ebook img

Ramification filtration of the Galois group of a local field via deformations PDF

0.32 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Ramification filtration of the Galois group of a local field via deformations

RAMIFICATION FILTRATION OF THE GALOIS GROUP OF A LOCAL FIELD VIA DEFORMATIONS VICTOR ABRASHKIN 7 Abstract. Let K be a field of formal Laurent series with co- 1 efficients in a finite field of chracteristic p, G — the maximal <p 0 quotient of Gal(K /K) of period p and nilpotent class < p and sep 2 {G<(vp)}v>0 — its filtration by ramification subgroups in the up- r a per numbering. Let G<p = G(L) be the identification of nilpo- M tent Artin-Schreier theory: here G(L) is the group obtained from a suitable profinite Lie F -algebra L via the Campbell-Hausdorff p 7 composition law. We give a new proof of the construction of gen- 2 eratorsofthe ideals L(v) suchthat G(L(v))=G(v) from[1]. Given <p ] v0 > 0 we construct epimorphism of Lie algebras η¯† : L −→ L¯† T and the action Ω of the formal group α = Spec F [U], Up = 0, U p p N of order p on L¯†. Then dΩ = B†U, where B† ∈ DiffL¯†, and U . Lv0) = (η¯†)−1L¯†[v0], where L¯†[v0] is the ideal of L¯† generated by h t the elements of B†(L¯†). a m [ 2 Introduction v 7 Everywhere in the paper v is a fixed positive real number. 0 0 2 Suppose s ∈ N. For any topological group G, we denote by C (G) s 2 theclosureofthesubgroupofG generatedbythecommutators oforder 0 . > s. If L is a toplogical Lie algebra then C (L) is the closure of the 1 s 0 ideal generated by commutators of degree > s. For any topological 7 A-modules M and B we use the notation M = M⊗ˆ B. B A 1 If L is any Lie algebra over F of nilpotent class < p (i.e. C (L) = 0) : p p v then we use the notation G(L) for the p-group obtained from the set i X L via the Campbell-Hausdorff composition law (l ,l ) 7→ l ◦ l , cf. 1 2 1 2 r Subsection 1.2. a Let k = F . Consider a free pro-finite Lie algebra L over k with pN0 k the set of free generators {D | a ∈ Z+(p),n ∈ Z/N }∪{D }, where an 0 0 Z+(p) is the set of all natural numbers prime to p. e If σ is the absolute Frobenius automorphism of k (for any α ∈ k, σ(α) = αp) we agree to use the same symbol σ for its continuation to an automorphism of L such that σ(D ) = D and σD = D . k an a,n+1 0 0 Then L := L | is a free Lie algebra over F . Clearly, L⊗ k = L . k σ=id p Fp k e Datee: Marceh 27, 2017. e e 2010 Mathematics Subject Classification. 11S15, 11S20. Key words and phrases. local field, Galois group, ramification filtration. 