dimension d 1 , then by [32, Théorème 6.3 (1b)] for a generic H ~ , the preimage f ~ 1 ( H ~ ) V is of dimension d 2 . To ensure flatness in (2), we choose a hyperplane whose intersection with f ~ ( V ) is not contained in f ~ ( S ~ ) .

We let h = i = 1 d a i x i a ( a i , a O K ) be any lifting of ℎ, and set H = V ( h ) D K d . In the remaining two steps, we shall prove that f η 1 ( H ) X is connected. By the claim in Step 5, this will finish the proof. To this end, if f η 1 ( H ) is disconnected, then we have Spec ( A / h A ) = T 1 T 2 for two non-empty closed subsets T 1 , T 2 such that T 1 T 2 is non-empty and set-theoretically contained in X ~ . In the final two steps, we shall derive a contradiction.

Step 7 [ T 1 T 2 does not contain any f ~ 1 ( H ~ ) Z i ]. Let ξ i be the generic point of f ~ 1 ( H ~ ) Z i ; we will show that ξ i T 1 T 2 . It is enough to show that A ξ i / h A ξ i is a domain. Indeed, since dim ( T 1 ) = dim ( T 2 ) = d if ξ i T 1 T 2 , then Spec ( A ξ i / h A ξ i ) would contain multiple components of Spec ( A / h A ) , but this would contradict that A ξ i / h A ξ i is a domain. To see that A ξ i / h A ξ i is a domain, consider the localization 𝑅 of O K x 1 , , x d at the generic point of H ~ and the map R A ξ i induced by 𝑓. Since 𝑓 is étale in a neighborhood of ξ i , the induced map R / h R A ξ i / h A ξ i is the composition of an étale morphism and a localization; in particular, it is weakly étale. Since O K x 1 , , x d / h O K y 1 , , y d 1 is normal, so is R / h R . Thus, by [49, Tag 0950], we conclude that A ξ i / h A ξ i is normal as well; in particular, it is a domain.

Step 8 [The contradiction]. It follows from the previous step that if we set

I = { i : ξ i T 1 } , J = { j : ξ j T 2 } ,

then these form a nontrivial partition of { 1 , , r } . Therefore, T 1 T 2 X ~ is contained in Z I , J . Note though that, since T 1 and T 2 both have codimension one in Spec ( A ) , the intersection T 1 T 2 X ~ necessarily contains a component 𝑉 of f ~ 1 ( H ~ ) Z I , J . This component can be taken to have codimension two in X ~ , and by our choice of hyperplane, we know that this component is generically Cohen–Macaulay and so not contained in S ~ .

Let 𝜁 be the generic point of 𝑉. By the previous paragraph, we know that 𝜁 is not contained in S ~ , and therefore A ~ ζ is a two-dimensional, Cohen–Macaulay, local ring. Moreover, h ~ is a non-zero divisor of A ~ ζ contained in its maximal ideal. Note that, since A ~ ζ / h ~ A ~ ζ is not zero-dimensional, its maximal ideal contains some non-zero divisor. Let g ~ in A ~ ζ be the lift of such an element. Then ( g ~ , h ~ ) is a regular sequence in A ~ ζ . Let 𝜛 be the pseudouniformizer from above and choose a lift 𝑔 of g ~ in A ζ . Note then that ( ϖ , g , h ) is a regular sequence in A ζ . Indeed, this is clear since A ζ / ϖ A ζ is a local ring and ( g , h ) have images in this ring that, in the further quotient ring, A ~ ζ have images forming a regular sequence. Therefore, ( ϖ , g ) is a regular sequence in the two-dimensional A ζ / h A ζ . The fact that we can permute a regular sequence follows from the argument given in [49, Tag 00LJ]. Indeed, while A ζ / h A ζ is not Noetherian, it is coherent (e. g. by [24, Chapter 0, Corollary 9.2.8]), and so the annihilator of any element of A ζ / h A ζ is finitely generated, which is all the referenced argument requires.

By Lemma A.5 below, we know that W = Spec ( A ζ / h A ζ ) V ( ϖ , g ) is connected. Note though that, since T 1 T 2 = Spec ( A / h A ) by construction, we have T 1 T 2 = W , where T i = T i W . Each T i is closed in 𝑊, and we have that T 1 T 2 V ( ϖ ) W . However, we have V ( ϖ ) W ) = : note that V ( ϖ ) W = Spec ( A ~ ζ / h ~ A ~ ζ ) V ( g ~ ) and this set is contained in the union of ξ i ’s that are in the intersection T 1 T 2 . As shown above, the set of those ξ i ’s is empty. Thus, T 1 and T 2 form a disconnection of 𝑊, a contradiction. ∎

Lemma A.5

Let 𝑅 be a local ring, let x , y R be such that ( x , y ) and ( y , x ) are regular sequences, and set W = Spec ( R ) V ( x , y ) . Then Γ ( W , O W ) = R ; in particular, 𝑊 is connected.

