arXiv:0810.0229v1 [math.NT] 1 Oct 2008
RECIPROCITY LAWS À LA IWASAWA-WILES
FRANCESC BARS, IGNAZIO LONGHI
Abstract. This paper is a brief survey on explicit reciprocity laws of
Artin-Hasse-Iwasawa-Wiles type for the Kummer pairing on local fields.
1. Introduction
Let K be a complete discrete valuation field, OK its ring of integers,
mK its maximal ideal and kK the residue field. Suppose that K is an ℓdimensional local field: this means that there is a chain of fields Kℓ =
K, Kℓ−1 , . . . , K0 where Ki is a complete discrete valuation field with residue
field Ki−1 and K0 is a finite field. We shall always assume that char(K0 ) = p.
Suppose char(kK ) = p > 0: then we have the reciprocity law map
(1)
( , K ab /K) : KℓM (K) → Gal(K ab /K),
where K ab is the maximal abelian extension of K, Gal denotes the Galois
group and KℓM is the Milnor K-theory.
Assume char(K) = 0 and ζpm ∈ K, where ζpm is a primitive pm -th root of
unity. The classical Hilbert symbol ( , )m : K ∗ × KℓM (K) →< ζpm >=: µpm
is:
(2)
(α0 , {α1 , . . . , αℓ })m
β ({α1 ,...,αℓ },K
:=
β
ab /K)
,
m
where β is a solution of X p = α0 and ({α1 , . . . , αℓ }, K ab /K) is the element
given by the reciprocity law map.
When ℓ = 1, K is the completion at some place of a global field (i.e.,
a finite extension of Fp ((T )) or of the p-adic field Qp ), K1M (K) ∼
= K ∗ and
ab
( , K /K) is the classical norm symbol map of local class field theory.
Historically there was deep interest to compute this Hilbert symbol (or,
better, Kummer pairing) in terms of analytic objects, as a step in the program of making local class field theory completely explicit. Vostokov [31]
suggests the existence of two different branches of explicit reciprocity formulas: Kummer’s type and Artin-Hasse’s type (later extended by Iwasawa
and Wiles). Kummer’s reciprocity law [21] is
Theorem 1 (Kummer 1858). Let K = Qp (ζp ), p 6= 2, and α0 , α1 principal
units. Then
−p
(α0 , α1 )1 = ζpres(log α˜0 (X)dlog α˜1 (X)X )
where α˜0 (X), α˜1 (X) ∈ Zp [[X]]∗ are power series such that α˜1 (ζp − 1) = α1 ,
α˜0 (ζp −1) = α0 , res means the residue and dlog is the logarithmic derivative.
1
2
FRANCESC BARS, IGNAZIO LONGHI
Artin-Hasse’s reciprocity law [2] is:
Theorem 2 (Artin-Hasse 1928). Let K = Qp (ζpm ), p 6= 2, and α1 ∈ K ∗ a
principal unit. Take π = ζpm − 1 a prime of K: then
T rK/Qp (− log α1 )/pm
(ζpm , α1 )m = ζpm
T rK/Qp (π −1 ζpm log α1 )/pm
, (π, α1 )m = ζpm
,
where log is the p-adic logarithm. Later Iwasawa [17] gave a formula for
K (p)
(α0 , α1 )m with α0 any principal unit such that valK (α0 − 1) > 2val
p−1 .
Roughly speaking the difference between these two branches is that Kummer’s type refers to residue formulas involving a power series for each component of the pairing, while Artin-Hasse-Iwasawa’s type are non-residue
formulas evaluating some generic series at the K-theory component of the
Hilbert pairing.
There is a big amount of articles in the literature that contribute to
and extend the above seminal works of Kummer and Artin-Hasse-Iwasawa.
The Hilbert symbol can be extended to Lubin-Tate formal groups, and also
to p-divisible groups. These extensions are defined from Kummer theory:
hence one often speaks of Kummer pairing instead of Hilbert symbol. Wiles
extended Iwasawa’s result to Lubin-Tate formal groups.
In this survey we intend to review some of the results on Artin-HasseIwasawa-Wiles’ type reciprocity laws. We list different variants of the Kummer pairing; afterwards we sketch some of the main points in the proof of
the Artin-Hasse-Iwasawa-Wiles reciprocity law for 1-dimensional local fields.
Finally we review Kato’s generalization of Wiles’ reciprocity law, which is
done in a cohomological setting. Kato’s work also extends the explicit reciprocity law to higher dimensional local fields and to schemes.
For Kummer’s type, one should cite also a lot of contributions by Vostokov, and many others, for example Shafarevich, Kneser, Brückner, Henniart, Fesenko, Demchenko, and Kato and Kurihara in the cohomological
setting. We refer to Vostokov’s paper [31] for a list of results and references
for reciprocity laws in this case, adding to it the recent works of Benois
[5] (for cyclotomic extensions), Cao [8] (for Lubin-Tate formal groups) and
Fukaya [12] and the works of Ankeny and Berrizbeitia [6]. Given α a unit of
OK , K a totally ramified finite extension of Qp and π a uniformizer of K,
n
α has a factorization modulo K ∗p by a product of E(π k ), k ∈ N, where E
is the Artin-Hasse exponential. The contribution by Shafarevich is to compute (α, β)n in terms of the above factorization for α and β. Berrizbeitia [6]
recovers Shafarevich results with different methods but also using the above
factorization.
Notations
Let L be a valuation field: we denote OL the ring of integers, mL its maximal
ideal, kL the residue field and valL the valuation function. We write L for
the algebraic closure of L if char(L) = 0 and the separable closure otherwise;
RECIPROCITY LAWS FOR LOCAL FIELDS
3
ˆ will be the completion (which is algebraically closed). Denote Gal(L/L)
L
by GL and the continuous Galois cohomology group H i (GL , A) by H i (L, A).
For any ring R, R∗ means the invertible elements of R. As usual Fq is
the finite field of cardinality q.
The symbol M will always denote a finite extension of Qp : M0 is the
subfield of M such that M/M0 is totally ramified and M0 /Qp is unramified.
For N a Zp -module, N (r) will be its r-th Tate twist, r ∈ Z (recall that
Zp (1) := lim
µpn )
←−
n
2. The Kummer pairing
Let K be an ℓ-dimensional local field. Then K is isomorphic to one of
• Fq ((X1 )) . . . ((Xℓ )) if char(K) = p
• M ((X1 )) . . . ((Xℓ−1 )) if char(K1 ) = 0
• a finite extension of M {{T1 }} . . . {{Tn }}((Xn+2 )) . . . ((Xℓ )) if
char(Kn+1 ) = 0 and char(Kn ) = p > 0
where L{{T }} is
)
(+∞
X
i
ai T : ai ∈ L, inf(valL (ai )) > −∞, lim valL (ai ) = +∞ .
i→−∞
−∞
P
Defining valL{{T }} ( ai T i ) := min{valL (ai )} makes L{{T }} a discrete valuation field with residue field kL ((t)) (see Zhukov [33]).
Since we assume char(kK ) = p > 0 it follows that the ℓ-dimensional
local field K has residue field isomorphic to Fq ((X1 )) . . . ((Xℓ−1 )). Then we
have the reciprocity map (1) obtained by Kato [18] (see the exposition [23]),
which had already been proved by Parshin [25] when char(K) = p > 0.
