arXiv:2107.09165v2 [math.NT] 20 Sep 2021 where oki oivsiaeti question. this investigate to is work rmti,w bantecmlxuiomzto of uniformization complex the obtain we this, From e sfis suppose first us Let ena etrgopover group vector a as seen with eoea bla ait over variety abelian an denote ensteeatsequence: exact the defines owne fasmlrcasfiaineit o bla varie abelian for exists classification similar a if wonder to hti h nrdcin nfc ntePoou f[ of Prologue the in Let fact picture. in Introduction, the in that [ Scholze–Weinstein of article the rusover groups eoeteaslt aosgopof group Galois absolute the denote ON hswr a w anisiainsucs [ sources: inspiration main two has work This sacneuneo eiso rilssatn ihTate’s with starting articles of series a of consequence a As 2020 Date e od n phrases. and words Key τ p Λ h ope ojgto,or conjugation, complex the AI NFRIAINO BLA AITE IHGO REDUCT GOOD WITH VARIETIES ABELIAN OF UNIFORMIZATION -ADIC otieitga sotnijcie npriua,i sp is it particular, In injective. often is integral Fontaine ie egv w plctoso hsrsl.Frt eexte we of First, cover result. this universal of applications two give We tive. G oso onso h omlgopof group formal the of points torsion fiainpoete ftefil xeso of extension field the of Sec properties uniformization. ification complex classical the resembles which naie xeso of extension unramified lsial,teFnan nerlwsse saHodge–Tate a as seen was ϕ integral Fontaine the Classically, A etme 1 2021. 21, September : ahmtc ujc Classification. Subject Mathematics F A BSTRACT steiaeo h lattice the of image the is h rsn ril trswt h bevto hti edo we if that observation the with starts article present The h boueGli ru of group Galois absolute the ⊗ K 1 O C IHA PEDXB EKHYJSU A N LXNE PETROV ALEXANDER AND LAM JOSHUA HAY YEUK BY APPENDIX AN WITH eoetecmlto fanme edadcos nalgebraic an choose and field number a of completion the denote p C DINIVT,JCSNS ORW N LXNR ZAHARESCU, ALEXANDRU AND MORROW, S. JACKSON IOVITA, ADRIAN : p Let . T p ntrso hi og–aesrcue.W eiv ti le a is it believe We structures. Hodge–Tate their of terms in K ( A sarchimedean. is p ) ⊗ eartoa rm,let prime, rational a be p A ai nfriain bla aite,Fnan integr Fontaine varieties, Abelian Uniformization, -adic Z n hwta if that show and p 0 C → K p Q . p → H , K K 1 Lie ( A SW13 fdimension of oefie leri lsr of closure algebraic fixed some A F H ( ( Let . A ( nti aew have we case this In C 11 1K0 12,14L05). 11G25, (14K20, 11G10 C 1 G )( K ( ) ) A A F K , ,w o aeacmlt lsicto fthe of classification complete a have now we ], = ) T ∼ by . A Z ( p ⊗ = 1. C F ( Lie ea bla ait endover defined variety abelian an be ) F A eoeafiie naie xeso of extension unramified finite, a denote ) G C → { , ) Introduction K 1 G K Z ( p } A K bandb donn h oriae fthe of coordinates the adjoining by obtained ( h xoeta a seeyhr endand defined everywhere is map exponential The . ) Lie := 1 . )( g ) = n ssc ti ujcieadhsalrekernel. large a has and surjective is it such as and , 1 C ( Gal > 0 Fon03 A ) then A / Fon03 )( 1 Λ oe htif that roved ( ..a smrhs fcmlxLie-groups: complex of isomorphism an i.e. , n e Lie let and , C K dteFnan nerlt efcodlike perfectoid a to integral Fontaine the nd = A ) / ∼ n,w rv eut ocrn h ram- the concering results prove we ond, n [ and ] K ( K exp K C → ,J-.Fnan rsnstefollowing the presents Fontaine J.-M. ], isoe a over ties ) ) = oprsnmrhs,ie samap a as i.e. morphism, comparison g A n by and a yeof type a has K / o tensor not R and , A Λ BK07 ( , or T rgnl[ original C p ( ) ( A C K A → p C ) ) .T tr ih e srecall us let with, start To ]. = G p h opeinof completion the eoeteLeagbaof algebra Lie the denote T ation, h opeinof completion the K 0 p ai ed h olo this of goal The field. -adic F C . p ihgo reduction. good with , ( = A ai uniformization, -adic , Tat67 ) K 0 Q p then , with ai Dynamics. -adic = p , K C n nigwith ending and ] closure C iiaequestion gitimate h maximal the ϕ p = A hnthe then , p C sinjec- is -power K , Let . G p K K -divisible K of = Let . ION K { 1 We . , τ A A } , , Now suppose that K is a finite extension of Qp for p > 2 a rational prime. In this case, we will denote Cp := C. Then the logarithm is everywhere defined and we have a natural commutative diagram with exact rows:
logA 0 Ator(K) A(K) Lie(A)(K) 0
logA 0 Ator(Cp) A(Cp) Lie(A)(Cp) 0.
