HAL Id: hal-00011266
https://hal.archives-ouvertes.fr/hal-00011266
Preprint submitted on 14 Oct 2005
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de
On some p-adic power series attached to the arithmetic of Q(ζ _p).
Bruno Angles
To cite this version:
Bruno Angles. On some p-adic power series attached to the arithmetic of Q(ζ_p).. 2005. �hal-
00011266�
ccsd-00011266, version 1 - 14 Oct 2005
On some p-adic power series attached to the arithmetic of Q (ζ p )
Bruno Angl`es Universit´e de Caen,
Laboratoire Nicolas Oresme, CNRS UMR 6139, Campus II, Boulevard Mar´echal Juin,
BP 5186, 14032 Caen Cedex, France.
E-mail: angles@math.unicaen.fr October 11 2005
Let p be a prime number, p ≥ 5. Let θ be an even and non-trivial character of Gal(Q(µ
p)/Q). Let f (T, θ) ∈ Z
p[[T ]] be the Iwasawa power series associated to the p-adic L-function L
p(s, θ) (see [12]), i.e.:
∀n ≥ 1, n ≡ 0 (mod p − 1), f ((1 + p)
1−n− 1, θ) = L(1 − n, θ), where L(s, θ) is the usual Dirichlet L-series. In 1979, in their celebrated article [4], Ferrero and Washington have proved that (see also [1]):
f (T, θ) 6≡ 0 (mod p).
Thus we can write:
f(T, θ) ≡ T
λ(θ)U ¯ (T ) (mod p),
where ¯ U(T ) ∈ F
p[[T ]]
∗. The Iwasawa lambda-invariant λ(θ) is not well- understood. By a heuristic argument due to Ferrero and Washington, one could expect that for sufficiently large p (see [5]):
X
θeven,θ6=1
λ(θ) ≤ Log(p)
Log(Log(p)) .
Furthermore, if p < 4000000, we have λ(θ) ≤ 1, and one would reasonably expect (see [8] ):
λ(θ) < p.
The only known result is due to Ferrero and Washington ([4]): for sufficiently large p we have:
λ(θ) ≤ p
Log(p)4ϕ(p−1)4. Now observe that:
f
′(T, θ) ≡ λ(θ)T
λ(θ)−1U ¯ (T ) + T
λ(θ)U ¯
′(T ) (mod p).
Thus if λ(θ) ≥ 1 and λ(θ) 6≡ 0 (mod p), we have that f
′(T, θ) 6≡ 0 (mod p).
The aim of this paper is to prove that for all θ even, θ 6= 1, we have (Corollary 4.4):
f
′(T, θ) 6≡ 0 (mod p).
This fact comes from properties of some power series that are connected to polynomials introduced by Mirimanoff at the beginning of the XXth century.
1 Notations
Let p be a prime number, p ≥ 5. Let Q
pbe an algebraic closure of Q
p. All the extensions of Q
pconsidered in this paper are contained in Q
p. Let v
pbe the p-adic valuation on Q
psuch that v
p(p) = 1. We denote the Iwasawa p-adic logarithm on Q
pby Log
p.
If A is a commutative ring, we denote the set of invertible elements of A by A
∗.
For every integer d, d ≥ 1, set µ
d= {z ∈ Q
p| z
d= 1}. If ρ ∈ ∪
d≥1µ
d, we denote the order of ρ by o(ρ). Set µ
p∞= ∪
n≥0µ
pn+1. For all n ≥ 0, let ζ
pn+1∈ µ
pn+1such that: ζ
p6= 1 and ∀n ≥ 0, ζ
ppn+2= ζ
pn+1.
Let:
- K
n= Q
p(µ
pn+1),
- O
n= Z
p[ζ
pn+1],
- π
n= ζ
pn+1− 1, - U
n= 1 + π
nO
n, - Γ
n= Gal(K
n/K
0), - ∆ = Gal(K
0/Q
p), - K
∞= Q
p(µ
p∞), - Γ = Gal(K
∞/K
0).
Let γ
0∈ Γ be such that ∀ζ ∈ µ
p∞, ζ
γ0= ζ
1+p. For a ∈ Z
∗p, we write:
a = ω(a) < a >,
where ω is the Teichm¨ uller character (i.e. ω(a) = lim
na
pn∈ µ
p−1) and
< a >≡ 1 (mod p).
Let n ≥ 0. For a ∈ Z
∗p, let σ
n(a) ∈ Gal(K
n/Q
p) be such that:
ζ
pσn+1n(a)= ζ
pan+1.
We have σ
n(a) = δ(a)γ
n(a), where δ(a) ∈ ∆ and γ
n(a) ∈ Γ
n. Observe that:
ζ
pγn+1n(a)= ζ
p<a>n+1. For a ∈ Z
∗p, let σ(a) ∈ Γ × ∆ be such that:
∀ζ ∈ µ
p∞, ζ
σ(a)= ζ
a.