Now we introduce the Kummer pairing through the classical Hilbert symbol (2). Assume first char(K) = 0 and ζpm ∈ K and restrict the Hilbert
pairing (2) to (1 + mK ) × KℓM (K): then
(1 + δ0 , {α1 , . . . , αℓ })m
β ({α1 ,...,αℓ },K
=
β
ab /K)
= ζpcm ,
m
where β is a solution of X p = 1 + δ0 and c ∈ Z/pm Z is determined by δ0
n
and the αi ’s. Recall that Milnor K-groups KnM (K) are defined as (K ∗ )⊗
modulo the subgroup generated by a ⊗ (1 − a), a ∈ K ∗ − 1.
We rewrite the above restricted pairing as
m
ˆ
+ 1)p − 1 = 0}
mK × KℓM (K) → Wpm := {ζpjm − 1|j ∈ Z} = {w ∈ K|(w
(3)
(δ0 , {α1 , . . . , αℓ })m := ζpcm − 1.
b m is the formal group given by X + b Y := X + Y + XY and
Recall that G
Gm
P∞
i X i . Summation pm -times in G
b m is given by [pm ](X) =
(−1)
−G
bmX =
i=1
4
FRANCESC BARS, IGNAZIO LONGHI
m
(1 + X)p − 1. Observe that
(4)
(δ0 , {α1 , . . . , αℓ })m = (({α1 , . . . , αℓ }, K ab /K) −G
b m 1)(β)
m
where β is a solution of (1 + X)p − 1 = δ0 .
b m can be exThis reformulation of the classical Kummer pairing for G
tended to classical Lubin-Tate formal groups, to Drinfeld modules and in
greater generality to p-divisible groups.
2.1. Classical Lubin-Tate formal groups. Lubin-Tate formal groups are
defined for 1-dimensional local fields: classically the emphasis is on the
unequal characteristic case and in this survey we shall restrict to such setting
when considering Lubin-Tate formal groups. We refer the interested reader
to [15] and [28] for proofs and detailed explanations.
Recall that a formal 1-dimensional commutative group law F over OM is
F (X, Y ) ∈ OM [[X, Y ]], satisfying:
(i) F (X, Y ) ≡ X + Y mod deg 2;
(ii) F (X, F (Y, Z)) = F (F (X, Y ), Z);
(iii) F (X, Y ) = F (Y, X).
We fix π a uniformizer of M . Define
l
Fπ := {f ∈ OM [[X]] f ≡ πX mod deg 2, f ≡ X p mod mM },
where pl is the number of elements of kM .
Theorem 3 (Lubin-Tate). We have:
1. for every f ∈ Fπ it exists a unique 1-dimensional commutative formal
group law Ff defined over OM and called of Lubin-Tate, such that f ∈
End(Ff ) (i.e. Ff (f (X), f (Y )) = f (Ff (X, Y ))).
2. Given f, g ∈ Fπ then Fg and Ff are isomorphic over OM .
3. Given f ∈ Fπ and a ∈ OM , it exists a unique [a]f (X) ∈ OM [[X]] with
[a]f ∈ End(Ff ) and [a]f (X) ≡ aX mod deg 2; the map a 7→ [a]f gives an
embedding OM → End(Ff ). We denote also [a]Ff by [a]f .
b m over Qp corresponds to the
Example The formal multiplicative group G
Lubin-Tate formal group with π = p and f = (1 + X)p − 1 ∈ Fp .
We denote by Ff (B) the B-valued points of Ff . Consider
WFf ,m := {zeroes of [π m ]Ff in Ff (m ˆ )}.
K
Remark 4. We want to emphasize that we have an “embedding” of OM
into End(Ff ), from which we get a tower of field extensions M (WFf ,m ) in
ˆ . For example take F = G
b and consider [pm ]
as m varies in N:
M
f
m
m
bm
G
p − 1 are ζ m − 1, thus M (W
the roots of [pm ]G
p
b m ,m ) is
b m (X) = (1 + X)
G
the cyclotomic extension M (ζpm ). The groups WFf ,m are the Lubin-Tate
analogs of µpm .
RECIPROCITY LAWS FOR LOCAL FIELDS
5
Let K be a finite extension of M such that WFf ,m ⊆ K. The Kummer
pairing for Ff is:
( , )m : Ff (mK ) × K ∗ → WFf ,m
(5)
(a, u)m := ((u, K ab /K) −Ff 1)β
where β is a solution of [π m ]Ff (X) = a.
There is a more general notion of Lubin-Tate formal group, called relative
Lubin-Tate formal group, which involves the unique unramified extension of
M of degree d. The corresponding formulation of the Kummer pairing differs
slightly from (5). We refer to [10, Chapter 1, §1.1,§1.4,§4.1] for the precise
statement.
We also remark that there is a notion of n-dimensional Lubin-Tate formal
group [15]. Also in this case we have an embedding of OM into End(Ff ),
and some generalized Kummer pairing appears. Since Kummer pairing on
p-divisible groups will also include this case we do not discuss it any further
here.
2.2. 1-dimensional local fields in char = p > 0: Drinfeld modules.
The key property of Lubin-Tate formal groups is the embedding: OM →
End(Ff ). This suggests to define a Kummer pairing for 1-dimensional local
fields K with char(K) = p > 0 in the following way.
For simplicity we take F = Fq (T ) and put A = Fq [T ]. Observe that
EndFq (Ga /F ) ∼
= F {τ }, where τ a = aq τ for a ∈ F , τ 0 = id. We can think of
F {τ } as a subset of F [X], via τ0 ↔ X, τ ↔ X q : then F {τ } consists of the
additive polynomials and the product in F {τ } corresponds to composition.
Drinfeld [11] defined elliptic modules (now called Drinfeld modules) as
embeddings
Φ : A → EndFq (Ga /F )
a 7→ Φa = a + (. . .)τ,
Fq -linear and non-trivial (i.e., there is a ∈ A such that Φa is not equal to a).
For a ∈ A we write Φ[a] := {zeroes of Φa (X)}. We have that Φ[a] ∼
=
(A/(a))d and define rank(Φ) := d. Let p = (π) be a place of A: under some
technical assumptions (see [3] for a reference) we can extend Φ to
b a) ∼
Ap → EndFq (G
= Fp{{τ }}
which we also call Φ, a 7→ Φa , where Fp is the completion of the field F at
p and Fp{{τ }} is the ring of skew power series.
Denote by WΦ,πm the set of roots of Φπm (X) in Fp. For any finite extension K/Fp containing WΦ,πm we define the Kummer pairing:
( , )m : mK × K ∗ → WΦ,πm
(6)
(a, u)m := ((u, K ab /K) − 1)(β)
where β is a root of Φm
π (X) = a.
6
FRANCESC BARS, IGNAZIO LONGHI
For example for rank 1, the simplest Drinfeld module is the Carlitz module
defined by ΦT (τ ) = T id + τ . Then ΦT 2 (τ ) = ΦT (τ ) ◦ ΦT (τ ) = (T id + τ ) ◦
(T id + τ ) = T 2 id + τ T + T τ + τ 2 = T 2 id + (T q + T )τ + τ 2 and similarly
2
ΦT 3 (τ ) = T 3 id + (T 2q + T q+1 + T 2 )τ + (T q + T q + T )τ 2 + τ 3 . If we take
p = (T ) (π = T ) then Fp ∼
= Fq ((T )) and Wφ,T = {zeroes of T X + X q },
2
Wφ,T 2 = {zeroes of T 2 X + (T q + T )X q + X q }, Wφ,T 3 = {zeroes of T 3 X +
3
2
2
(T 2q + T q+1 + T 2 )X q + (T + T q + T q )X q + X q }, ...