1 2 VICTOR ABRASHKIN Set L = L/C (L). Note that the images of D and D in L (which p an 0 k will be denoted by the same symbols) give a fixed system of generators of L . Wee alsoeset for n ∈ Z/N , D = σn(α )D , where α is a k 0 0n 0 0 0 fixed element from k such that Tr α = 1. From time-to-time we k/Fp 0 allow in the definition of D that n ∈ Z under the agreement that an D = D . an a,nmodN0 Let A be the enveloping algebra of L with augmentation ideal J. Consider an element e ∈ L such that in A K K exp(e) ≡ 1+ η(a ,...,a )t−(a1+···+as)D ...D modJp 1 s a10 as0 K 1X6s<pa1X,...,as (hgere exp is the truncated exponential) where the “structural con- stants” η(a ....,a ) ∈ k are such that exp(e) is “diagonal modulo 1 s degree p” , cf. Subsection 1.3. Fix f ∈ G(L ) such that σf = e◦f g Ksep and consider the map η0 from G := Gal(K /K) to G(L) such that e gsep for any τ ∈ G, η0 : τ 7→ (−f)◦τ(f). This map (due to the nilpotent e version of the Artin-Schreier theory, cf. [2]), provides us with the in- duced identification η : G ≃ G(L), where G := G/GpC (G) is the e <p <p p maximal quotient of G of period p and nilpotent class < p. Forv > 0,letG(v) bethesubgroupoframificationwithuppernumber v. Denote by G(v) the image of G(v) in G . Then η (G(v)) = L(v) is an <p <p e <p ideal in L. In this paper we give a substantially new proof of the main result from [1]. This result can be stated as follows. For α > 0 and N ∈ N, introduce F0 ∈ L such that α,−N k F0 = a η(a¯,n¯)[...[D ,D ],...,D ]. α,−N 1 a1n1 a2n2 asns 16s<p aXi,ni Here: — a¯ = (a ,...,a ) and n¯ = (n ,...,n ) with n = 0; 1 s 1 s 1 — α = α(a¯,n¯) = a pn1 +a pn2 +···+a pns; 1 2 s — if 1 6 s < s < ··· < s = s and n = ··· = n > ··· > 1 2 r 1 s1 n = ··· = n > −N then sr−1+1 sr η(a¯,n¯) := σns1η(a1,...,as1)...σnsrη(asr−1+1,...,asr), otherwise, η(a¯,n¯) = 0. Theorem 0.1. For any v > 0, there is N(v) such that if N > N(v) is fixed then the ideal L(v) is the minimal ideal in L such that its extension of scalars L(v) contains all F0 with αe> v. e k α,−N Our method can be briefly described as follows. First, we introduce a decreasing central filtration by the ideals L = L(1) ⊃ ··· ⊃ L(s) ⊃ ... andprove that L(p) ⊂ L(v0). This gives us the RAMIFICATION VIA DEFORMATIONS 3 projection η : G −→ G(L¯) := G(L/L(p)) and it remains to describe e¯ <p the ideal L¯(v0) of L¯ such that G(L¯(v0)) = η (G(v0)). e¯ <p Then we: a) introduce a lift V : L¯† −→ L¯ where L¯† is a Lie algebra over F of p nilpotent class < p with a central filtration L¯†(s) such that V(L¯†(s)) = L¯(s) and L¯†(p) = 0; b) prove the existence of a group epimorphism η : G −→ G(L¯†) e¯† <p such that Vη = η ; e¯† e¯ c) introduce an action of the group Z/p on L¯†; d) prove that L¯(v0) is the minimal ideal in L¯ such that for any γ ∈ Z/p, L¯(v0) ⊃ V(γ(KerV)) and the induced action of Z/p on L¯/L¯(v0) is trivial (Note that this condition is very difficult to study because it involves very complicated group composition law coming from the Campbell-Hausdorff formula.); e) the above action of the group Z/p comes from the co-action Ω : L¯† −→ L¯† ⊗ F [U] of the group scheme α = F [U] such that U p p p Up = 0 and the coaddition is such that ∆U = U ⊗1+1⊗U; f) if dΩ = B†U