A note on fierce ramification
H\'el\`ene Esnault, Lars Kindler, Vasudevan Srinivas

TL;DR
This paper demonstrates that bounding ramification at infinity implies fierce ramification, providing a positive answer to a question posed by Deligne regarding ramification behavior.
Contribution
It establishes a link between bounded ramification at infinity and fierce ramification, resolving a question in ramification theory.
Findings
Bounding ramification at infinity implies fierce ramification.
Provides a positive answer to Deligne's question.
Advances understanding of ramification behavior at infinity.
Abstract
We show that bounding ramification at infinity bounds fierce ramification. This answers positively a question of Deligne posed to the first named author.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMathematics and Applications · Advanced Differential Equations and Dynamical Systems · Geometric Analysis and Curvature Flows
A note on fierce ramification
Hélène Esnault
Freie Universität Berlin, Arnimallee 3, 14195, Berlin, Germany
,
Lars Kindler
and
Vasudevan Srinivas
School of Mathematics, Tata Institute of Fundamental Research, Homi Bhabha Road, Colaba, Mumbai-400005, India
1. Introduction
Let be an algebraically closed field of characteristic and let be a prime different from . One fixes an algebraic closure of . Let be a nonsingular connected projective curve over , with a dense Zariski open subset , and a geometric point . To any continuous representation , thus with values for some non-zero natural number , one associates its Swan conductor in the group is zero-cycles. It is an effective divisor supported on , which measures the wild ramification of ([Ser67, III.20], [Lau81, 1.1.2], [KR14, 4.4].)
The definition of the Swan conductor is local: if is the function field of , then for each closed point , consider the corresponding complete discrete valuation field obtained from by completion at , and the associated local Galois representation into (well-defined up to conjugation). The non-negative integer invariant assigned to it, see e.g. [KR14, 4.82,4.84], is the coefficient of in the zero cycle . We will use the term “Swan conductor” to mean either the local or global invariant, depending on the context.
When is a normal connected variety of finite type over , a modulus condition for continuous representations is defined in [EK12, Defn. 3.6]: if is a normal compactification of , and is a Cartier divisor supported in , we say that has ramification bounded by , if for any morphism from a connected nonsingular projective curve , such that is nonempty, the induced representation verifies
[TABLE]
with respect to the order on . As is normal, the intersection of its smooth locus with is dense, so as a divisor is a sum , where is an irreducible divisor and are the called the multiplicities of . Let be a natural number. We say that * has ramification bounded by if it has ramification bounded by for an effective divisor supported in with multiplicities for all . * Similarly we say that * has ramification bounded by along a divisorial discrete valuation * if there exists some normal compactification as above, with a boundary component divisor corresponding to , such that has ramification bounded by an effective Cartier divisor , where the coefficient of in is .
A natural question is whether, for a fixed effective Cartier divisor as above, or for a chosen divisorial valuation , the class of representations with ramification bounded by (or with ramification bounded by along ) has other “finiteness properties” with respect to wild ramification.
One such is the notion of fierce ramification along an irreducible component of . Let be the Galois cover of Galois group determined by the quotient , and be the normalization of in the field of functions of . So the smooth locus has complement of codimension at least . Let be an irreducible component of . Then the fierce ramification index of is the purely inseparable degree of the function field extension . It depends on the local system defined by and , not on the choices of and . Indeed, acts transitively on the set of components of , on the set of extensions of preserving the separable closures and the purely inseparable ones. Changing conjugates the representation. The conjugation sends defined for to the corresponding set defined for the other base point, preserving the separable closures and the purely inseparable ones. We say that has fierce ramification bounded by a natural number if for all , the fierce ramification index of is at most .
Similarly we have the notion of fierce ramification index along a divisorial valuation , which equals the fierce ramification index along for any normal compactification with a boundary component associated to the divisorial valuation . This notion depends only on the discrete valuation, since it can be defined using the extension of discrete valuation rings associated to .
The aim of this note is to prove that bounding the ramification along a divisorial valuation also bounds the fierce ramification index.
Theorem 1**.**
Let be as above. Let be a natural number. Then there is a natural number such that for all continuous representations of ramification bounded by along , the fierce ramification of along is bounded by .
The theorem answer positively a question posed by Pierre Deligne in [Del16].
The proof consists of two parts. In Section 2, we first make a local analysis of ramification at a point on a nonsingular curve, in relation to a bound on the Swan conductor at that point. This is formulated in terms of a boundedness assertion for representations of the corresponding local Galois group (see Proposition 2). To globalize the argument and perform the proof of Theorem 1, we notably use a version of the Bertini theorem controlling that the inverse image of a curve by a ramified Galois cover remains unibranch [Zha95].
Acknowledgements: The first author thanks Pierre Deligne for his letter [Del16] sent in relation to [Esn17]. This note, which answers one of the two questions in his letter, has been circulating among experts for two years. The second question, asking whether fixing algebraically closed, and one can find a curve such that the restriction of any irreducible representation of rank and ramification bounded by remains irreducible, remains unanswered. By Deligne’s theorem [EK12, Thm. 1.1] and the standard Lefschetz theorem over finite fields (see [Esnl17] and references in there), this is true over for the representations which descend to for a fix natural number .