Proof

Since W = D ( x ) D ( y ) , we have Γ ( W , O W ) = ker ( R x × R y R x y ) . As 𝑥 is a non-zero divisor, R R x is injective, and hence so is R Γ ( W , O W ) . To show it is surjective, take an element ( a / x n , b / y m ) R x × R y in the kernel, i.e. ( x y ) N ( a y m b x n ) = 0 . Since x y is a non-zero divisor, we have a y m = b x n . We may assume that n = 0 or that a x R . If n > 0 , then we have a y m = 0 in R / x R , and hence a x R since 𝑦 is a non-zero divisor in R / x R . Therefore, n = 0 , analogously m = 0 , and hence a = b . ∎

We can bootstrap this up to connect points on connected rigid 𝐾-spaces which are not necessarily classical, at the expense of base field extension.

Corollary A.6

Let 𝑋 be a connected rigid 𝐾-space. Fix maximal points 𝑥 and 𝑦 in 𝑋. Then there exists a complete extension 𝐿 of 𝐾 and smooth connected affinoid 𝐿-curves C i with maps C i X such that, for all 𝑖, we have that im ( C i X ) im ( C i + 1 X ) is non-empty, x im ( C 1 X ) , and y im ( C m X ) .

Proof

As in the second step of the proof of Proposition A.2, we may assume that 𝑋 is geometrically connected. Let us then note that, by Gruson’s theorem [27, Théorème 1], one has that the completed tensor product k ( x ) ^ K k ( y ) is non-zero. Thus, by taking a point of Spa ( k ( x ) ^ K k ( y ) ) (which exists by [4, Theorem 1.2.1]) and looking at its residue field, we get a valued extension 𝐿 of 𝐾 containing both k ( x ) and k ( y ) . This gives two maps Spa ( L ) X . They give rise to two 𝐿-points x and y of X L mapping to 𝑥 and 𝑦, respectively. We then note that, by Proposition A.2, there exists smooth, connected, affinoid curves C 1 , , C n over 𝐿 and morphisms C i X L satisfying x im ( C 1 X L ) , y im ( C n X L ) , and

im ( C i X L ) im ( C i + 1 X L )

is non-empty. Clearly, then the compositions C i X L X satisfy the desired properties. ∎

Remark A.7

The referee pointed out to us a different (and in many ways simpler) proof of Proposition A.2, avoiding formal models and hence non-Noetherian difficulties. Instead, it relies on the foundational work of Ducros in [19, 21] surrounding the study of flatness in Berkovich analytic geometry. Because we think that both approaches contain valuable techniques, we have elected to include them both.

The key idea is to locally produce a smooth morphism with connected fibers to a curve, and then use the existence of multisections to deduce the curve-connectedness of the space from that of the fibers, so that we conclude by induction on dimension. This is enabled by two results of Ducros, or rather their following corollaries for rigid 𝐾-spaces.

Proposition A.8

Proposition A.8 (Existence of multisections, cf. [19, Theorem 9.1.3])

Let φ : Y X be a flat surjective map between quasi-compact rigid 𝐾-spaces, with 𝑌 quasi-separated and 𝑋 separated. Then there exist a flat quasi-finite surjective map ψ : X X and an 𝑋-morphism ι : X Y . If 𝜑 is smooth, then 𝜓 may be taken to be étale.

Proposition A.9

Proposition A.9 (Local découpage, cf. [21, Théorème 3.11])

Let φ : Y X be a flat quasi-compact and quasi-separated map with geometrically reduced fibers[14] between rigid 𝐾-spaces. Then there exists an étale cover { X i X } such that, for all 𝑖, the space X i is affinoid and the space Y i = Y × X X i admits a cover { Y i j } by quasi-compact opens such that the maps φ i j : Y i j X i have geometrically connected fibers.