is the differential of Ω , where B† ∈ DiffL¯†, then U U the conditions from above item d) mean that VB†(L¯†) ⊂ L¯(v0); g) the above linearization procedure allows us to find explicit expres- sions for the generating elements F0 from Theorem 0.1. α,−N There is the following non-formal aspect of our construction. We could have started with the Lie algebra L¯ = L/L(p) and the ¯ ¯ image e¯= e¯(t) ∈ L of e ∈ L to fix the projection η : G −→ G(L) K K e¯ <p via methods of nilpotent Artin-Schreier theory. In order to define a deformation of η we choose auxiliary parameters N∗ = N∗(v ) ≡ e¯ 0 0modN and b∗ ∈ Z+(p) such that r∗(v ) := b∗/(q − 1) is “close” to 0 0 v (here q = pN∗). If we apply σN∗ to e¯ we obtain σN∗e¯= e¯(tq) and η 0 e¯ will be not changed. On the other hand, we can use a “deformation” of σN∗ such that t 7→ tqexp(tb∗). This operation gives us an element e¯′ ∈ L¯ and the nilpotent Artin-Schreier theory provides us with a new K projection η : G −→ G(L¯). The initial expectation is that (after e¯′ <p g a careful choice of the parameters N∗ = N∗(v ) and r∗ = r∗(v )) the 0 0 ramification ideal L¯(v0) will appear as the minimal ideal in L¯ such that ηe¯ coincides with its “deformation” ηe¯′ modulo L¯(v0). When following this idea we introduced a lift V : L¯† −→ L¯ and defined a deformation of σN∗ as an automorphism of some subquotient of L¯† . An interesting K aspect of this construction is that this deformation induces a group automorphism of the quotient G(L¯†) of G which is not induced by an <p automorphism of K. As a matter of fact, it comes in the form exp(B†) where B† is a differentiation of L¯† induced by a higher derivation taq 7→ ataq−b∗ (on the above subquotient of L¯† ). Finally, we prove that the K g 4 VICTOR ABRASHKIN ramificationideal L¯(v0) isgeneratedbytheelements ofV(B†(KerV)), or equivalently, L¯(v0) is the minimal ideal in L¯ such that the deformation B† descends to a trivial deformation of L¯/L¯(v0). The methods from this paper admit a generalisation to the “modulo pM” situation as well as to the case of higher dimensional local fields in the characteristic p case (in preparation). They also provide us with much better background for the paper [8] and its upcoming generalisa- tions to the “modulo pM” situation and the case of higher dimensional local fields in the mixed characteristic case. 1. Identification of nilpotent Artin-Schreier theory Suppose K is a field of characteristic p, K is a separable closure of sep K and G = Gal(K /K). We assume that G acts on K as follows: sep sep if g ,g ∈ G and a ∈ K then g (g a) = (g g )a. Denote by σ the 1 2 sep 1 2 1 2 p-power morphism. In [1, 2] we developed a nilpotent analogue of the classical Artin- Schreier theory of cyclic extensions of fields of characteristic p. We are going to use the covariant analogue of this theory, cf. the discussion in [7], for