2.3. p-divisible groups. A vast extension of the theory above considers
p-divisible groups (also called Barsotti-Tate groups). For the definition and
main properties see [29].
Let G be a p-divisible group scheme over OK of dimension d and finite
height h, where K is any ℓ-dimensional local field with char(kK ) = p > 0.
We denote by [pm ]G the pm -th power map G → G and let G[pm ] be the
group scheme kernel. As usual, X(B) denotes the B-points of the scheme
X. We put WG,pm := G[pm ](OK ) and we impose WG,pm = G[pm ](OK ), i.e.
WG,pm ⊂ K. Then Kummer pairing is defined by:
( , )m : G(mK ) × KℓM (K) → WG,pm
(7)
(a, u)m := ((u, K ab /K) −G 1)(β)
where β is a root of [pm ]G (X) = a. Finally we observe that p-groups with an
action of OK have also been studied [29]. In this case (7) can be reformulated
with respect to [π m ]G , π a uniformizer of K.
2.4. Cohomological interpretation. Let K be an ℓ-dimensional local
field with char(kK ) = p > 0 and char(K) = 0 . The Kummer pairing
also admits an interpretation in terms of Galois cohomology. We do this in
the generality of p-divisible groups. Consider the exact sequence
0 → G[pm ](mK ) → G(mK ) → G(mK ) → 0.
This induces:
δ1,G,m : G(mK ) → H 1 (K, G[pm ](OK )).
We assume WG,pm ⊂ K, i.e. G[pm ](OK ) = G[pm ](OK ). The Galois symbol
map:
hrK : KrM (K) → H r (K, Zp (r))
is obtained by cup product h1K ∪ . . . ∪ h1K , where h1K is the connecting
morphism h1K : K ∗ → H 1 (K, Zp (1)) from the usual Kummer sequence.
There is a canonical isomorphism [18] H ℓ+1 (K, Zp (ℓ)) ∼
= Zp defining the
following pairing:
(d
, )m : G(mK ) × KℓM (K) → H ℓ+1 (K, Zp (ℓ) ⊗ G[pm ](OK )) ∼
= G[pm ](OK )
(8)
[
(a,
u)m := (−1)ℓ δ1,G,m (u) ∪ hℓK (a).
RECIPROCITY LAWS FOR LOCAL FIELDS
7
[
Then by [12, Proposition 6.1.1] (a,
u)m = ((u, K ab /K) −G 1)(β) = (a, u)m
m
where β is a root of [p ]G (X) = a.
If one could define a good analog in characteristic p of the Galois symbol
maps for the cohomological groups from Illusie cohomologies used by Kato
[18] to obtain the reciprocity law map ( , K ab /K), then a cohomological
interpretation should appear when char(K) = p > 0 (following arguments
like the proof of [12, Proposition 6.1.1]). We refer to [23, §5] for a very quick
review of the definitions of these cohomological groups and of Kato’s higher
local class field theory.
2.5. Limit forms of the Kummer pairing. In this subsection ( , )m
denotes any of the pairings (4) to (7).
Let W♯,m be one of WFf ,πm , WΦ,πm or WG,pm . We shorten K(W♯,m ) to
K♯,m and write O♯,m (resp. m♯,m ) for the ring of integers (resp. the maximal
ideal). By definition the Tate module is Tp (♯) := lim
W♯,m where the limit is
←−
m
taken with respect to the map [♯] (which means [π]f , Φπ or [p]G ). Consider
W♯,∞ := lim
W♯,m = ∪m W♯,m .
−→
m
The symbol M♯,m denotes one of Ff (m♯,m ), m♯,m or G(m♯,m ). Consider
lim
M♯,m as the direct limit of [♯] : it consists of sequences (an )n≥N (for
−→
m
some N ∈ N) such that an ∈ K♯,n and an+1 = [♯]an .
We need to impose that the K♯,m ’s are abelian extensions of K (this is
satisfied for example by 1-dimensional classical Lubin-Tate formal groups
and rank 1 Drinfeld modules). Then we have a limit version of the Kummer
pairing:
( , ) : lim
M♯,m × lim
KℓM (K♯,m ) → W♯,∞
−→
←−
(9)
m
m
where lim
KℓM (K♯,m ) is with respect to the Norm map. The limit pairing is
←−
m
well defined: by the abelian assumption we have (Kato-Parshin’s reciprocity
ab /K ) acting on the roots
ab
/K♯,n−1 ) = (u′ , K♯,n
law) (NK♯,n /K♯,n−1 (u′ ), K♯,n−1
♯,n
of [♯]n−1 (X) = a, when u′ ∈ KℓM (K♯,n ).
We remark that the above is often formulated without taking the whole
limit tower. That is, suppose that one wants to compute explicitly (a, u)m
and u′ ∈ KℓM (K♯,m+k ) is given so that NK♯,m+k /K♯,m (u′ ) = u: then by a
similar argument as needed for defining (9) we have
(10)
(a, u)m = ([♯]a, NK♯,m+k /K♯,m+1 (u′ ))m+1 = . . . = ([♯]k a, u′ )m+k .
We can also write the above limit form (9) as:
(11)
KℓM (K♯,n ) → Hom(M♯,m , Tp (♯))
lim
←−
n
8
FRANCESC BARS, IGNAZIO LONGHI
sending u = (un )n to a 7→ lim (a, un )n with fixed m. These homomorphisms
n→∞
are continuous because of the continuity of the Kummer pairing: see [3,
Lemma 15] for the case ℓ = 1.
3. Explicit reciprocity law formulas à la Wiles’ for
1-dimensional local fields
In this section we fix ℓ = 1 and sketch some of the main ideas to obtain
explicit reciprocity laws for a 1-dimensional local field K in the context of
classical Lubin-Tate formal groups and rank 1 Drinfeld modules, introducing
Coleman power series. See the last section for different approaches to this
explicit reciprocity law through the exponential or dual exponential map.
Let K be either M or Fp. In the Drinfeld module case Φ is required to be
of rank 1 (in order to obtain abelian extensions) and sign-normalized. We
refrain from explaining this last technical condition (the reader is referred
to [3, §2.1] and the sources cited there) and just notice that when A = Fq [T ]
as in §2.2 this means that Φ is the Carlitz module, i.e. ΦT (X) = T X + X q .
Furthermore we take π ∈ A a monic polynomial.
We lighten the notation introduced in §2.5 by shortening K♯,m to Km .
The extensions Km /K, are totally ramified and abelian: they are generated by roots of Eisenstein polynomials and Gal(Km /K) ∼
= (OK /(π)n )∗ .
The Tate module Tp (♯) is a rank 1 OK -module with OK -action given by
α · γ := [α]f γ or Φα (γ). Let (εn )n be an OK -generator of Tp (♯): then K♯,m =
K(εm ), since εm generates W♯,m , and one has [♯]εn+1 = εn , [♯]n+1 εn = 0
where [♯] is respectively [π]f or Φπ . Moreover the εn ’s form a norm compatible system. Denote ∪m K(W♯,m ) by K♯,∞ .
In this section ( , )m refers to (5) or (6) and ( , ) refers to (9).
First, notice that ( , )m is bilinear, additive in the first variable and
multiplicative in the second variable. In particular ( , ζ)m = 0 for any root
of unity ζ.