For visual aid,

We shall now deduce these propositions from their Berkovich geometry versions, which requires some preparations that the reader may safely skip. We shall use the equivalence of categories

( ) Berk : { Taut rigid K -spaces } { Hausdorff strictly K -analytic spaces }

from [31, Remark 8.3.2], defined on affinoids by Spa ( A ) Berk = M ( A ) . In particular, 𝑋 is affinoid if and only if X Berk is affinoid, and a map 𝜑 of rigid 𝐾-spaces is a closed immersion if and only if φ Berk is.

By [31, Lemma 8.1.7], the composition of ( ) Berk with the forgetful functor to topological spaces agrees with universal separated quotient functor [ ] from § 2.1. In particular, as observed in [31], we have that 𝑋 is, respectively, connected and quasi-compact if and only if X Berk is (see Proposition 2.2.3). Finally, one sees from [31, Proposition 8.3.7] that 𝑋 is good if and only if X Berk is good.

Moreover, as the source category and target category are closed under fiber products with the category of all rigid 𝐾-spaces and strictly 𝐾-analytic spaces by [31, Lemma 5.1.4] and the fact that the natural map | Y × X X | | Y | × | X | | X | is compact in the sense of [5, § 1.4], respectively, the functor ( ) Berk commutes with fiber products in the naive sense. Continuing further, one sees that, for a morphism 𝜑 in the source category, one has that Δ φ Berk = Δ φ Berk and thus, from our previous observations concerning Zariski closed embeddings, that 𝜑 is separated if and only if φ Berk is.

The following lemma lists additional properties of the functor ( ) Berk which will be useful in the proof. For part (1), recall that, by definition, a morphism of rigid 𝐾-spaces φ : Y X is flat if and only if, for every y Y , the stalk O Y , y is flat over O X , φ ( y ) . In Berkovich geometry, such a naive notion is problematic (see [19, § 4.4]), and the actual definition of flat maps of 𝐾-analytic spaces is more involved; in particular, it imposes stability under base change (see [19, Definition 4.1.8]).

Lemma A.10

Let φ : X Y be a map of taut rigid 𝐾-spaces. Then

  1. 𝜑 is flat if and only if φ Berk is, in which case 𝜑 is open,[15]

  2. 𝜑 is quasi-finite if and only if φ Berk is,[16]

  3. 𝜑 is étale or partially proper étale if and only if φ Berk is quasi-étale or étale, respectively,

  4. if 𝑋 and 𝑌 are good, 𝜑 is smooth or partially proper and smooth if and only if φ Berk is quasi-smooth or smooth, respectively,

  5. if 𝜑 is either (i) quasi-compact and flat or (ii) partially proper, then 𝜑 is surjective if and only if φ Berk is.[17]

Proof

For (a), one quickly reduces to the case when X = Spa ( B ) and Y = Spa ( A ) . We then note that, by [19, Corollary 10.3.3], the map φ Berk is flat if and only if the map A B is flat. On the other hand, 𝜑 is flat if and only if A B is flat (cf. [48, Lemma B.4.3 and Remark B.4.5]). The claim concerning openness follows from [9, Corollary 9.4/2] and (b) follows from [31, Lemma 1.5.2]. Claim (c) follows from the discussion on [31, p. 427]. Claim (d) follows from [19, Theorem 5.4.6]. For claim (e), part (ii), the claim also follows from [19, Theorem 5.4.6]. For claim (e), part (i), we observe that, in this case, we quickly reduce to the case when 𝑌 is affinoid, and so φ ( X ) is quasi-compact open subset of 𝑌. The claim then follows from [30, Corollary 4.3]. ∎

Proof of Proposition A.8

Let us first observe that, by [31, Lemma 5.1.3], the spaces 𝑌 and 𝑋 are taut, and thus it makes sense to consider the map φ Berk : Y Berk X Berk . From the above discussion, in particular Lemma A.10, we see that Y Berk is compact and that X Berk is separated. The claim then follows by [19, Theorem 9.1.3] and the above discussion (in particular Lemma A.10). ∎

Proof of Proposition A.9

As 𝜑 is quasi-compact and quasi-separated, and the result is local on the target, by passing to an affinoid open cover of 𝑋, we may assume that 𝑋 is affinoid and 𝑌 is quasi-compact and quasi-separated. Again by [31, Lemma 5.1.3], the rigid 𝐾-space 𝑌 is taut, and so we may consider the map φ Berk : Y Berk X Berk . This map is flat and has geometrically connected fibers by assumption. Moreover, as Y Berk is compact and X Berk Hausdorff, we know that φ Berk is automatically topologically proper. The claim then follows by applying [21, Théorème 3.11] and the above discussion (in particular Lemma A.10). ∎

We now return to the second proof of Proposition A.2. In the process of the proof, we shall use the following reduction steps.