explicit description of the group G = G/GpC (G) as follows. <p p 1.1. Lie algebra L. Suppose K = k((t)) where t is a fixed uniformizer and k ≃ F with N ∈ N. Fix α ∈ k such that Tr (α ) = 1. pN0 0 0 k/Fp 0 Let Z+(p) = {a ∈ N | (a,p) = 1} and Z0(p) = Z+(p)∪{0}. LetLbeaprofinitefreeLieF -algebrawiththe(topological)module p of generators K∗/K∗p and L = L/C (L). Then we can take as a system p of topoelogical generators of L the set k e e {D }∪{D | a ∈ Z+(p),n ∈ Z/N } 0 an 0 due to the following identifications: (K∗/K∗p)⊗ˆ k = Hom (K/(σ −id)K,k) = Fp Fp Hom (⊕ kt−a ⊕F α ,k) = Hom (kt−a,k)×kD Fp a∈Z+(p) p 0 Fp 0 a∈Z+(p) Y and Hom (kt−a,k) = kD , where for any α ∈ k and a,b ∈ Fp n∈Z/N0 an Z+(p), D (αt−b) = δ σn(α). Note also that the first identification an aQb uses the Witt pairing [9, 6] and D comes from t⊗1 ∈ (K∗/K∗p)⊗ˆ k. 0 Fp For any n ∈ Z/N , set D = t⊗(σnα ) = (σnα )D . 0 0n 0 0 0 1.2. Groups and Lie algebras of nilpotent class < p. The basic ingredient of the nilpotent Artin-Schreier theory is the equivalence of the category of p-groups of nilpotent class s < p and the category of 0 Lie Z -algebras of the same nilpotent class s . In the case of objects p 0 killed by p, this equivalence can be explained as follows. Let L be a Lie F -algebra of nilpotent class < p, i.e. C (L) = 0. p p RAMIFICATION VIA DEFORMATIONS 5 Let A be the enveloping algebra of L. Then there is a natural em- bedding L ⊂ A, the elements of L generate the augmentation ideal J of A and we have a morphism of algebras ∆ : A −→ A⊗A uniquely determined by the condition ∆(l) = l⊗1+1⊗l for all l ∈ L (in other words all l ∈ L are primitive). Applying the Poincare-Birkhoff-Witt Theorem we obtain that: — L∩Jp = 0; —LmodJp = {amodJp | ∆(a) ≡ a⊗1+1⊗amod(J⊗1+1⊗J)p}; — the set exp(L)modJp is identified with the set of all ”diagonal elements modulo degree p“, i.e. with the set of a ∈ 1+J modJp such that ∆(a) ≡ a⊗amod(J ⊗1+1⊗J)p (here exp(x) = xi/i! is g 06i<p the truncated exponential). P g In particular, there is a natural embedding L ⊂ A/Jp and in terms of this embedding the Campbell-Hausdorff formula appears as 1 (l ,l ) 7→ l ◦l = l +l + [l ,l ]+..., l ,l ∈ L, 1 2 1 2 1 2 1 2 1 2 2 where exp(l )exp(l ) ≡ exp(l ◦l )modJp. This composition law pro- 1 2 1 2 vides the set L with a group structure and we denote this group by G(L). Note that I is an ideal in L iff G(I) is a normal subgroup g g g in G(L). Clearly, G(L) has period p and nilpotent class < p. Then the correspondence L 7→ G(L) is the above mentioned equivalence of the categories of p-groups of period p and nilpotent class s < p and Lie F -algebras of the same nilpotent class s. This equivalence can p be naturally extended to the categories of pro-finite Lie algebras and pro-finite p-groups. 