The group of points A(K) has a natural topology (see Subsection 2.14) whose completion is A(Cp) and which induces the discrete topology on Ator(K) and the natural, p-adic topology on Lie(A)(K). Let Ator(K) denote the torsion subgroup of A(K). We have the following decomposition
Ator(K)= Ap−tor(K) ⊕ Ap ′−tor(K) into the p-power torsion Ap−tor(K) and the prime to p-torsion Ap ′−tor(K). Fontaine has constructed a section s: A(Cp) Ap ′−tor(K) of the natural inclusion Ap ′−tor(K) ⊂ A(Cp) and we denote the (p) kernel of s by A (Cp) (see Section 6 or [Fon03] for more details on the construction of s). As such, we have the decompositions→ (p) A(Cp)= A (Cp) ⊕ Ap ′−tor(K) and (p) A(K)= A (K) ⊕ Ap ′−tor(K) (p) (p) where we let A (K) := A (Cp) ∩ A(K). (p) (p) Fontaine states, as a remarkable fact, that we can recover A (K), and so also A (Cp), as topo- logical abelian groups with GK-action, from the knowledge of Ap−tor(K). Moreover, this implies that we can recover A(K), respectively A(Cp) from the knowledge of Ator(K) (cf. [Fon03, Proposi- tion 1.1]). 1.1. Main contributions. The present article has as objective to show that, under certain circum- stances, by changing the topology of A(p)(K), this new topological group can be determined in a different way from Ap−tor(K). First, we recall that in [Fon82], Fontaine constructed an integration map ϕA : Tp(A) Lie(A)(K) ⊗K Cp(1), and when this map was tensored with Cp, it realizes the Hodge–Tate comparison morphism. In particular, the map →
ϕA ⊗ 1Cp : Tp(A) ⊗Zp Cp Lie(A)(K) ⊗K Cp(1), is surjective and it has a large kernel, and to the best of our knowledge, the Fontaine integral has always been seen as a Hodge–Tate comparison morphism.→ A starting point for this article is the observation that if we do not tensor Tp(A) with Cp, then the Fontaine integral is often injective. To state the results precisely, we need to establish some notation. Let now K denote the maximal unramified extension of Qp, let A denote an abelian variety defined over some subfield F ⊂ K such that [F : Qp] < , with good reduction over F. Let A denote the Néron model of A over Spec(OF). G We present a proof, whose sketch was supplied by Pierre Colmez, that if Tp(A) K = 0, then Fontaine’s integration∞ map ϕA : Tp(A) Lie(A)(F) ⊗F Cp(1) is injective. There is another proof of this fact in the Appendix by Yeuk Hay Joshua Lam and Alexander Petrov (independently). In fact, we have more; to describe this,→ we need to briefly recall some definitions. We define the universal covering space of A(K)= A(OK) to be [p] [p] [p] B := A(O ) − A(O ) − · · · − A(O ) · · · A lim− K K K , 2 ← ← ← ← and we call an element u = (un)n>0 of BA a path. It is clear that BA is a GK-module which sits in the following exact sequence: ∼ α 0 Tp(A) = Tp(A) BA A(OK) 0, with α ((un)n>0) := u0 for all u = (un)n>0 ∈ BA. In Definition 2.12, we extend the classical Fontaine integral to a non-zero,→GF-equivariant map→ → →
ϕA : BA Lie(A)(OF) ⊗F Cp(1), and with this definition, it is clear that if a path u = (un)n>0 is periodic (i.e., there exists some k > 1 such that u0 = uk), then ϕA(u)(ω)=→ 0 (see Remark 2.13 for details). Our first result proves that the kernel of this extended Fontaine integral is precisely the periodic paths. Theorem A. Let A be an abelian variety over F with good reduction, and let A denote its Néron model. G Suppose that Tp(A) K = 0. Then, the kernel of the Fontaine integral ϕA extended to BA is precisely the subgroup of periodic paths of BA (Definition 3.6). G We give two applications of Theorem A. First, we show that if A satisfies Tp(A) K = 0, then (p) A (K), endowed with a new topology, can be determined in a different way from Ap−tor(K). ur Theorem B. Let K := Qp denote the maximal unramified extension of Qp. Let A denote an abelian variety defined over some subfield F ⊂ K such that [F : Qp] < , with good reduction over F. Let A denote the Néron model of A over Spec(OF). G Suppose that A satisfies Tp(A) K = 0. Then, there∞ exists a canonical, injective, continuous, GF- equivariant map (p) ,((ιA : A (K) ֒ (Lie(A)(F) ⊗F Cp(1)) /ϕA(Tp(A where the topology on A(p)(K) which makes this embedding continuous is the w-topology (Definition 6.1).
Moreover, an element x ∈ (Lie(A)(O→F) ⊗OF Cp(1)) /ϕA(Tp(A)) lies in the image of ιA if and only if x is crystalline (Definition 6.4). In particular, one can recover A(p)(K) from the triple
. ((Tp(A),Lie(A)(F) ⊗F Cp(1), ϕA : Tp(A) ֒ Lie(A)(F) ⊗F Cp(1) Remark 1.2. The fact that the field F of definition of the abelian variety A in the above theorem, is unramified is only used in Section 4.7 and in the Appendix.→ It seems clear that with more work all the results of this paper, suitably adjusted, should hold when F is a finite extension of Qp. Remark 1.3. In fact, strangely enough, we are able to prove analogues of the above results, on the kernel of the Fontaine integral and the uniformization, for the rigid analytic multiplicative group, that is G := Spm FhX, X−1i , with the multiplicative group law. Then the K-points of this group are G(K) = (O×, ·), i.e. the multiplicative group of O . K K This is strange as we do not have such a result for the multiplicative group over C. After some meditation on these results, we dare speculate that they should be shadows of the fact that such results would hold for Tate-curves, and what we see is the "crystalline" part of the logarithmic uniformization of some Tate-curve. Our second application concerns interesting properties regarding the ramification of the p- power torsion points of an abelian variety with good reduction over F. In particular, we prove the following result. Theorem C. Let A be an abelian variety over F with good reduction, and let A denote the formal group associated to the Néron model of A. Then there is n0 > 1 such that for every m > n0 and 0 6= P ∈ m (1) (1) A[p ](O), we have P ∈/ A(O ) where O is the kernel of the canonical derivativeb map on OF. Finally, we give a criterion for when the integer n from in Theorem C is equal to 1 (see Section b b 0 8.4, Definition 8.5, Theorem 8.8, and Proposition 8.9). 3 1.4. Related results. Theorem A shows that the zeros of the Fontaine integral coincide with the G periodic paths on BA when Tp(A) K = 0. The consideration of zeros of p-adic integrals (in the sense of Berkovich [Ber06], Coleman [Col85b], Coleman–de Shalit [CDS88], and Zarhin [Zar96]) has led to several deep results in Diophantine geometry. We refer the reader to [Cha41, Col85a, Col85b, Col87, Sto18, KRZB16, Kim05, Kim09] for details and the survey [KRZB18]. G Our Theorem B states that if A satisfies Tp(A) K = 0, then A(K) has a type of p-adic uniformiza- tion, which resembles the classical complex uniformization. The history of p-adic uniformization of abelian varieties is rich and beautiful, and we briefly exposit it below. The first work in this area was due to Tate [Tat71] who showed that an elliptic curve with mul- tiplicative reduction is uniformized, as a rigid analytic space, by the rigid analytification of the multiplicative group. Later, Bosch–Lütkebohmert [BL85, BL84] constructed a topological univer- sal cover for the rigid analytic space associated to an abelian variety and isolated a class of abelian varieties, namely those with toric reduction, whose topological uniformization resembles complex uniformization. Subsequent developments in p-adic geometry by Berkovich [Ber90, Ber93] gave rise to p-adic analytic spaces which have topological properties similar to those of complex man- ifolds. Using this theory, Berkovich [Ber99] showed that a smooth, connected Berkovich analytic space has a topological universal cover, and this result was later generalized by Hrushovshi– Loeser [HL16] to quasi-projective Berkovich analytic spaces. While Berkovich spaces have nice topological properties, adic spaces (in the sense of Huber [Hub94]) do not (e.g., they do not pos- sess paths). That being said, one can use the theory of perfectoid spaces [Sch12] and diamonds [SW20, Sch21] to construct certain pro-étale uniformizations of the adic space associated to an abelian variety (c.f. [BGH+20, Heu21]). To the best of our knowledge, Theorem B is the first result which gives a p-adic uniformization of abelian varieties with good reduction which is similar to the classical complex uniformization.