Consider the character
∗
χ : Gal(K♯,∞ /K) → OK
σ 7→ χσ defined by σ((εn )n ) = χσ ·((εn )n ); it can be thought of as a character
b m this χ corresponds to the cyclotomic
of Gal(K ab /K). We remark that for G
∗ → Gal(K
character χcycl . The main point is the following: χ−1 : OK
♯,∞ /K)
coincides with the inverse of the local norm symbol map. Therefore:
(
([NKm /K u]−1
f −Ff 1)(ε2m ) K = M
(12)
(εm , u)m =
K = Fp.
Φ(NKm /K u)−1 (ε2m ) − ε2m
Inspired by (12) one can ask: could we find h such that (am , um )m =
[h(a, u)]f (εm ) for classical Lubin-Tate formal groups or = Φh(a,u) (εm ) for
rank 1 Drinfeld modules? Observe that h(a, u) modulo π m defines the same
action over εm (because W♯,m is isomorphic to OK /(π m )).
RECIPROCITY LAWS FOR LOCAL FIELDS
9
In this direction one obtains the following results. From class field theory
it follows (a, a)n = 0. Exploiting linearity and continuity of the pairing
( , )m one obtains (assuming v(c) greater than a fixed value which depends
linearly of n),
dlog w
, εn )n
(13)
(c, w)n = (cεn
dεn
where
∗
dlog : O♯,n
→ Ω1O♯,n /OK
1
is the map x 7→ dx
x (recall that the module of Kähler differentials ΩO♯,n /OK
is free with generator dεn over O♯,n /dKn /K , where dKn /K is the different of
Kn over K). Notice that one can pick a power series g ∈ OK ((X)) such that
g(εn ) = w and dlog w/dεn = g′ (εn )/g(εn ).
Finally one proves that exists m > n such that
(an , εn )n = −(εm , 1 + am ε−1
m )m
giving a positive answer to the question above.
This suggests the following definition of an analytic pairing:
dlog u
−n
· εn .
[a, u]n := T rKn /K π λ(a)
dεn
Here · is the action on the Tate module at level n and λ is the logarithm
map defined by λ(a) := lim π1n [♯]n a (the limit exists for valK (a) sufficiently
big).
In order to compare (a, u)n and [a, u]n , Iwasawa (theorem 2) imposes
the condition that there exists m such that u = NQp (ζpm )/Qp (ζpn ) (u′ ) : then
a, u′ )m and it is at level m that he compares the two
(a, u)n = ([p]m−n
b
Gm
m−n
′
′
pairings obtaining ([p]m−n
b m a, u ]m .
b m a, u )m = [[p]G
G
The general case follows Iwasawa’s argument. Here we will state a limit
version, hence we suppose that (un ) ∈ lim Kn∗ (limit w.r.t. the norm).
←
To express in compact form the limit of the pairings [ , ]n it is convenient
to introduce Coleman’s power series.
Theorem 5. Let K, Ff , Φ be as above. Then
(1) [9] (case K = M ) There exists a unique operator N (the Coleman
norm) defined by the property
Y
(N h) ◦ f (X) =
h(X +Ff w)
w∈WFf ,0
for any h ∈ M ((x))1 (the set of those Laurent series which are convergent in the unit ball).
(2) [3] (case K = Fp) There exists a unique operator N such that
Y
h(x + v)
(N h) ◦ Φπ =
v∈Φ[p]
10
FRANCESC BARS, IGNAZIO LONGHI
for any h ∈ Fp((x))1 .
Moreover, in both cases the evaluation map ev : f 7→ (f (εn ))n gives an
isomorphism
∗
.
(OK ((x))∗ )N =id ∼
= lim K♯,n
←
∗
Denote by Colu the power series associated to u ∈ lim K♯,n
. Then we
←
define the limit form of the analytic pairing by
∗
[ , ] : lim
M♯,m × lim
K♯,m
→ W♯,∞
−→
←−
n
n
[a, u] := T rK♯,n /K (π −n λ(an )dlog Colu (εn )) · εn .
un
Notice that dlog Colu (εn ) = dlog
dεn .
Finally one observes that [a, u] has similar properties to (a, u); in particular [a, u] = [an εn dlog Colu (εn ), εn ]n (compare with (13)) and to prove the
reciprocity law one is reduced to show [ , εn ]n = ( , εn )n for n big enough.
Theorem 6. Under the above notation, we have
(1) [32] ( , ) = [ , ] for classical Lubin-Tate formal groups Ff .
(2) [3] ( , ) = [ , ] for Fp and Φ.
In [32] Wiles proved this result for classical Lubin-Tate formal groups
without using Coleman power series: he takes m big enough in order to
b m . This strategy
compare ( , )m and [ , ]m , like Iwasawa had done for G
requires a very precise valuation calculation. For a detailed explanation one
can look also at [24, §8,9]. The above limit version with Coleman power
series for classical Lubin-Tate formal groups can be found in [10, I,§4].
For the Carlitz module, Anglès in [1] obtained Wiles’ version of theorem
6, i.e. without Coleman power series.
4. Explicit reciprocity laws and higher K-theory groups
The starting point for the results in §3 is explicit local class field theory
applied to ε and the dlog homomorphism. In [19] and [20] Kato reinterpreted Wiles’ reciprocity law as an equality between two maps obtained by
composition of natural maps from cohomology groups and gave generalized
reciprocity laws in higher K-theory. Here we introduce his approach to the
classical Hilbert symbol for ℓ-dimensional local fields, ℓ > 1, and reformulate
parts of it for rank 1 Drinfeld modules.
4.1. Exponential map in the Hilbert symbol. In this paragraph we
assume char(K) = 0 and char(kK ) = p > 0. Take:
∗
expδ : OK → OK
valK (p)
(p−1) where
2valK (p)
≥ (p−1) , the
P
Xm
m≥0 m!
given by expδ (a) = exp(δa) if valK (δ) >
exp(X) =
is the exponential. Assuming valK (δ)
map expδ extends to
RECIPROCITY LAWS FOR LOCAL FIELDS
11
a group morphism [22]:
r−1
M
→ K\
expδ,r : ΩO
r (K)
K
adlog b1 ∧ · · · ∧ dlog br−1 7→ {exp(δa), b1 , . . . , br−1 },
Vr 1
M (K) := lim K M (K)/pn K M (K), Ωr :=
where K\
Ω , Ω1 are the Kähler
r
←−
n
r
r
R
R
R
R
differentials and r is a strict positive integer. The function expδ,r factors
Vr−1 1
through O
Ω̂OK where Ω̂1OK is the p-adic completion of Ω1OK .
K
Sen [27] generalizes theorem 2:
Theorem 7. Assume ζpn ∈ M . Fix π a uniformizer of M , α1 ∈ OM − {0},
K (p)
+ valK (α1 ). Then
and α ∈ K with valK (α − 1) ≥ 2val
p−1
(α, α1 )n = ζpcn , with c =
ζpm g′ (π)
−1
log α)
T
r
(
M/Q
p
pn
h′ (π) α1
where g, h ∈ OM0 [T ] are such that g(π) = α1 and h(π) = ζpn , and the
pairing is (2).
M (p)
n
From now on take valM (η) = 2val
p−1 , η ∈ M0 (ζp ). The proof of theorem
7 reduces to theorem 2 because T rM/M0 (ζpm ) (adπ) = T rM/M0 (ζpm ) ( h′ a(π) )dζpm
and the commutativity of the following diagram ([22, p.217]):
expη,2
Ω̂1OM
−−−−→
T rM/M (ζ )
y
0 pm
Ω̂1OM
0 (ζpn )
h
d
M
K
−−−M−→
2 (M )
NM/M (ζ )
y
0 pm
hM
Z/pn (1)
yid
n
expη,2
b M (M0 (ζpn )) −−−0−−p−
−−−−→ K
→ Z/pn (1)
2
(ζ
)
where h∗ is the Hilbert symbol {a, b} 7→ (a, b)n .