1.3. Epimorphism η0 : G −→ G(L). Let L := L . Then the e sep Ksep elements of G = Gal(K /K) and σ act on L through the second sep sep factor, L | = L and (L )G = L . The covariant nilpotent Artin- sep σ=id sep K Schreier theory states that for any e ∈ G(L ), the set K F(e) = {f ∈ G(L ) | σ(f) = e◦f} sep is not empty and for any fixed f ∈ F(e), the map τ 7→ (−f)◦τ(f) is a continuous group homomorphism π (e) : G −→ G(L). The correspon- f dence e 7→ π (e) has the following properties: f a) if f′ ∈ F(e) then f′ = f ◦l, where l ∈ G(L), and π (e) and π (e) f f′ are conjugated via l; b) for any continuous group homomorphism π : G −→ G(L), there are e ∈ G(L ) and f ∈ F(e) such that π (e) = π; K f c) for appropriate elements e,e′ ∈ G(L ) and f,f′ ∈ G(L ), we K sep have π (e) = π (e′) iff there is an x ∈ G(L ) such that f′ = x◦f and, f f′ K therefore, e′ = σ(x)◦e◦(−x). 6 VICTOR ABRASHKIN In[1,2,3]weappliedthistheorytotheLiealgebraLfromSubsection 1.1 via a special choice of e ∈ L such that K (1.1) expe ≡ 1+ η(a ,...,a )t−(a1+···+as)D ...D modJp . 1 s a10 as0 L,K 16s<p X Here Jgis the augmentation ideal in the envelopping algebra A of L. L L In the above sum the indices a ,...,a run over Z0(p) and the “struc- 1 s tural constants” η(a ,...,a ) ∈ F satisfy the following identities: 1 s p I ) η(a ) = 1; e 1 II ) if 0 6 s 6 s < p then e 1 η(a ,...,a )η(a ,...,a ) = η(a ,...,a ), 1 s1 s1+1 s π(1) π(s) πX∈Is1s where I consists of all permutations π of order s such that the se- s1s quences π−1(1),...,π−1(s ) and π−1(s +1),...,π−1(s) are increasing 1 1 (i.e. I is the set of all “insertions” of the ordered set {1,...,s } into s1s 1 the ordered set {s +1,...,s}). 1 Under the assumption I ) the map π (e)modGpC (G) induces a e f 2 group isomorphism of Gab⊗ˆF and G(L)/C (G(L)) = Lab = K∗/K∗p, p 2 which coincides with the inverse to the reciprocity map of local class field theory, cf. [6]. This also implies that η0 := π (e) (when taken e f modulo GpC (G)) induces a group isomorphism η : G ≃ G(L). In p e <p particular, we can work with the elements of L = “G ⊗ˆk” in terms k <p of our explicit generators {D | a ∈ Z+(p),n ∈ Z/N }∪{D }. an 0 0 Assumption II ) means that e ∆(exp(e)) ≡ exp(e)⊗exp(e)mod(J ⊗ˆ1+1⊗ˆJ )p, L,K L,K i.e. exp(e) is diagonal modulo Jp . L,K g g g Under a fixed choice of the structural constants η(a ,...,a ) the 1 s element e depends on the uniformiser t ∈ K. We underline this fact g in some cases by writing e = e(t). Note that in most examples or interpretations of results from [1, 2, 3] we used the identification G ≃ <p G(L) coming from the simplest choice e = t−aD (in this a∈Z0(p) a0 case all η(a ,...,a ) = 1/s!). But as a matter of fact, all proofs in the 1 s P papers [1, 2, 3] were adjusted to the above more general choice of e. ¯ 1.4. Lie algebra L and epimorphism η . Introduce a weight func- e¯ tion wt : L −→ N on L by setting on its generators wt(D ) = s if k k an (s−1)v 6 a < sv . Then we obtain a central decreasing filtration of 0 0 