1.5. Future work. The firstquestiontoask is if our resultsextendto abelian varieties with semistable reduction over K, or even any abelian variety over K. The first author together with Nicola Maz- zari and Khai Hoan Nguyen have started to work on this and hope to soon report on it. In a different direction, Fontaine, in [Fon03], associates two interesting objects to Ap−tor(K): the (p) (p) topological group A (Cp), endowed with its GK-action and the rigid analytic group A , whose (p) Cp-points form the topological group A (Cp) above. Additionally, if the abelian variety has good reduction, we can associate to Ap−tor(K) a third object, namely a sheaf of abelian groups for the flat topology. The main goal of the article [Fon03] is to generalize loc. cit. Proposition 1 and these three constructions to motives. We believe that the constructions from Theorem B could also generalize to certain motives at- tached to algebraic varieties over p-adic fields.
1.6. Organization of paper. In Section 2, we recall the construction of Fontaine integration and discuss the role this integration theory plays in the context of Hodge–Tate and de Rham compari- son isomorphisms for abelian varieties. In Section 3, we state Theorem A as well as how it relates to previous literature and to p-adic dynamical systems. In Section 4, we show that Theorem A can be reduced to the injectivity of the Fontaine integral restricted to the Tate module of the formal group and prove this, following a sketch provided by Pierre Colmez. The last three sections describe our applications of Theorem A. InSection6, weproveTheorem B, which is our p-adic uniformization result for A(p)(K), and in Section 7, we discuss the analogues of our works to the rigid multiplicative group. In Section 8, we give a different perspective on Fontaine integration, which allows us to prove Theorem C, and then in Section 8.4, we give a criterion to make Theorem C effective.
1.7. Conventions. We establish the following notations and conventions throughout the paper. 4 ur Fields. Fix a rational prime p > 2. Let K := Qp denote the maximal unramified extension of Qp, let K be a fixed algebraic closure of K, and let Cp denote the completion of K with respect to the unique extension v of the p-adic valuation on Qp (normalized such that v(p) = 1). For a tower of field extensions Qp ⊂ F ⊂ K, we denote by GK and respectively GF the absolute Galois groups of K and F respectively. We denote O := OK. We remark that working over K, as we do in this article, ur is not essential. One could start by fixing a finite extension L of Qp, and end-up working on L .
Abelian varieties. We will consider an abelian variety A defined over some subfield F ⊂ K such that [F : Qp] < , with good reduction over F. Let A denote the Néron model of A over Spec(OF) and also denote by A the formal completion of A along the identity of its special fiber, i.e. the formal group of A∞. We note that the formation of Néron models commutes with unramified base change. We will denote theb Tate module of A (resp. the Néron model A of A) by Tp(A) (resp. Tp(A)). We ∼ note that Tp(A) = Tp(A) as GF-modules. We recall that A is a formal group of dimension dim(A) and of height h which satisfies dim(A) 6 h 6 2 dim(A). b Acknowledgments. This article owes much to many people. We thank Robert Benedetto, Olivier brinon, Henri Darmon, Eyal Goren and Ralph Greenberg for helpful conversations and email ex- changes on topics related to this research. We are grateful to Pierre Colmez for sending us a sketch of the proof of Theorem4.8 and to Yeuk Hay Joshua Lam and Alexander Petrov for providing us with the proof presented in the Appendix. We also thank Pierre Colmez and Jan Nekovar for pointing out some errors in earlier drafts of this paper.
2. Fontaine integration for abelian varieties with good reduction In this section, we recall the construction of the Fontaine integration and the extension of this integration theory to a certain universal cover of an abelian variety. We also describe several topological aspects of the integration theory and discuss how these integration theories realize the Hodge–Tate and de Rham comparison isomorphisms for abelian varieties.
2.1. The differentials of the algebraic integers. First, we recall for the reader’s convenience the ur notation established above. Let K := Qp denote the maximal unramified extension of Qp, let K be a fixed algebraic closure of K, and let Cp denote the completion of K. Let GK denote the absolute Galois group of K. We denote O := OK. Fix a finite extension F of Qp in K. In [Fon82], Fontaine studied a fundamental object related to these choices, namely the O- module Ω := Ω1 =∼ Ω1 of Kähler differentials of O over O , or over O . The O-module Ω O/OK O/OF K F is a torsion and p-divisible O-module, with a semi-linear action of GF. Let d: O Ω denote the canonical derivation, which is surjective. Important examples of algebraic differentials arise as follows: Let (εn) denote→ a compatible sequence of primitive pth roots of unity in K. Then
dεn dε + dεn = d(log ε ) ∈ Ω and p n 1 = . ε n ε ε n n+I n Next, we recall a theorem of Fontaine.