Note that ∆ = {σ(ω(a)) | a = 1, · · · , p − 1}. For θ ∈ ∆ = Hom(∆, µ b
p−1), we set:
e
θ= 1 p − 1
X
δ∈∆
θ
−1(δ) δ ∈ Z
p[∆].
Let T be an indeterminate over Q
p. Let L be a finite extension of Q
p. Let O
Lbe the valuation ring of L. The restriction maps Res
n+1,n: Γ
n+1→ Γ
ninduce maps Res
n+1,n: O
L[Γ
n+1] → O
L[Γ
n]. Thus we can take the inverse limit of the group rings O
L[Γ
n] with respect to these maps, and by [12], Theorem 7.1,we have:
O
L[[T ]] ≃ lim
←
O
L[Γ
n],
where T corresponds to γ
0− 1. We set Λ
L= O
L[[T ]] and Λ = Λ
Qp. For all n ≥ 0, we set ω
n(T ) = (1 + T )
pn− 1. Recall that:
∀n ≥ 0, Λ
Lω
n(T )Λ
L≃ O
L[Γ
n].
Note that forl all n ≥ 0, N
Kn+1/Kn(U
n+1) ⊂ U
n, where N
Kn+1/Knis the norm map from K
n+1to K
n. We denote the inverse limit of the principal units U
nwith respect to the norm maps by U
∞. Note that U
∞is a Λ[∆]- module.
2 Logarithmic derivatives
Let u = (u
n)
n≥0be an element in U
∞, recall that:
∀n ≥ 0, N
Kn+1/Kn(u
n+1) = u
n.
There exists an unique element f
u(T ) ∈ Λ such that ([12], Theorem 13.38, see also [3]):
∀n ≥ 0, f
u(π
n) = u
n. Furthermore, if x = P
b
aσ(a) ∈ Z
p[Γ × ∆] and if u ∈ U
∞, then we have ([12], Lemma 13.48):
f
ux(T ) = Y
f
u((1 + T )
a− 1)
ba. For u ∈ U
∞and for n ≥ 0, we set:
D
n(u) = ζ
pn+1f
u′(π
n)
f
u(π
n) ∈ O
n.
We call the map D
n: U
∞→ O
nthe nth logarithmic derivative of Coleman.
Lemma 2.1
i) Let u ∈ U
∞, ∀n ≥ 0, D
n(u
ωn(T)) ≡ 0 (mod p
n+1).
ii) ∀θ ∈ ∆, b ∀u ∈ U
∞, ∀n ≥ 0, D
n(u
eθ) = e
θω−1D
n(u).
Proof i) We have:
f
uωn(T)(T ) = f
u((1 + T )
(1+p)pn− 1) f
u(T ) . Thus:
(1+T ) f
u′ωn(T)(T )
f
uωn(T)(T ) = (1+p)
pn(1+T )
(1+p)pnf
u′((1 + T )
(1+p)pn− 1)
f
u((1 + T )
(1+p)pn− 1) −(1+T ) f
u′(T )
f
u(T ) .
Note that:
(1 + π
n)
(1+p)pn− 1 = π
n. Therefore:
D
n(u
ωn(T)) = ζ
pn+1((1 + p)
pn− 1) f
u′(π
n)
f
u(π
n) ≡ 0 (mod p
n+1).
ii) Let a ∈ Z
∗pand let u ∈ U
∞. Observe that:
D
n(u
σ(a)) = aσ
n(a)(D
n(u)).
Let u ∈ U
∞, we get:
D
n(u
eθ) = 1 p − 1
X
p−1a=1
θ
−1(a)D
n(u
σ(ω(a))).
We have:
D
n(u
eθ) = 1 p − 1
X
p−1a=1
θ
−1(a) ω(a) σ
n(ω(a))(D
n(u)).
The Lemma follows. ♦
Proposition 2.2 Let θ ∈ ∆, θ b 6= 1, ω. The map D
ngives rise to a mor- phism of Z
p[Γ
n]-modules:
φ
θn: U
neθU
npn+1eθ→ e
θω−1O
np
n+1e
θω−1O
n,
where T acts on the right via (1 + p)(1 + T ) − 1.
Proof We have ([12], Theorem 13.54):
U
neθ≃ U
∞eθU
∞ωn(T)eθ.
Let v ∈ U
neθ, then there exists u = (u
n)
n≥0∈ U
∞eθsuch that v = u
n. We set:
φ
θn(v) ≡ D
n(u) (mod p
n+1).
By Lemma 2.1, we get a morphism of Z
p-modules:
φ
θn: U
neθU
npn+1eθ→ e
θω−1O
np
n+1e
θω−1O
n.