ideals L(s) = {l ∈ L | wt(l) > s}. This weight function gives us also a decreasing filtration of ideals J (s) in A such that J (1) = J and for L L L L any s, (J (s)+Jp)∩L = L(s) (use the Poincare-Birkhoff theorem). L L Let L¯ := L/L(p) and consider the natural projection p¯r : L −→ L¯. ThenL¯(s) = p¯r(L(s))isadecreasing filtrationinL¯such thatL¯(p) = 0. RAMIFICATION VIA DEFORMATIONS 7 Consider an F -submodule N in L¯ generated by all t−bl where l ∈ p K L¯(s) and b < sv . Then N has a natural structure of a Lie algebra k 0 over F . For any i > 0, let N(i) be the F -submodule in L¯ generated p p K by all t−bl where l ∈ L¯(s) and b < (s−i)v . Then N(i) is an ideal in k 0 N and we can introduce the Lie F -algebra N¯ = N/N(p). Note that p N(p) ⊂ L¯ , where m = tk[[t]]. m Lemma 1.1. e := p¯r e ∈ N. L¯ K Proof. Let N be an F -submodule in L generated by the elements of L p K L(p) and all t−bl such that l ∈ L(s) , b < sv and 1 6 s < p. It will K k 0 be sufficient to prove that e ∈ N . L Let N be an F -submodule in A generated by the elements of AL p K J (p) and all t−bj such that j ∈ J (s) and b < sv . It will be L K L k 0 sufficient to prove that expe ∈ N . Indeed, N is closed under AL AL multiplication and, therefore, log(expe) ∈ N (here log is the trun- AL cated logarithm). It remagins to notice that e ≡ log(expe)modJp and K NAL ∩LK = NL. f g f Verify that any term of the form t−(a1+···+ar)Daf10.g..Dar0 belongs to N . Indeed, if wt(D ) = s for i = 1,...,r, then a + ··· + a < AL ai0 i 1 r (s + ...s )v = wt(D ...D )v . So, expe ∈ N and the lemma 1 r 0 a10 ar0 0 AL is proved. (cid:3) g Let e¯:= e modN(p) ∈ N¯. L¯ Let η := p¯r·η : G −→ G(L¯). Verify that η depends only on the e¯ e <p e¯ image e¯ of e in N¯. Proposition 1.2. Let e′ ∈ L and e′ := (p¯r )e′ ≡ e modN(p). K L¯ K L¯ Then there is a unique f′ ∈ L modL(p) such that σf′ = e′ ◦ f′ sep sep and p¯r·π (f′) = η . e′ e¯ Proof. Note that σ is topologically nilpotent on N(p)modL(p) . This K implies the existence of a unique x ∈ N(p)modL(p) such that e′ = K L¯ (σx)◦e ◦(−x). Let f = (p¯r )f ∈ L¯ . Then σ(x◦f ) = e′ ◦(x◦f ). L¯ L¯ sep sep L¯ L¯ L¯ If f′ ∈ L is such that σf′ = e′ ◦ f′ and f′ := p¯r f′ then for sep L¯ sep g = (−f′)◦x◦f ∈ L¯ we have that σg ≡ gmodL¯ . This implies L¯ L¯ sep sep that g = l modL(p) , where l ∈ L. Therefore, we can replace f′ by 0 sep 0 f′ ◦l to satisfy the requirements of our proposition. From the above 0 construction it is clear that f′ is unique modulo L(p) . (cid:3) sep Remark. From now on we can use the construction of the morphism ¯ ¯ η : G −→ G(L) in terms related only to the Lie algebra N. Namely, e¯ <p there is a unique f¯ ∈ N¯ := N¯ ⊗ K (which comes from our sep K sep ¯ ¯ fixed f ∈ L ) such that σf = e¯◦ f and for any τ ∈ G , η (τ) = sep <p e¯ (−f¯)◦τ(f¯) ∈ N¯ | = L¯. sep σ=id 8 VICTOR ABRASHKIN 2. Lie algebra L† In this Section we introduce a Lie F -algebra L¯† together with epi- p morphism of Lie algebras V : L¯† −→ L¯ and its section j0 : L¯ −→ L¯†. Let α = SpecF [U], Up = 0, be the group scheme over F with p p p the coaddition ∆(U) = U ⊗ 1 + 1 ⊗ U. We introduce the action Ω : L¯† −→ F [U]⊗L¯† of α on L¯†. U p p 2.1. Special choice of parameters r∗ = r∗(v ) and N∗ = N∗(v ). 