Theorem 2.2 ([Fon82, Théorème 1’]). Let (εn) denote a compatible sequence of primitive pth roots of unity in K. The morphism ξ: K(1) Ω defined by
dεr → ξ(α ⊗ (εn)n)= a εr 5 r where α = a/p for some a ∈ O is surjective and GK-equivariant with kernel 1 ker(ξ)= a := x ∈ K : v(x) > − . K p − 1
∼ ∼ ∼ Moreover, Ω = K(1)/aK(1) = (K/aK)(1) and Vp(Ω)= HomZp (Qp, Ω) = Cp(1). Theorem 2.2 implies the following:
p p p T (Ω) ⊗ Q := lim Ω[pn] ⊗ Q =∼ lim Ω Ω · · · Ω · · · ⊗ Q =∼ C (1) p Zp p − Zp p − Zp p p n ! as GF-modules. ← ← ← (1) ← ← We denote by O := ker(d), the kernel of d, which is an OK-sub-algebra of O. Indeed, if a, b ∈ O(1), then d(ab) = ad(b)+ bd(a) = 0, and so ab ∈ O(1). In order to better understand O(1), we recall a construction from [IZ99]. Definition 2.3. Let a ∈ O. Let L/K be a finite extension which contains a, let π be a uniformizer of L, and let f ∈ OK[x] be such that a = f(π). Then, define f′(π) δ(a) := min v , 0 ∆ L/K where ∆L/K denotes the different ideal of L/K. Note that δ does not depend on π, f, or F, and so it defines a function δ: O (− , 0]. Lemma 2.4 (Properties of δ). The function δ from Definition 2.3 satisfies the following properties. → ∞ (1) If a, b ∈ O, then δ(a + b) > min(δ(a), δ(b)), and if δ(a) 6= δ(b), then we have equality. (2) If a, b ∈ O, then δ(ab) > min(δ(a)+ v(b), δ(b)+ v(a)). ′ (3) If f ∈ OK[x] and α ∈ O, then δ(f(α)) = min(v(f (θ)) + δ(θ), 0). (4) If x, y ∈, then xdy = 0 if and only if v(x)+ δ(y) > 0. (5) For a ∈ O, δ(a)= 0 if and only if a ∈ O(1). (6) The formula δ(adb) := min(v(a)+ δ(b), 0) is well-defined and give a map δ: Ω (− , 0], which makes the obvious diagram commutative. We will use the follow properties of δ in our study of the Fontaine integral. → ∞
Lemma 2.5 ([IZ99, Lemma 2.2]). Let a, b ∈ O be such that δ(a) 6 δ(b). Then there exists c ∈ OK[a,b] such that cda = db. Proposition 2.6 ([IZ99, Theorem 2.2]). Let L/K be an algebraic extension. Then L is deeply ramified (loc. cit. Definition 1.1) if and only if δ(OL) is unbounded. To conclude our discussion on differentials of the algebraic integers, we note that the derivation map d: O Ω is not continuous with respect to the p-adic topology on O and the discrete topology on Ω. Nevertheless, there is another topology on O for which d is continuous, namely the topology→ defined by the valuation w: O − Z ∪ { } defined by w(a) := sup{n ∈ Z | a ∈ pnO(1)}, where O(1) = ker(d). We summarize this result in the following lemma. Lemma 2.7. Consider O with the topology defined→ by the∞ valuation w and endow Ω with the discrete topology. Then, the derivation map d: O Ω is continuous. 2.8. A universal cover of an abelian variety. Classically, the Fontaine integral is defined as a map → ϕA : Tp(A) Lie(A)(F) ⊗F Cp(1). For our purposes, we will want to integrate along a larger collection of paths, which leads to the following definition. We note that this construction→ appears in [CI99, Part II, Section 3]. 6 Definition 2.9. We define the universal covering space of A(K) to be
[p] [p] [p] [p] [p] [p] B := A(O) − A(O) − · · · − A(O) · · · = A(K) − A(K) − · · · − A(K) · · · A lim− lim− where the equality← comes from← the properness← of A. We call an element← u =← (un)n>0←of BA a path. ← ← We immediately have that BA is a GF-module which sits in the following exact sequence: α 0 Tp(A)= Tp(A) BA A(O) 0, (2.1) with α (u) := u0 for all u = (un)n>0 ∈ BA. → → → → Remark 2.10. The universal covering space from Definition 2.9 is similar to the construction of the perfectoid universal cover of an abelian variety (see e.g., [PS16, Lemma 4.11] and [BGH+20]). We refrain from using this terminology as we work over the field K, which is not perfectoid, and hence BA cannot come from any perfectoid space. We also mention that [Col98, Appendice B] discusses universal covering spaces of abelian varieties over p-adic fields. 2.11. The definition of Fontaine integration. We are now ready to define Fontaine integration. Let H0(A, Ω1 ) and respectively Lie(A)(O ) denote the O -modules of invariant differentials A/OF F F on A and respectively its Lie algebra.