  1. If X = X i is an open cover with X i connected and the conclusion of Proposition A.2 holds for the X i , then it holds for 𝑋 (see Step 1 of the previous proof).

  2. If X X is a surjective map with X connected, and the conclusion of Proposition A.2 holds for X , then it holds for 𝑋.

  3. If 𝑋 is irreducible and U X is a non-empty Zariski open subspace, and the conclusion of Proposition A.2 holds for 𝑈,[18] then it holds for 𝑋.

To see (C), it is enough to show that, for any classical point x X U , there exists a classical point 𝑦 of 𝑈 which is connected to 𝑥 by curves. By (A), we may assume that 𝑋 is affinoid. Let f : X D K n be a finite map, which exists by Noether normalization (cf. [10, Corollary 6.1.2/2]). As 𝑋 is irreducible, we know that X U has strictly smaller dimension than 𝑋, and thus f ( X U ) is a proper Zariski closed subset of D K n (cf. [10, Proposition 9.6.3/3]). Take a point 𝑝 in D K n f ( X U ) and a line ℓ in D L n connecting f ( x ) and 𝑝, where 𝐿 is a finite extension of 𝐾 containing the residue fields of f ( x ) and 𝑝. Consider then any connected component 𝐶 of X × D K n . Observe that, as C is finite, it is clear that 𝐶 is an affinoid curve, and this map is surjective since the image is a Zariski closed subset (by [10, Proposition 9.6.3/3]) of dimension 1. In particular, the image of C X must contain a point 𝑞 of f 1 ( p ) which cannot be in X U and so must be in 𝑈.

Second proof of Proposition A.2

The proof will proceed by induction on dimension. The base case of dimension at most 1 being trivial, we assume that the claim has been verified for dimension less than dim ( X ) = n > 1 . As in [16, § 6.1 and 6.2], we reduce to the case of 𝑋 integral normal affinoid with non-empty smooth locus. By (C) and (A), we may assume that 𝑋 is a smooth connected affinoid, or even that it admits an etale morphism X D K n .

Let φ : X D K 1 = C be the composition of this map with projection onto the first coordinate, which is a smooth map whose fibers have dimension n 1 . Applying Proposition A.9 to 𝜑, we find an étale cover { C i C } and open covers { X i j } of X i = X × C C i by quasi-compact opens such that each map X i j C i has geometrically connected fibers. By applying (A) and (B), it suffices to prove the claim for each X i j . Let us then rename X i j to 𝑋 and φ ( X i j ) C i to 𝐶 so that 𝐶 is still a separated rigid 𝐾-curve and we have a smooth surjective φ : X C with (geometrically) connected fibers.

Applying Proposition A.8 to φ : X C gives an étale surjection ψ : C C and a 𝐶-morphism ι : C X . Let { C i } be the connected components of C , C i their images in 𝐶, and let X i denote the preimage of C i in X i . Applying (A) and (B) again, we may replace the triple ( X , C , C ) with ( X i , C i , C i ) , and so we may assume without loss of generality that C is connected. So let x 1 and x 2 be classical points of 𝑋 and set X i = φ 1 ( φ ( x i ) ) . By construction, dim ( X i ) n 1 , and so, if q i is any point of C lying over φ ( x i ) , then we have by induction hypothesis that x i may be connected to ι ( q i ) . But as evidently ι ( q 1 ) and ι ( q 2 ) are connected by C , we are done. ∎

Acknowledgements

We would like to thank Jeremy Brazas, Antoine Ducros, Gabriela Guzman, David Hansen, Johan de Jong, Kiran Kedlaya, Nakyung “Rachel” Lee, Emmanuel Lepage, Shizhang Li, Wiesława Nizioł, Jérôme Poineau, Peter Scholze, and Maciej Zdanowicz for useful conversations and comments on the preliminary draft of this manuscript. We especially thank Ofer Gabber for pointing out several inaccuracies in previous versions of this manuscript and suggesting fixes. We would also like to thank an anonymous referee who gave numerous suggestions which greatly improved the readability of the document, as well as for suggesting an alternate proof to the result in the appendix.