We rewrite Sen’s theorem as a commutative diagram. Take γ ∈ OM with
M (p)
valM (γ) = valp−1
. Fontaine proved that we have the following isomorphism
−1 −1
ΨM : Ω̂1OM /pn γ −1 → γ −1 d−1
dM/M0 (ζ
M/M0 /γ
pn )
(1)
induced from the map adlog ζpm 7→ ap−m ⊗ (ζpm )m>0 where dL1 /L2 is the
different of the finite extension of fields L1 /L2 . Commutativity of the diagram
Ψ
−1 d−1
−→ γ −1 d−1
Ω̂1OM /pn γ −1 −−−M
M/M0 /γ
M/M0 (ζpn ) (1)
−expη y
T rη ⊗idy
h
K2M (M )/pn −−−M−→
Z/pn (1)
(where T rη is the map induced by x →
7 T rM/Qp (ηx)) is Sen’s theorem [22,
§4].
Now we consider ℓ > 1. Assume ζpn ∈ K. Consider K0 with K/K0 a
finite and totally ramified extension such that p is a prime element of OK0 .
12
FRANCESC BARS, IGNAZIO LONGHI
Recall kK = Fq ((t1 )) . . . ((tℓ−1 )). Take M such that M = M0 , M ⊆ K0
and the residue field is Fq : then K0 = M {{T1 }} · · · {{Tℓ−1 }}. Kurihara
extends Sen’s result to higher Milnor K-theory (theorem 8) by means of the
commutativity of the diagram:
expη,ℓ+1
−−−−−→
Ω̂ℓOK
T rK/K (ζ )
y
0 pm
h
[
M
K
−−−K−→
ℓ+1 (K)
NK/K (ζ )
y
0 pm
h
K0 (ζpn )
expη,ℓ+1 [
M (K (ζ n )) −
Ω̂ℓOK (ζ ) −−−−−→ K
−−−−−
→
0
p
ℓ+1
n
0 p
Res
yRes
y
expη,2
Ω̂1OM (ζ
−−−−→
pn )
where Res : Ω̂ℓOK
0 (ζpn )
w where w ∈ Ω̂1OM (ζ
pn )
h
M (ζpn )
d
M
K
2 (M (ζpn )) −−−−−→
Z/pn (1) = µpn
yid
Z/pn (1)
yid
Z/pn (1)
→ Ω̂1OM is defined by Res(wdlog T1 ∧· · ·∧dlog Tℓ−1 ) =
, Res is the Kato’s residue homomorphism in Milnor
K-groups and hL is the Hilbert symbol {a1 , . . . , aℓ+1 } 7→ (a1 , {a2 , . . . , aℓ+1 })n .
∗ and α ∈ O ∗ with val (α − 1) ≥
Theorem 8 ([22]). Take α1 , . . . , αℓ ∈ OK
K
K
2valK (p)
p−1 . Take fi (T, T2 , . . . , Tℓ ) ∈ OK0 [T ] such that fi (π, T2 , . . . , Tℓ ) = αi ,
and h(T ) ∈ OK0 [T ] with h(π) = ζpn where π is a fixed uniformizer of K.
Then (α, {α1 , . . . , αℓ }) = ζpcn where
!
T2 · · · Tℓ ζpn
∂fi
1
.
)1≤i,j≤ℓ
det(
c = − n T log α
p
α1 · · · αℓ h′ (π)
∂Tj
|T1 =π
P
Here T := T rM (ζpn )/Qp ◦cK0 (ζpn )/M (ζpn ) ◦T rK/K0(ζpn ) , with cL{{T }}/L ( ai T i ) :=
a0 and cL{{T2 }}{{T3 }}···{{Tk }}/L defined recursively as composition of the maps
cL{{T2 }}{{T3 }}···{{Ti+1 }}/L{{T2 }}{{T3 }}···{{Ti }} .
Remark 9. Benois in [4] extends theorem 7 to formal groups.
4.2. Kato’s generalized explicit reciprocity laws. Kato generalizes Wiles’
reciprocity law for an unequal characteristic local field K giving a natural
interpretation in the context of cohomology groups. Here we introduce two
generalizations.
4.2.1. Local approach. Let us first rewrite (11) in the context of Wiles’ reciprocity law: let π be a fixed uniformizer of M , Ff a formal Lubin-Tate
group and ε a fixed generator for the Tate module. For simplicity we assume that p is prime in K. Recall [20, Remark 4.1.3] that a Lubin-Tate
formal group G over OM has the following characterization in p-divisible
groups: dim(G) = 1, and the canonical map End(G) → EndOM (LieG) ∼
=
OM is an isomorphism, where Lie(G) is the tangent of G at the origin
and coLie(G) is HomOM (Lie(G), OM ). Put G′ := G ×OM OKFf ,m ; then
RECIPROCITY LAWS FOR LOCAL FIELDS
13
Lie(G′ ) ⊗OM Q ∼
= KFf ,m and coLie(G′ ) ⊗OM Q ∼
= KFf ,m . Consider the map
∗
̺ε : lim
K
→
K
given
by
the
composition
Ff ,m
Ff ,n
←−
n
( , )(11)
◦exp
lim
KF∗ f ,n −−−−−→ HomOM ,cont(Ff (mf,m ), OM ) −−−−→ HomOM (Lie(G′ ), M )
←−
n
Tr
−−−−→
KFf ,m ∼
= Q ⊗OM coLie(G′ )
where the last isomorphism is Kato’s trace pairing [19, II,§2] and exp is the
exponential map.
Wiles’ reciprocity law affirms that ̺ε (a) has an expression in terms of
dlog , which can be reformulated by defining a natural map δε . Here we
define ̺ε and δε in Kato’s generality [20, §6]: this essentially includes LubinTate formal groups and the generalized Wiles’ reciprocity law obtained in
§3 ([20, 6.1.10]).
Let K be an ℓ-dimensional local field with char(K) = 0 and char(kK ) =
p > 0. Take G a p-divisible group over OK with dim(G) = 1. Suppose
Λ ֒→ End(G) with Λ an integral domain over Zp which is free of finite rank
as Zp -module. Suppose Tp G is a free Λ-module of rank 1 and fix a generator ε. Let Kn /K be the field extension corresponding to Ker(GK →
AutΛ (Tp G/pn )). Denote the p-adic GK -representation Tp G ⊗ Qp by Vp G:
to it one can apply Fontaine’s theory. By Fontaine’s ring BdR,K (here we
just recall that it is a filtered GK -module - see [19, §2] for more) one constructs the filtered module DdR,K (Vp G) := H 0 (K, BdR,K ⊗Qp Vp G). One
has that gr −1 DdR,K (Vp G) is canonically isomorphic to Q ⊗ Lie(G) and
dimK (DdR,K (Vp G)) = dimQp (Vp G). Then one defines a map [20, Proposition(2.3.3),p.118]
ℓ−1
⊗OK coLie(G)
FDR : H ℓ−1 (K, Zp (ℓ) ⊗ HomΛ (Tp G, Λ)) → Ω̂K
where Ω̂rK is the p-adic completion of ΩrOK /Z tensored by Q.