2. Local arguments
Let be an algebraically closed field of characteristic and let be a discretely valued, complete field of characteristic with residue field . Fix algebraic an algebraic closure and write .
Proposition 2**.**
Fix a prime , and positive natural numbers . There exists a number with the following property. For any continuous representation with bounded by , there exists a finite Galois extension of of degree , such that is tamely ramified.
Proof.
Given as in the statement, we write , and we let denote the (unique) -Sylow subgroup of . The Galois theory of discretely valued fields shows that there is a short exact sequence
[TABLE]
where . Writing with , we have .
Claim 1**.**
The exponent is bounded by a constant only depending on .
Proof.
In there exists an element of order precisely (e.g. by the theorem of Schur-Zassenhaus). Considering as an element of , we can write , with a nilpotent matrix such that but . This shows that . ∎
Claim 2**.**
The order of is bounded by a constant only depending on .
Proof.
Let be the preimage of . We proceed in several steps.
- (a)
As , and is the unique -Sylow subgroup of . 2. (b)
As , we see that . Recall Jordan’s theorem according to which there is a constant , the Jordan constant, depending only on , such that, given a subgroup of order prime to , it contains an abelian normal subgroup of index at most (the constant is the same as for subgroups of ). It yields a normal abelian subgroup , such that is of order bounded by . Consider the following diagram:
[TABLE]
To prove the claim, it suffices to show that the order of is bounded by a constant only depending on . 3. (c)
Translating the group theoretic picture via Galois theory, we obtain the following diagram of field extensions, where the arrows are labeled by the corresponding Galois groups.
[TABLE] 4. (d)
If is the induced representation, then there exists a bound depending only on , such that is bounded by .
To see this, apply 5 below to and . This yields the desired estimate with . 5. (e)
For every element , we have . Indeed, if are the jumps of the ramification filtration of in ascending order, and if , then
[TABLE]
where is the -dimensional -vector space on which acts. It follows that , as the jumps are integers, according to the theorem of Hasse-Arf. But the associated gradeds of the filtration are products of copies of . It follows that every element of is killed by . 6. (f)
Finally, as , the elements of the finite abelian subgroup are simultaneously diagonizable matrices with eigenvalues -th roots of unity. It follows that the cardinality of is bounded by , which is a constant depending only on and , as claimed.
∎
Remark 3**.**
Our argument yields the upper bound for the order of . Since (since ), we get an upper bound for the order of to be .
We finish the proof of Proposition 2. The short exact sequence (1) induces a homomorphism of groups , given by conjugation. Write and let be the preimage of . By construction , with both and characteristic subgroups of . We obtain an injective map , such that the composition is the identity on , and is normal in . This shows that there is a short exact sequence
[TABLE]
The orders of the two outer terms are bounded by constants depending only on (because of Claim 2 and because ). It follows that the order of is bounded by such a constant.
Let be the Galois extension of corresponding to . Then the restriction of to is tame, as it takes values in the prime-to--group . On the other hand, by construction we have . This finishes the proof.
∎
Remark 4**.**
Our argument gives an upper bound for the index of the subgroup to be the order of the automorphism group of . Thus, if is our bound for the order of , a crude upper bound for is the order of the permutation group, that is,
Lemma 5**.**
Let be an extension of finite Galois extensions of complete discretely valued fields with algebraically closed residue fields of characteristic . If is a continuous representation with bounded by , then is bounded by .
Proof.
Write and . Then there is a finite extension , such that and . Since the rank is fixed, and since the Swan conductor is additive, we may assume that is irreducible. In this case, the existence of the “break decomposition” implies that there exists a unique such that . As , is also the only index where the ramification filtration of jumps (we assume ). One has
[TABLE]
and
[TABLE]
where denotes the Herbrand function. We know that
[TABLE]
and for any , one has
[TABLE]
It follows that
[TABLE]
as we wanted to prove. ∎
3. Global arguments
Let be an algebraically closed field, a normal projective -variety and let be an open subscheme such that is divisor. Let be an irreducible component of , with generic point . This corresponds to a divisorial valuation with discrete valuation ring . We fix , a continuous representation with finite image and with ramification bounded by an effective Cartier divisor which has multiplicty at most along .
We then introduce some notations.
- •
Let be finite Galois covering with group and let be the normalization of in . Denote by the codimension points of lying over .
- •
As is Galois, the ramification index of the extension of discrete valuation rings is independent of ; we denote it by .
- •
Let be the degree of the residue extensions , and the separable, resp. inseparable degree. Note that is independent of , and so are , : indeed, the integral closure of in is a Dedekind domain and if is a prime ideal lying over the maximal ideal , then any Galois automorphism induces an isomorphism of -extensions , and acts transitively on the set of prime ideals of lying over .