0 1 Definition 2.12. Let u = (u ) N ∈ B and ω ∈ H (A, Ω ). Each u ∈A(O) corresponds to n n∈ A A/OF n a morphism un : Spec(O) A, and hence we can pullback ω along this map giving us a Kähler ∗ ∗ differential un(ω) ∈ Ω. The sequence (un(ω))n>0 is a sequence of differentials in Ω satisfying ∗ ∗ ∼ pun+1(ω)= un(ω), and hence→ defines an element in Vp(Ω) = Cp(1). The Fontaine integration map
ϕA : BA Lie(A)(OF) ⊗OF Cp(1) is a non-zero, GF-equivariant map defined by → ∗ ∼ ϕA(u)(ω):=(un(ω))n>0 ∈ Vp(Ω) = Cp(1). Remark 2.13. Using Theorem 2.2 and the function δ from Definition 2.3, we can give an alternative 0 1 description of the Fontaine integration map. Let u = (u ) > ∈ B and ω ∈ H (A, Ω ). Each n n 0 A A/OF un ∈ A(O) corresponds to a morphism un : Spec(O) A, and hence we can pullback ω along ∗ this map giving us a Kähler differential un(ω) ∈ Ω. > > ∗ For every n 0, there is a maximal m(n) 0 such that→ un(ω)= αn(dεm(n)/εm(n)) with αn ∈ O m(n) where εm(n) is some primitive p -th root of unity. To see this, we first note that dε 1 δ r =−r − ε pr(p − 1) r for any primitive pr-th root of unity. This result follow from the definition of δ and a result of Tate [Tat67, Proposition 5] on the valuation of the different ideal of K(εr)/K. By taking m(n) = ∗ −[δ(un(ω))] where [x] denotes the greatest integer of the real number x, we can use Lemma 2.4.(6) and Lemma 2.5 to deduce the above equality. Now using Theorem 2.2, we have that n−m(n) ϕA(u)(ω)= lim p αn ∈ Cp. n→∞ G Moreover, using the definition of δ and this above interpretation, we can see that if u ∈ (BA) K (i.e., if u is an unramified path), then ϕA(u)(ω)= 0. Indeed, it is clear from the definition of δ that m(n)= 0. 7 2.14. Topological aspects of the extended Fontaine integral. Recall the notation from Subsection 1.7. We have two GF-modules, namely A(K)= A(O) and BA, which was defined in Definition 2.9. A priori, A(K) = A(O) and BA are just abelian groups, but we can endow them with natural topologies to enhance them to topological abelian groups as follows. Fontaine defines the following natural topology on A(K), namely the coarsest topology for which the maps A(O) A(O/pnO) are continuous, for all n > 1, with the discrete topology on the target. We denote A(O) with this topology by AFo(O), which makes A(K) = A(O) into a topological abelian group→ and induces the discrete topology on Ator(K), the subgroup of torsion points of A(K) and the p-adic topology on the points of A. Let ψA : A(O) Lie(A)(OF) ⊗OF Ω ∗ denote the Fontaine integral i.e., defined by sending a ∈A(O) to ψA(a)(ω) := a (ω) ∈ Ω where ω ∈ H0(A, Ω1 ). As the derivative map d is not continuousb with respect to the natural p-adic A/OF → Fo topology on O, we see that the map ψA is not continuous for A (O) and the discrete topology on Lie(A)(OF) ⊗OF Ω. In order to get continuity of the Fontaine integral ψA and the extended Fontaine integral ϕA, we will need to define a new topology on which resembles the w-topology from Lemma 2.7. Definition 2.15. The w-topology on A(O) = A(K) is defined to be the coarsest linear topology for which A(O(1)) ⊂ A(O) is open and all the maps A(O(1)) A(O(1)/pnO(1)) are continu- ous with respect to the discrete topology on the target. In other words, this will be the linear n (1) ∼ topology on A(K) such that a base of open neighborhoods of 0 ∈→A(K) is given by A(p O ) = n (1) g ∼ n (1) g > p O , ⊕A = p O , ⊕A , for all n 1 where we denote by ⊕A the formal group ( ) ( ) b law of Aand by ⊕A the group law in A K . Moreover, if x ∈ A O is a point, then the subsets n (1b) b w x ⊕A A p O for n > 1 of A(O) give a base of open neighborhoods of x in A (O). We denotedb A(O) with the w-topology by Aw(O). b Remark 2.16. The w-topology on A(O) induces the w-topology on A(O(1)), and later, we will see GK that when Tp(A) = 0, it induces the discrete topology on Ator(K). b w To conclude this discussion, we will show that the action of GF on A (O) and the Fontaine w integral ψA defined on A (O) are continuous. w Lemma 2.17. The action of GF on A (O) is continuous. Fo Proof. We know that the action of GF on A (O) is continuous. Moreover, the topologies induced Fo w on A(OL)= A(L), for a finite extension L/F in K, by A (O) and by A (O) are the same, therefore the conclusion follows. Lemma 2.18. The Fontaine integral ψ : Aw(O) × H0(A, Ω1 ) Ω defined by (x, ω) x∗(ω) ∈ A A/OF Ω is continuous. Proof. We recall that on Aw(O) we have the w-topology, on H0→A, Ω1 we have→ the p-adic A/OF topology, and on Ω we have the discrete topology. Let x ∈ A(O) and ω ∈ H0 A, Ω1 and denote by η := x∗(ω) ∈ Ω. Let U := ψ−1(η) ⊂ A/OF A w 0 1 A (O) × H A, Ω . We claim U is a neighborhood of (x, ω). For this let Vx := x ⊕A A/OF m (1) A p O for some m > 1 be a neighbourhood of x. Let ω1, . . . , ωg denote an OF-basis of 0 1 n ∗ H A, Ω and let n > 1 be an integer such that p x (ωi) = 0 for i = 1, 2, . . . , g. Then for A/OF 0 1 n ∗ n 0 1 anyb β ∈ H A, Ω , we have p x (β)= 0. Let Vω := ω + p H A, Ω denote the neigh- A/OF A/OF borhood of ω. We claim that Vx × Vω ⊂ U. m (1) n Let (y, γ) ∈ Vx × Vω, i.e. y = x ⊕A z, with z ∈ A p O and γ = ω + p β, with β ∈ H0 A, Ω1 . We have A/OF ∗ n ∗ n ∗ n b ∗ n ∗ ∗ (x ⊕A z) (ω + p β)= x (ω + p β)+ z (ω + p β)= x (ω)+ p x (β)= x (ω)= η, 8 which gives our desired result. Note that the first equality holds because the differential ω + pnβ is invariant. For the second equality, we have, denoting δ := ω + pnβ ∈ H0 A, Ω1 , that A/OF g g ∗ ∗ ∗ z (δ)= z (δ|A) = (z1, z2,..., zg) Fi(X1,..., Xg)dXi = Fi(z1, . . . , zg)dzi = 0 i=1 ! i=1 b X X where A = Spf(OF[[X1, . . . , Xg]]) and Fi(X1, . . . , Xg) ∈ OF[[X1, . . . , Xg]]. The final equality holds because z ∈ A pmO(1) . b Definition 2.19. We define the w-topology on B to be the projective limit topology with the w- b A topology defined on A(O). Using Lemma 2.17 and Lemma 2.18, it is a simple exercise, which we leave to the reader, to show that ϕA : BA Lie(A)(OF) ⊗OF Cp(1) is GF-equivariant and continuous where BA has the w-topology and Lie(A)(OF) ⊗OF Cp(1) is endowed with the p-adic topology. → 2.20. Comparison isomorphisms via p-adic integration. To conclude this section, we discuss the role of Fontaine integration in the Hodge–Tate and de Rham comparison isomorphisms for abelian varieties. Fontaine [Fon82] originally defined this integration map in order to re-prove the Hodge–Tate comparison isomorphism for abelian varieties. More precisely, let A∨ denote the dual abelian variety, which is also defined over F and with good reduction and we denote by A∨ its Néron model over OF. See[BLR90, Theorem 8.4.5] for details. We write
ϕA ⊗ 1Cp : Tp(A) ⊗Zp Cp Lie(A)(F) ⊗F Cp(1) ∨ ∨ ∨ ϕA ⊗ 1Cp : Tp(A ) ⊗Zp Cp Lie(A )(F) ⊗F Cp(1), → which are both surjective maps. By duality, the Weil pairing, and further arguments, we obtain the following isomorphism →
∨ ∼ ∨ Tp(A) ⊗Zp Cp = Lie(A )(F) ⊗F Cp ⊕ (Lie(A)(F) ⊗F Cp(1)) , which is the Hodge–Tate decomposition of Tp(A). We also mention work of Coleman [Col84], which complements [Fon82] and gives another proof of the Hodge–Tate decomposition. In addition to the Hodge–Tate decomposition, there is a de Rham comparison isomorphism for 1 an abelian variety A. More precisely, let HdR(A) denotes the first de Rham cohomology group of the abelian variety A; it has a natural filtration expressed by the exact sequence: 0 1 1 0 H (A, ΩA/F) HdR(A) Lie(A)(F) 0. In [Col92], Colmez defined a p-adic integration pairing, now called Colmez integration, which is a functorial, perfect pairing→ → → → 1 + 2 h·, ·iCz : Tp(A) × HdR(A) BdR/I , which realizes the de Rham comparision isomorphism. Here I ⊂ B+ is the maximal ideal of B+ → dR dR and the above pairing is GF-equivariant in the first argument and respects filtrations in the second 0 1 1 argument. By restricting to H (A, ΩA/F) ⊂ HdR(A), Colmez integration induces a pairing: 0 1 2 ∼ h·, ·iF : Tp(A) × H (A, ΩA/F) I/I = Cp(1), which gives a map Tp(A) Lie(A)(F) ⊗F Cp(1). By[Col92, Proposition 6.1], this map coincides → with Fontaine’s integration (ϕA)|Tp(A), when A has good reduction. → 9 3. The result on the zeroes of the Fontaine integral
In this section, we investigate the kernel of ϕA : BA Lie(A)(F) ⊗F Cp(1). We recall that the abelian variety is defined over F, a finite extension of Qp contained in K and denote IF := GK, the inertia subgroup of GF. As stated in Remark 2.13,→ the Fontaine integral will vanish on an G element u = (un)n>0 ∈ BA if un ∈ OK for all n i.e., if the path u lies in (BA) K . It is natural to speculate whether the Fontaine integral restricted to Tp(A) will be injective away from unramified paths which live in Tp(A). To make this question easier to study, we will impose the following assumption throughout. G Assumption 3.1. We assume that the abelian variety A defined over F satisfies Tp(A) K = 0. Remark 3.2. Assumption 3.1 is not very restrictive. For example, if A is an elliptic curve, Assumption 3.1 is equivalent to the property that A does not have CM by a quadratic imaginary field M, in which p is split (see e.g., [Oze10, Theorem 2.11] and [Ser98, A.2.4] for details). We refer the reader to [Ser98, A.2.3] for further discussion of the relationship between complex multiplication and Assumption 3.1.
We begin our analysis of ker(ϕA) by identifying two of its subgroups. First, we again remark G that (BA) K ⊂ ker(ϕA). Indeed, this follows from the above discussion, or one may deduce it from the fact that GK (Lie(A)(OF) ⊗OF Cp(1)) = 0, which is a consequence of a result of Tate [Tat67, Theorem 2]. We can easily determine another significant subgroup of ker(ϕA).
Definition 3.3. A path u = (un)n>0 ∈ BA is periodic if there exists some k > 1 such that un = un+k for all n > 0. More precisely, a periodic path is of the form
k−1 k−2 k−1 k−1 u := u0, [p ]u0, [p ]u0 . . . , [p]u0, u0, [p ]u0, . . . , [p]u0, u0, [p ]u0 . . . , (3.1)
k where, let us remark that [p ]u0 = u0.
Lemma 3.4. Let u = (un)n>0 ∈ BA be a periodic path. Then u0 is a prime-to-p torsion point of A(O) G and u ∈ (BA) K . k k Proof. Let k > 1 be the smallest integer such that uk = u0. Then u0 = [p ]uk = [p ]u0 which k k implies [p − 1]u0 = 0 and therefore u0 ∈A[p − 1](O). Since A has good reduction at p, one can use formal groups to deduce that u0 ∈A(OK). Now, the latter claim follows from the description of a periodic path in (3.1). Lemma 3.5. Suppose that A satisfies Assumption 3.1, and let u be a prime-to-p torsion point of A. Then, G there is a unique periodic path u = (un)n>0 ∈ (BA) K such that u0 = u. G G Proof. If v := (vn)n>0 ∈ (BA) K is a path such that v0 = u0, then u − v ∈ Tp(A) K = 0 by (2.1) and Assumption 3.1.
GK Definition 3.6. We define per(BA) ⊂ (BA) to be the subgroup of periodic sequences of BA.
Lemma 3.7. The subgroup per(BA) is isomorphic to the subgroup of prime-to-p torsion points on BA which itself is isomorphic to the subgroup of prime-to-p torsion points on A(O).