Let v ∈ U
neθ, and let u = (u
n)
n≥0∈ U
∞eθsuch that v = u
n. We have:
φ
θn(u
T) ≡ D
n(u
γ0−1) (mod p
n+1).
But:
D
n(u
γ0−1) ≡ ((1 + p)γ
0− 1)(D
n(u)) (mod p
n+1).
The Proposition follows. ♦
Theorem 2.3 For all n ≥ 0, set T
n= P
nd=0
ζ
pd+1. Let θ ∈ ∆, θ b 6= 1, ω.
Then:
∀n ≥ 0, Im(φ
θn) = Z
p[Γ
n]e
θω−1T
np
n+1e
θω−1O
n.
Proof Let Λ
nbe the maximal order of Q
p[Γ
n]. Then ([6], see also [7]):
O
n= Λ
n[∆] T
n. Now observe that p
n+1Λ
n⊂ pZ
p[Γ
n]. Thus:
p
n+1e
θω−1O
n⊂ p Z
p[Γ
n]e
θω−1T
n.
Recall that there exists α ∈ µ
p−1\ {1} such that ([12], Corollary 13.37):
∀n ≥ 0, U
neθ= ( α − ζ
pn+1ω(α − 1) )
Zp[Γn]eθ. For α ∈ µ
p−1\ {1}, set:
ρ
∞(α) = ( α − ζ
pn+1ω(α − 1) )
n≥0∈ U
∞. We have:
D
n(ρ
∞(α)) = −ζ
pn+1α − ζ
pn+1.
Thus Im(φ
θn) is the Z
p[Γ
n]-module generated by the e
θω−1α−ζ−ζpn+1pn+1
, α ∈ µ
p−1\ {1}. For α ∈ µ
p−1\ {1}, for all n ≥ 0, set:
u
n(α) =
pn+1
X
−1k=1, k6≡0 (modp)
α
kσ
n(k) ∈ Z
p[Γ
n× ∆].
We have:
Res
n+1,n(u
n+1(α)) =
pn+2
X
−1k=1, k6≡0 (modp)
α
kσ
n(k).
Thus :
Res
n+1,n(u
n+1(α)) =
pn+1
X
−1k=1, k6≡0 (modp)
σ
n(k) ( X
p−1ℓ=0
α
k+pn+1ℓ).
Finally, we get:
Res
n+1,n(u
n+1(α)) = u
n(α).
Now, we have:
α − 1
αζ
pn+1− 1 − 1 = (αζ
pn+1)
pn+1− 1 αζ
pn+1− 1 − 1.
Thus:
α − 1
αζ
pn+1− 1 − 1 =
pn+1−1
X
k=1
α
kζ
pkn+1.
Therefore:
α − 1
αζ
pn+1− 1 − 1 =
pn+1
X
−1k=1,k6≡0 (modp)
α
kσ
n(k)(ζ
pn+1) +
p
X
n−1k=1
α
kζ
pkn. Finally, we get:
α − 1
αζ
pn+1− 1 − 1 = u
n(α)(T
n).
Now set:
u
n(θ, α) =
pn+1
X
−1k=1, k6≡0 (modp)
θ(k)ω
−1(k)α
kγ
n(k).
We have:
(u
n(θ, α))
n≥0∈ Λ.
Furthermore, we have:
e
θω−1−ζ
pn+1α − ζ
pn+1= e
θω−11 α
−1ζ
pn+1− 1 . Thus:
e
θω−1−ζ
pn+1α − ζ
pn+1= 1
α
−1− 1 u
n(θ, α
−1)(e
θω−1T
n).
Now: X
α∈µp−1\{1}
u
0(θ, α) = (p − 1)θ(−1)ω
−1(−1) ∈ Z
∗p.
Thus there exists α ∈ µ
p−1\ {1} such that for all n ≥ 0, u
n(θ, α) ∈ Z
p[Γ
n]
∗. The theorem follows. ♦
We will also need the following Lemma:
Lemma 2.4 Let Λ
nbe the maximal order of Q
p[Γ
n]. Under the isomorphism Λ/ω
n(T )Λ ≃ Z
p[Γ
n], p
n+1Λ
ncorresponds to an ideal U
nof Λ such that ω
n(T )Λ ⊂ U
n⊂ (p, ω
n(T )) and lim
nU
n= {0}.
Proof Set:
e
0= 1 p
nX
γ∈Γn
γ, and for 1 ≤ d ≤ n, set:
e
d= X
χ∈Γcn, fχ=pd+1
e
χ,
where the sum is over all the characaters of Γ
nof conductors p
d+1and : e
χ= 1
p
nX
γ∈Γn
χ
−1(γ)γ.