Kato extends the above ̺ε to
GSM
̺ε : lim
KℓM (Kn ) −−−−→ H ℓ (K, Zp (ℓ) ⊗ HomΛ (Tp G, Λ))
←−
n
F
DR
−−−
−→
ℓ−1
⊗OK coLie(G) .
Ω̂K
The recipe to construct GSM is: fix n, then compose the Galois symbol
map defined in §2.4 (but with coefficients in (Z/pn Z)(ℓ)) instead of Zp (ℓ))
ℓ
n
ℓ
n
with ∪ε−1
n : H (Kn , (Z/p Z)(ℓ)) → H (Kn , (Z/p Z)(ℓ) ⊗ HomΛ (Tp G, Λ))
and take the trace map to have the cohomology group over K, and finally
take the limit with respect to n.
As for δε : up to some technical details which we do not reproduce here
(see [20, p.119]) it is essentially the map dlog : KℓM (Kn ) → Ω̂ℓK given by
{α1 , . . . , αℓ } 7→ dlog (α1 ) ∧ · · · ∧ dlog (αℓ ).
Theorem 10. One has ̺ε = (−1)ℓ−1 δε [20, Theorem 6.1.9].
14
FRANCESC BARS, IGNAZIO LONGHI
Remark 11. Kato obtains a generalized reciprocity law when dim(G) = 1,
Tp G is Λ-free with rankΛ Tp G = ℓ ≥ 1 and some technical conditions
from Fontaine’s theory are satisfied [20, Theorem 4.3.4]. He also defines
̺s (from Fontaine’s theory) and δs (on the dlog side) with s ∈ Nℓ as maps
lim
KℓM (Yn ) → O(Ym ) ⊗OK coLie(G)⊗(ℓ+r(s)) where Yn is a scheme repre←−
n
senting a functor related with the pn -torsion points associated to G and
r(s) ∈ N.
Remark 12. Kato’s generalized reciprocity laws (theorem 10 and remark
11) did not include a generalization of Coleman power series. Fukaya [13]
obtains a K2 -analog for the Coleman isomorphism.
Remark 13. In the previous paper [19, II] Kato had obtained a reciprocity
law for a map ̺ˆε whose definition involves the Kummer map (11) for the
tower of fields given by any classical Lubin-Tate formal group (i.e. ℓ = 1 and
K = M ) and the dual exponential map (see below) associated to a certain
representation of the formal group (more precisely, to a power of a Hecke
character obtained from a CM elliptic curve). He relates ̺ˆε with a map δ̂ε
constructed mainly from dlog . This reciprocity law is expressed in terms of
Coleman power series [19, II]. Tsuji in [30, I] extends Kato’s reciprocity law
[19, II] to representations coming from more general Hecke characters.
4.2.2. Global approach. Now the base field is M , in particular ℓ = 1 and
car(kM ) = p > 0. As explained in [26, §3.3], Kato proves that Wiles’ reciprocity (or rather Iwasawa’s, since it is the case Ff = Gm ) is equivalent to
the commutativity of the following diagram (which follows from cohomological properties):
∗
Q ⊗ lim
OM
(ζpn )
←−
n,N orm
dlog y
Q ⊗ lim
Ω1OM (ζ
←−
n,T race
pn )
−−−−−−→
Kummer
ζ n 7→1
yp
n
Q ⊗ lim
OM [ζpn ]/pn
←−
cor
y
Q ⊗ lim
H 1 (Km , Z/pn )
←−
n
t−,−
M (ζpm )
n
Q ⊗ lim
H 1 (M (ζpn ), Z/pn )
←−
/OM
dlog ζpn 7→1y
(tn,m )n y
Homcont (GM (ζpn )) , µpn )
Q ⊗ lim
←−
∼
=
−−−−−−−−→
∪ p1m log χcycl
yinc
ˆ)
H 1 (M (ζpm ), M
1
where tn,m := pn−m
T rM (ζpn )/M (ζpm ) , cor is the corestriction map, χcycl is the
cyclotomic character as in §3, ∪ log χcycl is the cup product by the element
RECIPROCITY LAWS FOR LOCAL FIELDS
15
log χcycl ∈ Homcont (GM , Zp ), inc is the map induced by lim
Z/pn = Zp ֒→
←−
n
ˆ and Hom
n
∼ 1
M
cont (GM (ζpn ) , µpn ) = H (M (ζpn ), Z/p (1)) by ζpn 7→ 1.
Inspired by this cohomological approach, Kato formulates a new reciprocity law. To state it we need to introduce some more notation. Take
a smooth OM -scheme U , complement of a divisor Z with relatively normal
crossings in a smooth proper OM -scheme X. We assume there is a theory of
Chern classes for higher Quillen K-theory giving functorial homomorphisms
r
(U ⊗M M (ζpn ), Zp (r))
chr : Kr (U ⊗OM OM (ζpn ) ) → Het
where the Kr ’s are Quillen K-theory groups and Het denotes continuous
ètale cohomology. By the Hochschild-Serre spectral sequence we obtain a
map
HS ◦ chr : Kr (U ⊗OM OM (ζpn ) ) → H r (M (ζpn ), H r−1 (U ⊗M M , Zp (r))).
Now V := H r−1 (U ⊗M M , Zp (r)) is a p-adic GM -representation and Fontaine’s
theory applies. By Fontaine’s ring BdR,M (here we just recall that it is a
filtered GM -module - see [19, §2] for more) one constructs the filtered M vector space DdR,M (V ) := (BdR,M ⊗Qp V )GM . Suppose that V is de Rham,
that is dimM (DdR,M (V )) = dimQp (V ). Then the dual exponential map
0
exp∗ : H 1 (M, V ) → DdR,M
(V ) is given by the composition of
inclusion
H 1 (M, V ) −−−−−−→
∼
=
−−−−→
ˆ ⊗ V)
H 1 (M, M
Qp
∼
=∪ log χcycl
ˆ ) ⊗ D0
0
H 1 (M, M
M
dR,M (V ) −−−−−−−→ DdR,M (V ),
ˆ ) and
ˆ ) → H 1 (M, M
where ∪ log χcycl is the Tate isomorphism M = H 0 (M, M
ˆ ⊗
the first isomorphism is induced by the Hodge-Tate decomposition M
Qp
ˆ (−i) ⊗ gr i D
∗
V ∼
= ⊕i∈Z M
M
dR,M (V ) [26, p.407] (for more on exp and how it
fits into a motivic Tamagawa number framework see [7]).
On the dlog side we need a Chern character into de Rham cohomology
(cohomology of differentials). When X = Spec(R) is noetherian it exists a
map dlog : Kq (Spec(R)) → ΩqR/Z satisfying dlog (a ∪ b) = dlog (a) ∧ dlog (b),
i (U/O ) := H i (X, Ω·
and other properties [26, p.393]. Denote HdR
M
X/OM (Z))
the hypercohomology of the de Rham complex of differentials on X with
logarithmic singularities over Z: it has a natural filtration. We recall that
1
0
(U/OM ) ⊗OM M.
(V ) = H 0 (X, Ω1X/OM (log Z)) ⊗OM M = F il1 HdR
DdR,M
For simplicity assume M = M0 , so that Ω1O
dlog ζpn and dM (ζpn )/M is pn (ζp − 1)−1 OM [ζpn ].