Proof. We first note that the prime-to-p torsion on BA is isomorphic to the prime-to-p torsion on A(O). Indeed, this follows because the multiplication-by-p map is an isomorphism on A[n](O) for every n coprime to p. The result now follows from Lemma 3.4. With these definitions established, we can now state our main result. 10 Theorem 3.8 (=Theorem A). Let A be an abelian variety over F, with good reduction, and let A denote its Néron model. Suppose that A satisfies Assumption 3.1. Then we have that
GK per(BA) = (BA) = ker(ϕA). Below, we make two remarks relating Theorem 3.8 to previous literature on determining the zeros of p-adic integrals and to p-adic dynamical systems. Remark 3.9. We wish to highlight the similarities between our Theorem 3.8 and [Col85b, Theorem 2.11]. The result of Coleman says that the torsion points on an abelian variety correspond to the set of common zeros of p-adic abelian integrals of the first kind on the abelian variety. Our Theorem 3.8 shows that for an abelian variety A over K satisfying Assumption 3.1, the common zeros of the Fontaine integral are precisely the torsion points on BA. Indeed, recall that Lemma 3.7 shows the subgroup per(BA) coincides with the prime-to-p torsion on BA, and since BA will have no p-power torsion by construction, the statement follows. Remark 3.10. We want to give an interpretation of Theorem 3.8 in terms of p-adic dynamical sys- tems. First, we note that Theorem 3.8 reduces to showing that for u = (un)n>0 ∈ BA a non- periodic path, the valuation of αn from Remark 2.13 is bounded with respect to n for all n ≫ 0. Indeed, such a result will tell us that ϕA(u)(ω) is non-zero and hence not in the kernel. In order to study v(αn), we use the function δ and Lemma 2.4.(6). Recall from Remark 2.13 that we precisely r know the value of δ(dεr/εr) where εr is a primitive p -th root of unity, and so if we can compute ∗ (or more realistically, bound) δ(un(ω)), then we can understand v(αn). ∗ To determine δ(un(ω)), we need to understand various aspects of the field extensions K(un)/K for all n > 0. In particular, we would like to find explicit uniformizers for K(un)/K and to compute the different ideal of the extension K(un)/K. Both aspects are very difficult in general; we refer the reader to [Viv04, BL20] (resp. [Sve04, AHM05, Ber14, Ber16, Nor18]) for a discussion of these topics, respectively. That being said, understanding the ramification behavior of the extensions K(un)/K is a question of interest in the field of p-adic dynamical systems. Our Theorem 3.8 shows that the algebraic extension K∞ = n>0 K(un) is deeply ramified, or equivalently by Proposition 2.6, the values δ(O ∞ ) are unbounded. In fact, it says more. Using K S Remark 2.13, we can see that Theorem 3.8 states that the function |δ(OK(un))| grows like n plus a term which is bounded with respect to n for all n ≫ 0, and unwinding the definition of δ, this says the valuation of the different ideal of K(un)/K is at least n. If A is an elliptic curve and u = (un) ∈ Tp(A), then this type of result can be deduced using formal Weierstrass preparation, workof Katz[Kat73, Section 3.6], and some basics of p-adic dynamics (see e.g., [Ben19]). However, when A has dimensionb > 1, proving such a result using dynamical methods seems beyond reach due to the lack of techniques to analyze p-adic multivariable power series.
4. The proof of Theorem 3.8 In this section, we first describe the reduction of the proof of Theorem 3.8 to a statement about injectivity of the Fontaine integral restricted to the Tate module of the formal group associated to our abelian variety and then prove this injectivity following an idea of P. Colmez. 4.1. Reduction of Theorem 3.8 to the setting of the Tate module of the formal group. Let A/F be an abelian variety with good reduction, A its Néron model over OF and we assume that A satisfies Assumption 3.1. Let A denote the formal completion of A along the identity of its special fiber, i.e. the formal group of A. Then, if we denote by g the dimension of A over Spec(F), we have that b ∼ A = Spf (OF[[X1, . . . , Xg]]) as formal schemes, with a formal group-law given by g power series b A1(X1, . . . , Xg, Y1, . . . , Yg), . . . , Ag(X1, . . . , Xg, Y1, . . . , Yg). 11 b b Let Tp(A) ⊂ Tp(A) be the Tate module of the formal group and let us recall that the restriction to the formal group defines an isomorphism H0(A, Ω1 ) =∼ Inv(A), where the second O -module A/OF F is the moduleb of invariant differentials of the formal group. We also have natural isomorphisms ∼ Lie(A)(OF) = Lie(A)(OF), where the second module is the Lie algebrab of the formal group, i.e. its tangent space at the origin. Let (ϕ ) : Tpb(A) Lie(A)(O ) ⊗ Cp(1) denote the restriction of ϕ to Tp(A). The main A |Tp(A) F OF A result of this subsection proves that if (ϕ ) is injective, then Theorem 3.8 holds. A |Tp(A) b b → b b Theorem 4.2. Assume that A satisfies Assumptionb 3.1 and is such that ker((ϕ ) ) = 0. Then we A |Tp(A) have b (1) The restriction (ϕA)|Tp(A) : Tp(A) Lie(A)(OF) ⊗OF Cp(1) is injective. GK (2) We have ker(ϕA)= per(BA) = (BA) i.e., Theorem 3.8 holds. → Proof of Theorem 4.2.(1). We first prove (ϕA)|Tp(A) is injective. If we denote by A(p) and respec- tively A(p), the p-divisible groups attached to the abelian scheme A and respectively its formal group scheme, then we have an exact sequence of p-divisible groups b 0 A(p) A(p) A(p)/A(p) 0.