0 0 Denote by δ = δ (v ) the minimum of the positive values of 0 0 0 1 v − (a +a /pn2 +···+a /pnu), 0 1 2 u s where 1 6 s,u < p, all n ∈ Z and a ∈ [0,(p− 1)v ) ∩Z0(p). The i >0 i 0 existence of such δ > 0 can be proved easily by induction on u. 0 Let r∗ ∈ Q be such that r∗ = b∗/(q∗ − 1), where q∗ = pN0∗ with 0 0 0 N∗ > 2, b∗ ∈ N and gcd(b∗,p(q∗ − 1)) = 1. The set of such r∗ is 0 0 0 0 everywhere dense in R and we can fix r∗ = r∗(v ) ∈ (v −δ ,v ). >0 0 0 0 0 For 1 6 s < p, introduce the following subsets in Q: — A[s] is the set of all a +a /pm2 +···+a /pmu, where 1 6 u 6 s, 1 2 u all m ∈ Z , all a ∈ [0,(p−1)v )∩Z0(p) and a 6= 0 (in particular, i >0 i 0 1 0 ∈/ A[s]). IfM ∈ Z wedenotebyA[s,M]thesubset ofA[s]consisted >0 of the elements satisfying the additional restrictions m ,...,m 6 M. 2 u Note that A[s,M] is finite. — B[s] is the set of all numbers r∗(b +b /pm2+···+b /pmu) where 1 2 u 1 6 u 6 s, all b ∈ Z and b +···+b < p. For M ∈ Z , let B[s,M] i >0 1 u >0 be the subset of B[s] consisted of the elements satisfying the additional restrictions m ,...,m 6 M. The set B[s,M] is also finite. 2 u Lemma 2.1. For any s, A[s]∩B[s] = ∅. Proof. Note that A[s] ⊂ Z[1/p]\{0}. Prove that B[s]∩Z[1/p] = {0}. It will be sufficient to verify that for any m ,...,m ∈ Z and 1 u >0 b ∈ Z such that 0 < b +···+b < p, we have i >0 1 u b pm1 +···+b pmu 6≡ 0mod(q∗ −1). 1 u 0 Since q∗ ≡ 1mod(q∗ −1) we can assume that all m < N∗. But then 0 0 i 0 0 < b1pm1 + ··· + bupmu 6 (p − 1)pN0∗−1 < q0∗ − 1. The lemma is proved. (cid:3) For α,β ∈ Q, set ρ(α,β) = |α−β|. Lemma 2.2. If α ∈/ B[s] then ρ(α,B[s]) := inf{ρ(α,β) | β ∈ B[s]} =6 0. Proof. Use induction on s. If s = 1 there is nothing to prove because B[1] is finite. Suppose s 6 p−2 and ρ(α,B[s]) > 0. RAMIFICATION VIA DEFORMATIONS 9 Choose Ms ∈ Z>0 such that r∗(p−1)/pMs < ρ(α,B[s])/2. If β ∈ B[s + 1] \ B[s + 1,M ] then there is β′ ∈ B[s] such that s ρ(β,β′) < ρ(α,B[s])/2. Then ρ(α,β) > ρ(α,β′)−ρ(β′,β) > ρ(α,B[s])−ρ(α,B[s])/2 = ρ(α,B[s])/2, and we obtain ρ(α,B[s+1]) > min{ρ(α,B[s+1,M ]),ρ(α,B[s])/2} > 0. s The lemma is proved. (cid:3) Lemma 2.3. If β ∈/ A[s] then ρ(β,A[s]) 6= 0. Proof. The proof is similar to the proof of above Lemma 2.2. (cid:3) Lemma 2.4. For all 1 6 u,s < p, ρ(A[u],B[s]) > 0. Proof. If u = 1 it follows from Lemma 2.2 because A[1] is finite. Suppose u 6 p−2 and ρ(A[u],B[s]) = δ > 0. us Choose Mu ∈ Z>0 such that (p−1)v0/pMu < δus/2. If α ∈ A[u + 1] \ A[u + 1,M ] then there is α′ ∈ A[u] such that u ρ(α,α′) < δ /2. Then for any β ∈ B[s], we have us ρ(α,β) > ρ(α′,β)−ρ(α,α′) > δ /2. us Therefore, for any α ∈ A[s], ρ(α,B[s]) > min{ρ(A[u+1,M ],B[s]),δ /2}} > 0. u us Lemma is proved. (cid:3) Specify the choice of N∗ = N∗(v ) ∈ N by the following conditions: 0 C1) N∗ ≡ 0modN∗; 0 C2) pN∗ρ(A[p−1],B[p−1]) > 2r∗(p−1); C3) r∗(1−p−N∗) ∈ (v −δ ,v ). 