M (ζpn ) /OM
is generated by
Theorem 14. (Explicit reciprocity law) Take p > 2. Let X be a smooth
and proper curve over OM and take an affine U ⊂ X as above. Then the
16
FRANCESC BARS, IGNAZIO LONGHI
following diagram commutes:
⊗−1
lim (K2 (U ⊗ OM (ζpn ) )) ⊗ µpn
←−
n
1
lim H 0 (X ⊗ OM (ζpn ) , Ω2X⊗O
M (ζpn ) /OM
←−
n
1
−−−−−−→ lim H (M(ζpn ), Het (U ×M M , µpn ))
HS◦ch2
←−
n
?
?dlog
y
?
?
cor y
lim H 1 (M(ζpm ), H 1 (U ×M M , µpn ))
(log Z))(−1)
←−
n≥m
?
?
?=
y
lim Ω1O
←−
n
M (ζpn ) /OM
?
∼
=y
1
⊗ F il1 (HdR
(U/OM ))
H 1 (M(ζpm ), H 1 (U ×M M, Zp )(1))
?
?
y
1
lim OM (ζpn ) /dM (ζpn )/M ⊗OM F il1 HdR
(U/OM )
←−
n
?
?
exp∗ y
−
−−−−−
→
tr
1
M(ζpm ) ⊗OM F il1 HdR
(U/OM )
where tr denotes ( p1n trM (ζpn )/M (ζpm ) )n≥m .
Remark 15. Take X = P1 and U = A1 − {0} = Spec(OM [t, t−1 ]). Take
∗
OM
(un )n ∈ lim
(ζpn ) and consider {un , t} ∈ K2 (U ⊗ OM (ζpn ) ). Then from
←−
n
theorem 14 one recovers Iwasawa’s theorem 2 [26, Remark p.411]. See also
remark 13.
Remark 16. Theorem 14 has been vastly extended by Härkönen [16]: he
shows that, under some technical assumptions, the map exp∗ ◦corestriction◦
0
HS ◦ chr : lim
Kr (U ⊗OM M (εn )) → M (εm ) ⊗ DdR,M
(V ) can be computed
←−
n
in terms of a dlog map, for any r with 1 ≤ r ≤ p − 2 (as before ε is a fixed
generator of Tp Ff for a Lubin-Tate formal group Ff over OM ). We refer to
[16] for precise statements.
4.3. Rank 1 Drinfeld modules. Kato’s approach inspired the construction of the diagram presented in this subsection, which is the only new
material of this survey paper. As in the rest of this work, we restrict ourselves to the case of the Carlitz module; however, we remark that theorem
18 can easily be extended to any rank 1 sign-normalized Drinfeld module,
with exactly the same proof (mutatis mutandis: the changes are the same
necessary to pass from the function field results exposed in this paper to the
more general statements of [3]).
Remember that the Galois action on the module of Kähler differentials
ΩOΦ,n /O is given by σ(αdβ) = σ(α)dσ(β), σ ∈ Gal(KΦ,n /K), α, β ∈ OΦ,n .
In particular, one has
σ(dεn ) = d(σεn ) = d(Φχ(σ) (εn )) = χ(σ)dεn .
Besides, dεn = dΦπk (εn+k ) = π k dεn+k .
By lim ΩOΦ,n /O we denote the limit with respect to the trace map, defined
←
P
n
n (ω) :=
as usual by T rm
σ∈Gal(KΦ,n /KΦ,m ) σ(ω) where T rm = T rKΦ,n /KΦ,m .
RECIPROCITY LAWS FOR LOCAL FIELDS
17
Lemma 17. Let (ωn )n ∈ lim ΩOΦ,n /O and for each n choose xn ∈ On so
←
n (x ) converges to a limit
that ωn = xn dεn . Then the sequence yn := π −n T rm
n
in KΦ,m for any fixed m.
Proof. To lighten notation put Grn := Gal(KΦ,r /KΦ,n ), r > n. The diagram
Grn
?
?
≃y
0 −
−−−−→
−
−−−−→ Gal(KΦ,r /K) −
−−−−→ Gal(KΦ,n /K) −
−−−−→ 0
?
?
?
?
≃y
≃y
0 −
−−−−→ (1 + pn )/(1 + pr ) −
−−−−→
(A/pr )∗
−
−−−−→
(A/pn )∗
−
−−−−→ 0
pn
(where vertical maps are induced by χ) shows that χ(σ)−1 ∈
for σ ∈ Grn .
n+k
The equality ωn = T rn ωn+k can be rewritten as
X
xn π k dεn+k = xn dεn =
σ(xn+k )χ(σ)dεn+k ,
that is, π k xn =
σ∈Gn+k
n
P
σ(xn+k )χ(σ) + δn,n+k for some δn,n+k ∈ dKΦ,n+k /K . Let
X
zn,n+k :=
σ(xn+k )(χ(σ) − 1) + δn,n+k .
σ∈Gn+k
n
By [3, Lemma 3] we know v(δn,n+k ) > n+k−1; together with the observation
above, this implies that v(zn,n+k ) ≥ n. It is computed in [3, cor.4] that
n (a)) > v(a) + n − m − 1: applying it to
v(T rm
z
X
1
n,n+k
n
n
xn π k −
σ(xn+k ) = T rm
yn − yn+k = n+k T rm
π
π n+k
n+k
σ∈Gn
we see that v(yn − yn+k ) ≥ n − (m + k + 1), proving that the yn ’s form a
Cauchy sequence.
1
n ωn
By abuse of notation, we denote the limit in lemma 17 as lim n T rm
.
n→∞ π
dεn
Inspired by [20, §1.1] and [26, Theorem 3.3.15], we construct the following
diagram:
∗
lim KΦ,n
←
(2)
y
(1)
−−−−→
Homcont (mΦ,m , Tp Φ)
(4)y
dεn
(3)
⊕ Ω1OΦ,n /OF −−−−→ KΦ,m ≃ HomFp (Km,Φ , Fp)
p
εn
Arrow (1) is the Kummer map: it sends u = (un )n to a → lim (a, un )n .
lim Fp
←
n→∞
(This limit exists in Tp Φ:
√
ab
Φπ (a, un )n = ((un , KΦ,n
/KΦ,n ) − 1)Φπ ( πn a) = (a, un−1 )n−1
√
because Φπ ∈ O{τ } commutes with the action of GK ; here πn a is a root of
Φnπ (X) = a.)
18
FRANCESC BARS, IGNAZIO LONGHI
∗
i
As for (2), it is just dlog : u 7→ du
u , extended to KΦ,n by putting dlog εn :=
n
i dε
εn , as described in [3, §4.2.1]; the limit of differentials is taken with respect
dεn
to the trace and dε
ε denotes the inverse system εn .
The isomorphism KΦ,m ≃ HomFp (KΦ,m , Fp) is given by the trace pairing:
b ∈ KΦ,m is sent to a 7→ T rKΦ,m /Fp (ab) . The map (3) is
ωn
1
.
(ωn )n 7→ lim n T rKΦ,n /KΦ,m
n→∞ π
dεn
The definition of (4) needs more explanation. The logarithm λ has locally
n
→ mΦ,n for tn ≫ 0; this can be extended to ẽ : KΦ,n →
an inverse e : mtΦ,n
Fp ⊗Φ mΦ,n (the tensor product is taken on Ap, which acts on mΦ,n via Φ) by
putting ẽ(π i z) := π i ⊗ e(z) for i ≫ 0. In order to define (4), first remember
the isomorphism Tp Φ ≃ Ap, via a · ε ↔ a; then use composition with ẽ,
f 7→ f ◦ ẽ, to get Hom(mΦ,m , Ap) → Hom(KΦ,m , Fp) .