References

[1] P. Achinger, M. Lara and A. Youcis, Variants of the de Jong fundamental group, preprint (2021), http://arxiv.org/abs/2203.11750. Search in Google Scholar

[2] P. Achinger, M. Lara and A. Youcis, Specialization for the pro-étale fundamental group, Compos. Math. 158 (2022), no. 8, 1713–1745. Search in Google Scholar

[3] Y. André, Period mappings and differential equations. From ℂ to C p , MSJ Mem. 12, Mathematical Society of Japan, Tokyo 2003. Search in Google Scholar

[4] V. G. Berkovich, Spectral theory and analytic geometry over non-Archimedean fields, Math. Surveys Monogr. 33, American Mathematical Society, Providence 1990. Search in Google Scholar

[5] V. G. Berkovich, Étale cohomology for non-Archimedean analytic spaces, Publ. Math. Inst. Hautes Études Sci. 78 (1993), 5–161. Search in Google Scholar

[6] V. G. Berkovich, Smooth 𝑝-adic analytic spaces are locally contractible, Invent. Math. 137 (1999), no. 1, 1–84. Search in Google Scholar

[7] V. G. Berkovich, Integration of one-forms on 𝑝-adic analytic spaces, Ann. of Math. Stud. 162, Princeton University, Princeton 2007. Search in Google Scholar

[8] B. Bhatt and P. Scholze, The pro-étale topology for schemes, De la géométrie algébrique aux formes automorphes (I) Astérisque 369, Société Mathématique de France, Paris (2015), 99–201. Search in Google Scholar

[9] S. Bosch, Lectures on formal and rigid geometry, Lecture Notes in Math. 2105, Springer, Cham 2014. Search in Google Scholar

[10] S. Bosch, U. Güntzer and R. Remmert, Non-Archimedean analysis, Grundlehren Math. Wiss. 261, Springer, Berlin 1984. Search in Google Scholar

[11] S. Bosch, W. Lütkebohmert and M. Raynaud, Formal and rigid geometry. IV. The reduced fibre theorem, Invent. Math. 119 (1995), no. 2, 361–398. Search in Google Scholar

[12] J. Brazas, Semicoverings: A generalization of covering space theory, Homology Homotopy Appl. 14 (2012), no. 1, 33–63. Search in Google Scholar

[13] B. Conrad, Irreducible components of rigid spaces, Ann. Inst. Fourier (Grenoble) 49 (1999), no. 2, 473–541. Search in Google Scholar

[14] B. Conrad, Modular curves and rigid-analytic spaces, Pure Appl. Math. Q. 2 (2006), no. 1, 29–110. Search in Google Scholar

[15] B. Conrad and M. Temkin, Descent for nonarchimedean analytic spaces, Tunis. J. Math. 3 (2021), no. 4, 689–748. Search in Google Scholar

[16] A. J. de Jong, Crystalline Dieudonné module theory via formal and rigid geometry, Publ. Math. Inst. Hautes Études Sci. 82 (1995), 5–96. Search in Google Scholar

[17] A. J. de Jong, Étale fundamental groups of non-Archimedean analytic spaces, Compos. Math. 97 (1995), no. 1–2, 89–118. Search in Google Scholar

[18] P. Deligne and M. Rapoport, Les schémas de modules de courbes elliptiques, Modular functions of one variable. II (Antwerp 1972), Lecture Notes in Math. 349, Springer, Berlin (1973), 143–316. Search in Google Scholar

[19] A. Ducros, Families of Berkovich spaces, Astérisque 400, Société Mathématique de France, Paris 2018. Search in Google Scholar

[20] A. Ducros, La structure des courbes analytiques, https://webusers.imj-prg.fr/~antoine.ducros/trirss.pdf. Search in Google Scholar

[21] A. Ducros, Dévisser, découper, éclater et aplatir les espaces de Berkovich, Compos. Math. 157 (2021), no. 2, 236–302. Search in Google Scholar

[22] J. Fresnel and M. Matignon, Sur les espaces analytiques quasi-compacts de dimension 1 sur un corps valué complet ultramétrique, Ann. Mat. Pura Appl. (4) 145 (1986), 159–210. Search in Google Scholar

[23] K. Fujiwara and F. Kato, Rigid geometry and applications, Moduli spaces and arithmetic geometry, Adv. Stud. Pure Math. 45, Mathematical Society of Japan, Tokyo (2006), 327–386. Search in Google Scholar

[24] K. Fujiwara and F. Kato, Foundations of rigid geometry. I, EMS Monogr. Math., European Mathematical Society, Zürich 2018. Search in Google Scholar

[25] S. Gaulhiac, Comparison between admissible and de Jong coverings of rigid analytic spaces in mixed characteristic, preprint (2022), https://arxiv.org/abs/2201.08065. Search in Google Scholar