0 0 0 Under this choice we set q = pN∗ and b∗ = b∗(q − 1)/(q − 1) ∈ N 0 0 (note that r∗ = b∗/(q −1)). We need below the following property of such N∗. Proposition 2.5. If α ∈ A[p−1] and β ∈ B[p−1] then q|qα−(q −1)β| > b∗(p−1). Proof. Indeed, q2|α−β +β/q| > q2|α−β|−βq > q2ρ(A[p−1],B[p−1]) −r∗(p−1)q > 2r∗(p−1)q −r∗(p−1)q = r∗(p−1)q > b∗(p−1). (cid:3) 10 VICTOR ABRASHKIN 2.2. Lie algebra L¯†. Use the above r∗ = r∗(v ) and N∗ = N∗(v ). 0 0 Definition. A = A(r∗,N∗) is the set of all ι = pm(qα − (q − 1)β), where m ∈ Z , α ∈ A[p−1,m], β ∈ B[p−1,m] and |ι| 6 b∗(p−1). >0 Lemma 2.6. Suppose ι = pm(qα−(q −1)β) ∈ A. Then: a) m < N∗; b) pmα ∈ N and pmβ/r∗ ∈ Z do not depend on the presentation of >0 ι in the form pm(qα−(q−1)β) with α ∈ A[p−1,m] and β ∈ B[p−1,m]. Proof. If m > N∗ then by Proposition 2.5, |ι| > b∗(p −1) i.e. ι ∈/ A. Fromthe definitionof thesets A[p−1,m] andB[p−1,m] it follows that pmα,pmβ/r∗ ∈ Z . If ι = pm′(qα′−(q−1)β′) is another presentation >0 of ι then pmβ/r∗ and pm′β′/r∗ are non-negative congruent modulo q integersandthebotharesmallerthanq. Indeed, ifβ/r∗ = b +b p−m2+ 1 2 ···+b p−mu, where all m 6 m, then u i pmβ/r∗ 6 pm(b +···+b ) 6 pm(p−1) < pm+1 6 q 1 u because m < N∗. Similarly, pm′β′/r∗ < q. Therefore, they coincide and this implies also that pmα = pm′α′. (cid:3) Corollary 2.7. If ι = pm(qα − (q −1)β) ∈ A then the sum of digits b +···+b of the appropriate β/r∗ = b +b p−m2+···+b p−mu depends 1 u 1 2 u only on ι. Definition. With the above notation set ch(ι) := b +···+b . 1 u Definition. a) A+(p) is the set of all ι = pm(qα−(q−1)β) ∈ A (given in the above standard notation) such that: — ι > 0; — gcd(pmα,pmβ/r∗) 6≡ 0modp; — if s ∈ N and ι ∈ [(s−1)b∗,sb∗) then ch(ι) 6 p−1−s; b) A0(p) = A+(p)∪{0}; c) A0(p) = {ι ∈ A0(p) | ch(ι) = 0} = {qa | a ∈ Z0(p)∩[0,(p−1)v )}; 0 0 d)A0 (p) = {ι ∈ A0(p)|ch(ι) > 1}, A0(p) = {ι ∈ A0(p)|ch(ι) = 1}. >1 1 Proposition 2.8. If v is the normalized p-adic valuation then all p ιp−vp(ι), where ι ∈ A+(p), are pairwise different elements from Z+(p). Proof. Suppose ι = pm(qα−(q −1)β) ∈ A+(p). If ch(ι) = 0 then ι ∈ A0(p) and the correspondence ι 7→ ιp−vp(ι) 0 identifies A0(p) \ {0} ⊂ A+(p) with Z+(p) ∩ [0,(p − 1)v ). Indeed, 0 0 qpmα < b∗(p − 1) means that a = pmα < r∗(p − 1). According to property C3 from Subsection 2.1 this condition and the condition a < v (p−1)areequivalent because thebothimply thata 6 (v −δ )(p−1). 0 0 0 If ch(ι) > 1 then ιp−m ∈/ pN, i.e. m > v (ι). Indeed, ιp−m = p qα−(q −1)β ∈ pN implies (use that m < N∗) that p−m(b +b pm2 + 1 2

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.