Theorem 18. The diagram above is commutative.
Proof. Working out definitions, one sees that this is equivalent to part 2 of
∗
theorem 6. More precisely (3) ◦ (2) sends u = (un ) ∈ lim KΦ,n
to the map
←
w
dlog
Col
(ε
)
.
w 7→ T rKΦ,m /K
u
m
πm
On the other hand, the image of u under (4) ◦ (1) is the morphism mapping π i z, v(z) ≫ 0, to π i gu (e(z)), where gu ∈ Hom(mΦ,m , Ap) is uniquely
determined by the condition gu (a) · εn = (a, un )n for all n ≥ m. Recalling
that
«
„
“ z
”
λ(e(z))
dlog
Col
(ε
)
·εn = T rKΦ,m /K m dlog Colu (εm ) ·εn ,
[e(z), un ]n := T rKΦ,n /K
u n
n
π
π
it is clear that (3) ◦ (2) = (4) ◦ (1) iff (·, ·) = [·, ·].
Remark 19. The similarity of our diagram with the first diagram of §4.2.2
is rather vague. It would be nice to express (and prove) the reciprocity
law in the cohomological setting, as in [26, §3.3]; the big problem here is
to find a good analogue of H 1 (GQp , Cp ) in characteristic p > 0 (a naive
approach cannot work: Y. Taguchi proved that H 1 (GK , Cp) = 0 ). Recent
developments in extending Fontaine’s theory to the equal characteristic case
(for a survey see [14]) might be helpful.
References
[1] Anglès, B.: On explicit reciprocity laws for the local Carlitz-Kummer symbols.
J. Number Theory 78 (1999), no. 2, 228–252.
[2] Artin, E., and Hasse, H.: Die beiden Ergänzungssätze zum Reziprozitätsgesetz
der ln -ten Potenzreste im Körper der ln -ten Einheitswurzeln. Abh. Math. Sem.
Univ. Hamburg 6 (1928), 146-162.
[3] Bars,
F. and Longhi,
I.:
Coleman’s power series and
Wiles’ reciprocity for Drinfeld modules. Corrected version at
http://mat.uab.cat/∼francesc/publicac.html
RECIPROCITY LAWS FOR LOCAL FIELDS
19
[4] Benois, D.: Périodes p-adiques et lois de réciprocité explicites. J.reine angew.
Math. 493 (1997), 115-151.
[5] Benois, D.: On Iwasawa theory of crystalline representations. Duke Math. J.
104 (2000), 211-267.
[6] Berrizbeitia, P.J.: An explicit reciprocity theorem in finite extensions of Qp .
Thesis, Massachussets Institute of Technology (1986).
[7] Bloch, S. and Kato, K.: L-functions and Tamagawa numbers of motives. In
P. Cartier et al. eds.: Grothendieck Festschrift Vol. I., Birkhäuser (1990).
[8] Cao, L.: Explicit reciprocity law for Lubin-Tate formal groups. Acta Math.
Sinica, english series 22, No.5 (2006), 1399-1412.
[9] Coleman, R.: Division values in local fields. Inven.Math. 53 (1979), 91-116.
[10] de Shalit, E.: Iwasawa Theory of Elliptic Curves with Complex Multiplication.
Persp. in Mathematics 3, Acad.Press, Boston, MA 1987.
[11] Drinfeld, V.G.: Elliptic modules. Math.USSR Sbornik 23 (1974), No.4, 561592.
[12] Fukaya, T.: Explicit reciprocity laws for p-divisible groups over higher dimensional local fields. J.Reine Angew. Math. 531 (2001), 61-119.
[13] Fukaya, T.: The theory of Coleman power series for K2 . Journal of
Alg.Geometry12 (2003), 1-80.
[14] Hartl, U.:A dictionary between fontaine-theory and its analogue in equal characteristic. Arxiv preprint.
[15] Hazewinkel, M.: Formal groups and applications. Academic Press, New York
(1978).
[16] Härkönen, H.: Explicit reciprocity laws and cohomology of log schemes
Preprint June 2007.
[17] Iwasawa, K.: On explicit formulas for the norm residue symbol. J. Math. Soc.
Japan 20 (1968), 151-164.
[18] Kato, K.: A generalization of local class field theory by using K-groups
I J.Fac.Sci.Univ.Tokyo26 (1979), 303-376; II ibid.27 (1980), 603-683; III
ibid.29 (1982), 31-43.
[19] Kato, K.: Lectures on the approach to Iwasawa theory for Hasse-Weil Lfunctions via BdR . In Arithmetic Algebraic Geometry (Trento, 1991), 50-163.
LNM 1553. Springer (1993).
[20] Kato, K.: Generalized explicit reciprocity laws. Adv. Stud. Contemp. Math.
(Pusan) 1 (1999), 57-126.
[21] Kummer, E.: Über die allgemeinen Reziprozitätsgesetze der Potenzreste. J.
Reine Angew. Math. 56 (1858), 270-279.
[22] Kurihara, M.: The exponential homomorphisms for the Milnor K-groups and
an explicit reciprocity law. J. Reine Angew. Math. 498 (1998), 201-221.
[23] Kurihara, M.: Kato’s higher local class field theory. In Invitation to higher
local fields., 53-60. Geometry and Topology Monographs Vol 3, 2000, International Press.
[24] Lang, S.: Cyclotomic Fields I and II. Graduate Texts in Mathematics 121.
Springer-Verlag, New York, 1990.
[25] Parshin, A.N.: Class field theory and algebraic K-theory. Uspekhi Mat.Nauk
30, no.1 (1975), 253-254, English transl.:Russian Math.Surv.
[26] Scholl, A.J.: An introduction to Kato’s Euler systems In Galois representations in Arithmetic Algebraic Geometry (A.J.Scholl and R.L.Taylor eds.),
379-460. London Math.Soc. Lecture Notes 254, 1998.
[27] Sen, S.: On explicit reciprocity laws. J.Reine Angew. Math.313 (1980), 1-26.
[28] Serre, J.P.: Local class field theory. In Algebraic Number Fields. J.W.S.Cassels
and A. Fröhlich, eds., Academic Press 1967
[29] Tate, J.: p-divisible groups. In Proc. of a conf. on local fields., 158-183. Driebergen 1966, Springer (1967).
20
FRANCESC BARS, IGNAZIO LONGHI
[30] Tsuji, T.: Explicit reciprocity law and formal moduli for Lubin-Tate formal
groups. J.reine angew. Math.569 (2004), 103-173.
[31] Vostokov, S.: Explicit formulas for the Hilbert symbol. In Invitation to higher
local fields., 81-89. Geometry and Topology Monographs Vol 3, 2000, International Press.
[32] Wiles, A.: Higher explicit reciprocity laws. Ann. of Math. 107 (1978), no. 2,
235-254.
[33] Zhukov, I.B.: Structure theorems for complete fields In Concerning the Hilbert
16th problem. Proceedings of ths St.Petersburg Mathematical Society., 175192. Amer. Math. Soc. Transl. (Ser. 2) 165, 1995.
Francesc Bars Cortina, Depart. Matemàtiques, Universitat Autònoma de
Barcelona, 08193 Bellaterra. Catalonia. Spain.
E-mail: francesc@mat.uab.cat
Ignazio Longhi, Dipartimento di Matematica “Federigo Enriques”, Università degli Studi di Milano. Via Cesare Saldini 50, 20133 Milan. Italy.
E-mail: longhi@mat.unimi.it