[26] H. Grauert and R. Remmert, Über die Methode der diskret bewerteten Ringe in der nicht-archimedischen Analysis, Invent. Math. 2 (1966), 87–133. Search in Google Scholar

[27] L. Gruson, Théorie de Fredholm 𝑝-adique, Bull. Soc. Math. France 94 (1966), 67–95. Search in Google Scholar

[28] D. Hansen, Vanishing and comparison theorems in rigid analytic geometry, Compos. Math. 156 (2020), no. 2, 299–324. Search in Google Scholar

[29] E. Hrushovski and F. Loeser, Non-archimedean tame topology and stably dominated types, Ann. of Math. Stud. 192, Princeton University, Princeton 2016. Search in Google Scholar

[30] R. Huber, Continuous valuations, Math. Z. 212 (1993), no. 3, 455–477. Search in Google Scholar

[31] R. Huber, Étale cohomology of rigid analytic varieties and adic spaces, Aspects of Math. E30, Friedrich Vieweg & Sohn, Braunschweig 1996. Search in Google Scholar

[32] J.-P. Jouanolou, Théorèmes de Bertini et applications, Progr. Math. 42, Birkhäuser, Boston 1983. Search in Google Scholar

[33] M. Kashiwara and P. Schapira, Categories and sheaves, Grundlehren Math. Wiss. 332, Springer, Berlin 2006. Search in Google Scholar

[34] K. S. Kedlaya, More étale covers of affine spaces in positive characteristic, J. Algebraic Geom. 14 (2005), no. 1, 187–192. Search in Google Scholar

[35] K. S. Kedlaya and R. Liu, Relative 𝑝-adic Hodge theory: Foundations, Astérisque 371, Société Mathématique de France, Paris 2015. Search in Google Scholar

[36] E. Lepage, Géométrie anabélienne tempéré, preprint (2010), https://arxiv.org/abs/1004.2150. Search in Google Scholar

[37] E. Lepage, Tempered fundamental group and metric graph of a Mumford curve, Publ. Res. Inst. Math. Sci. 46 (2010), no. 4, 849–897. Search in Google Scholar

[38] J. Ludwig, Topics in the theory of adic spaces and the eigenvariety machine, Lecture notes, Heidelberg University, 2020, https://typo.iwr.uni-heidelberg.de/fileadmin/groups/arithgeo/templates/data/Judith_Ludwig/home/Notes_Adic_Spaces_2.pdf. Search in Google Scholar

[39] W. Lütkebohmert, Fortsetzbarkeit 𝑘-meromorpher Funktionen, Math. Ann. 220 (1976), no. 3, 273–284. Search in Google Scholar

[40] S. MacLane, Categories for the working mathematician, Grad. Texts in Math. 5, Springer, New York 1971. Search in Google Scholar

[41] P. Scholze, 𝑝-adic Hodge theory for rigid-analytic varieties, Forum Math. Pi 1 (2013), Paper No. 77. Search in Google Scholar

[42] P. Scholze, Étale cohomology of diamonds, preprint (2017), https://arxiv.org/abs/1709.07343. Search in Google Scholar

[43] M. Temkin, On local properties of non-Archimedean analytic spaces. II, Israel J. Math. 140 (2004), 1–27. Search in Google Scholar

[44] M. Temkin, Introduction to Berkovich analytic spaces, Berkovich spaces and applications, Lecture Notes in Math. 2119, Springer, Cham (2015), 3–66. Search in Google Scholar

[45] R. Vakil, The rising sea, Lecture notes, Stanford University, 2017, http://math.stanford.edu/~vakil/216blog/FOAGnov1817public.pdf. Search in Google Scholar

[46] E. Warner, Adic moduli spaces, PhD thesis, Stanford University, 2017, https://purl.stanford.edu/vp223xm9859. Search in Google Scholar

[47] S. Willard, General topology, Addison-Wesley, Reading 1970. Search in Google Scholar

[48] B. Zavyalov, Quotients of admissible formal schemes and adic spaces by finite groups, preprint (2021), https://arxiv.org/abs/2102.02762. Search in Google Scholar

[49] The Stacks Project Authors, Stacks Project, 2021, http://stacks.math.columbia.edu. Search in Google Scholar

Received: 2022-03-07
Revised: 2023-01-09
Published Online: 2023-04-20

© 2023 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 28.5.2023 from https://www.degruyter.com/document/doi/10.1515/crelle-2023-0013/html
Scroll to top button