Proposition 6**.**
With as above, a general complete intersection curve has the following properties:
- (a)
* is smooth and intersects transversely at smooth points.* 2. (b)
Write , , then is Galois étale with group . 3. (c)
Write . The normalization map is a universal homeomorphism. 4. (d)
For , there are precisely points in (and hence in ) mapping to .
Proof.
That is nonsingular projective and intersects only at non-singular points, and that the intersection is transverse, follows easily from Bertini’s theorem. This proves (a). For (c), we use [Zha95, Thm. 1.2] where it is shown that the inverse image scheme of a general complete intersection curve is geometrically unibranch at all points. In addition, it is connected, since is normal (see the proof of [Zha95, Cor. 1.7], for example). Since is Cohen-Macaulay outside a closed subset of codimension , we may assume is contained in the Cohen-Macaulay locus, and so (which is a complete intersection) is also Cohen-Macaulay, and hence reduced. Thus is irreducible, and the normalization morphism is bijective, and a universal homeomorphism. This proves (b). For (d), note that for every , the finite map
[TABLE]
has degree , and can be factored in a homeomorphism of degree followed by a generically étale morphism of degree . We may now assume is chosen so that its intersection with is in the open subset over which the morphism is étale, which again follows from Bertini’s theorem. ∎
Definition 7**.**
A curve as in Proposition 6 is said to be in good position relative to . If is in good position relative to each irreducible component of , we say that is in good position relative to .
Theorem 8**.**
We keep the assumptions and notations from above. Let be a curve which is in good position relative to . Fix a closed point and a closed point mapping to . Denote by be the associated decomposition group. Then
[TABLE]
Proof.
Let be the normalization morphism. We have the following diagram
[TABLE]
and we utilize the notations introduced at the beginning of the section. Note that acts on and that the normalization is an -equivariant homeomorphism.
For , the properties from 6 imply that . On the other hand, for mapping to we have
[TABLE]
and
[TABLE]
It follows that
[TABLE]
which is what we wanted to prove. ∎
Recall that a by [KS10, Thm. 1.1], a Galois cover is tamely ramified if and only if it is in restriction to all curves. We make the small observation, implicit in loc.cit., that there are test curves for all the generic codimension points at infinity. Explicitly:
Corollary 9**.**
The covering is tamely ramified along the codimension one points of with support in if and only if is tamely ramified for a curve which is in good position relative to .
Proof.
If is tamely ramified, then for all , and all mapping to , the order of is prime to . The theorem implies that, for each boundary component , and , so is tamely ramified with respect . ∎
Proof of Theorem 1.
Let be a curve as in 6, i.e. which is in good position relative to . Fix , and mapping to . As intersects transversely, the ramification of is bounded by .
Claim 2 in the proof of 2 showed that the order of the -Sylow subgroup of is bounded by a constant only depending on (an explicit bound is given by the constant ). As is a -power, 8 implies that is bounded by the same constant. ∎
4. Some comments
Using the notation from above, for any , and any lying on mapping to , we have
[TABLE]
The order of is , the order of is , so the kernel has order . It follows that ( ‣ 4) is an equality, and hence that the decomposition groups only depend on the component on which lies (recall that are the generic points of ).
We can also interpret as the inertia group of . As such, it has a unique -Sylow subgroup, which according to the discussion above, has to coincide with the -Sylow subgroup of for any lying on the closure of .
Then this raises the following interesting side question. On the -Sylow subgroup of we have two filtrations: the upper numbering ramification filtration associated to , for some lying on , and the Abbes-Saito ramification filtration on the inertia group of . Do they coincide?
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[Zha 95] Zhang, Bin, Théorèmes du type Bertini en caractéristique positive , Arch. Math. (Basel) 64 (1995), no. 3, 209–215.
- 2[Del 16] Deligne, P.: Email to Hélène Esnault , dated Feb. 29th, 2016.
- 3[EK 12] H. Esnault, M. Kerz, A finiteness theorem for Galois representations of function fields over finite fields (after Deligne) , Acta Mathematica Vietnamica 37 4 (2012), 531–562.
- 4[Esn 17] Esnault, H.: A remark on Deligne’s finiteness theorem , Int. Math. Res. Not. 16 (2017), 4962–4970.
- 5[Esnl 17] Esnault, H.: Survey on some aspects of Lefschetz theorems in algebraic geometry , Revista Matemática Complutense 30 (2) (2017), 217–232.
- 6[KS 10] Kerz, M., Schmidt, A.: On different notions of tameness in arithmetic geometry , Math. Ann. 346 (3) (2010), 641- 668.
- 7[KR 14] Kindler, L., Rülling, K.: Introductory course on ℓ ℓ \ell -adic sheaves and the ramification theory on curves , https://arxiv.org/abs/1409.6899
- 8[Lau 81] Laumon, G.: Semi-continuité du conducteur de Swan (d’après P. Deligne) , in The Euler-Poincaré characteristic, Astérisque, 83 (1981), 173–219, Soc. Math. France.