As A(p) is identified with the connected sub-p-divisible group of A(p), the p-divisible group → b → → b → A(p)/A(p) is an étale p-divisible group over OK, therefore it is isomorphic, as p-divisible groups, 2g−h to (Qbp/Zp) , where h is the height of A. The exact sequence of p-divisible groups above definesb and exact sequence of GK-representations b ψ 2g−h 0 Tp(A) Tp(A) (Zp) 0, (4.1)
Let e1, e2, . . . , e2g denote a Zp-basis of Tbp(A), such that e1, . . . , eh is a basis of Tp(A) and the images 2g−h → → → → ψ(eh+1), . . . , ψ(e2g) in (Zp) give a Zp-basis of this module. We can describe the action of GK on the basis elements as follows: if σ ∈ GK and 0 6 i 6 2g − h , then b
σ(eh+i)= eh+i + ξi(σ) where ξi : GK Tp(A) is a continuous 1-cocycle, for all 0 6 i 6 2g − h.
Suppose for the sake of contradiction that (ϕA)|Tp(A) is not injective, and let b → 2g−h
0 6= x = f + aieh+i, = Xi 0 6 6 denote an element of ker((ϕA)|Tp(A)) with f ∈ Tp(A) and ai ∈ Zp for all 0 i 2g − h. Let σ ∈ GK and consider the relations: b ( ) ( ) + 2g−h ( ) ( ) = ϕA |Tp(A) f i=0 ai ϕA |Tp(A) eh+i 0 P ( ) ( ) + 2g−h ( ) ( ( )) + 2g−h ( ) ( ) = σ ϕA |Tp(A) f i=0 ai ϕA |Tp(A) ξi σ i=0 ai ϕA |Tp(A) eh+i 0. By substracting these rows,P one obtains P
2g−h
(ϕA)|Tp(A) f − σ(f)− aiξi(σ) = 0. i=0 ! 12 X As (ϕ ) is injective, it follows that A |Tp(A)
b 2g−h aiξi(σ)= f − σ(f), for all σ ∈ GK, = Xi 0 which implies that the cohomology class
2g−h
ai[ξi]= 0 = Xi 0 1 in H (GK, Tp(A)), where we denoted [ξi] the class of the cocycle ξi. The exact sequence (4.1) gives the long exact sequence in cohomology: b GK 2g−h δ 1 0 = Tp(A) (Zp) H (GK, Tp(A)) G where Tp(A) K = 0 by Assumption 3.1 and δ (ψ(eh+i)) = [ξi] for all 0 6 i 6 2g − h. Therefore → → b 2g−h 2g−h
δ(ψ(x)) = δ aiψ(eh+i) = ai[ξi]= 0. − ! = Xi 0 Xi 0
As δ is injective, it follows that ψ(x)= 0, i.e. x = f ∈ Tp(A), and since (ϕA)|Tp(A)(x) = (ϕA)|Tp(A)(f)= 0, we have x = 0. This is a contradiction, which proves our desired result. b Before we prove Theorem 4.2.(2), we need a lemma.
Lemma 4.3. Keep the notations and assumptions from Theorem 4.2. Let u ∈ BA be such that u0 = α(u) ∈ G A(OL), for L/F a finite extension in K. If ϕA(u)= 0, then u ∈ (BA) L , where GL = Gal(K/L).
Proof. First note that for every σ ∈ GL we have that α (σ(u)− u) = 0, i.e. σ(u)− u ∈ Tp(A).
Theorem 4.2.(1) implies that (ϕA)|Tp(A) is injective. Therefore if (ϕA)(u) = 0, then we have
(ϕA)|Tp(A)(u − σ(u)) = ϕA(u)− σ(ϕA(u)) = 0, and hence σ(u)− u = 0 for all σ ∈ GL. Thus, for all σ ∈ GL we have σ(u)= u.
Proof of Theorem 4.2.(2). We already know that per(BA) ⊂ ker(ϕA), so it suffices to prove the re- verse inclusion. We break the proof up into three cases.
Case 1. Let u ∈ ker(ϕA) such that u0 := α(u) ∈ A(O)[m] for m > 1 with (m, p) = 1. Then G G u0 ∈ A(OK) and by Lemma 4.3, we see that u ∈ (BA) K . However, the only element u ∈ (BA) K with u0 ∈A(OK)[m] for m above is the periodic sequence associated to u0, and so u ∈ per(BA).
Case 2. Let us now supposethat u ∈ ker(ϕA) with u0 = α(u) ∈A(O) such that u0 is a torsion point a a a of A of order a power of p, say u0 ∈A(O)[p ] for some a > 1. Then p (u) = ([p ]un)n>0 ∈ Tp(A) a a a and ϕA(p (u)) = p ϕA(u) = 0. As (ϕA)|Tp(A) is injective, we have that p (u) = 0 in BA. Since BA[p]= 0, we see that u = 0.
Case 3. Finally, let us suppose that u ∈ ker(ϕA) with u0 = α(u) not a torsion point of A(O).
Observe that u ⊗ 1Qp ∈ BA ⊗Z Q whose image via α ⊗ 1Q in A(O) ⊗Z Q is u0 ⊗ 1 6= 0. We have a natural exact sequence of Qp-vector spaces, with continuous action by GK, namely
0 Vp(A) BA ⊗Z Q A(O) ⊗Z Q 0. (4.1) ∼ Notice that as Tp(A) is naturally a Zp-module, we have that Vp(A) := Tp(A) ⊗Z Q = Tp(A) ⊗Zp Qp. → → 13 → → Let us now suppose that u ∈ ker(ϕA) has the property that u0 = α(u) ∈A(OL) for some finite extension L of F, in K and that, as stated above, 0 6= u0 ⊗ 1 ∈ A(O) ⊗Z Q. We denote, as usual GL = Gal(K/L) and let αL : BA,L A(OL) be the fiber product of the diagram of GL-modules: α BA − A(O) → ∪ → A(OL) In other words we have a cartesian, commutative diagram of GL-modules, with exact rows α⊗1 0 Vp(A) BA ⊗Z Q A(O) ⊗Z Q 0 || ∪ ∪ αL⊗1 0 → Vp(A) → BA,L ⊗Z Q → A(OL) ⊗Z Q → 0 It is easy to see that we have, on the one hand → → → → GL (BA) ⊂ BA,L ⊂ BA,
GL GL therefore we have BA,L ⊗Z Q = BA ⊗Z Q . On the other hand the exact sequence