We have:
p
n+1Λ
n= ⊕
nd=0Z
p[Γ
n]p
n+1e
d⊂ pZ
p[Γ
n].
Thus it is clear that ω
n(T )Λ ⊂ U
n⊂ (p, ω
n(T )). Now:
p
n+1e
0≡ p(
pn−1
X
ℓ=0
(1 + T )
ℓ) (mod ω
n(T )).
Thus:
p
n+1e
0≡ p ω
n(T )
T (mod ω
n(T )).
Let d be an integer, 1 ≤ d ≤ n. We have:
p
n+1e
d≡ p(
pn−1
X
ℓ=0
(1 + T )
ℓ( X
fχ=pd+1
χ(1 + p)
ℓ)) (mod ω
n(T )).
Thus:
p
n+1e
d≡ p(
p
X
n−1ℓ=0
T r
Kd−1/Qp(ζ
pℓd)(1 + T )
ℓ) (mod ω
n(T )).
Now recall that T r
Kd−1/Qp(ζ
pℓd) = 0 if v
p(ℓ) < d − 1. Therefore:
p
n+1e
d≡ p
d(
pn−d+1
X
−1ℓ=0
(1 + T )
ℓpd−1T r
K0/Qp(ζ
pℓ)) (mod ω
n(T )).
But:
pn−d+1
X
−1ℓ=0,ℓ≡0 (modp)
(1 + T )
ℓpd−1T r
K0/Qp(ζ
pℓ) = (p − 1) ω
n(T ) ω
d(T ) , and
pn−d+1−1
X
ℓ=0,ℓ6≡0 (modp)
(1 + T )
ℓpd−1T r
K0/Qp(ζ
pℓ) = ω
n(T )
ω
d(T ) − ω
n(T ) ω
d−1(T ) . Thus:
p
n+1e
d≡ p
d+1ω
n(T )
ω
d(T ) − p
dω
n(T )
ω
d−1(T ) (mod ω
n(T )).
The Lemma follows. ♦
3 Mirimanoff ’s power series
Recall that Mirimanoff has introduced the following polynomials in F
p[T ] :
∀j, 1 ≤ j ≤ p − 1, ϕ
j(T ) = X
p−1a=1
a
j−1T
a.
These polynomials have many beautiful properties and we refer the inter- ested reader to [9], [11] and [2]. In this section, we will introduce some power series that are related to these polynomials.
Let L be a finite extension of Q
p. Let π
Lbe a prime of L. Let θ ∈ ∆. b Let a ∈ O
Lsuch that a(a − 1) 6≡ 0 (mod π
L). For all n ≥ 0, we set:
M
n(θ, a) =
pn+1
X
−1k=1,k6≡0 (modp)
a
ka
pn+1− 1 θ(k)ω
−1(k)γ
n(k) ∈ O
L[Γ
n].
Lemma 3.1
(M
n(θ, a))
n≥0∈ lim
←
O
L[Γ
n].
Proof We must prove that:
∀n ≥ 0, Res
n+1,n(M
n+1(θ, a)) = M
n(θ, a).
Now observe that:
Res
n+1,n(M
n+1(θ, a)) =
pn+1−1
X
k=1,k6≡0 (modp)
θ(k)ω
−1(k)γ
n(k) 1 a
pn+2− 1 (
X
p−1ℓ=0
a
k+ℓpn+1).
The Lemma follows.♦
By the above Lemma, (M
n(θ, a))
n≥0corresponds to a power series M (T, θ, a) ∈ Λ
L. If θ = ω
j, 1 ≤ j ≤ p − 1, observe that:
M(0, θ, a) ≡ ϕ
j(a)
a
p− 1 (mod π
L).
Therefore we call M (T, θ, a) the Mirimanoff’s power series attached to θ and a.
Lemma 3.2
M (T, θ, a) = −θ(−1)ω
−1(−1)M (T, θ, a
−1).
Proof
We have:
M
n(θ, a) =
pn+1
X
−1k=1,k6≡0 (modp)
a
pn+1−ka
pn+1− 1 θ(p
n+1− k)ω
−1(p
n+1− k)γ
n(p
n+1− k).
Thus:
M
n(θ, a) = −
pn+1
X
−1k=1,k6≡0 (modp)
(a
−1)
k(a
−1)
pn+1− 1 θ(p
n+1−k)ω
−1(p
n+1−k)γ
n(p
n+1−k).
The Lemma follows. ♦ Theorem 3.3
M
′(T, θ, a) ≡ 0 (mod π
L) if and only if a ≡ −1 (mod π
L) and θ is odd.
Proof By Lemma 3.2, M (T, θ, −1) = 0 if θ is odd. Thus, if a ≡ −1 (mod π
L) and if θ is odd, we have M
′(T, θ, a) ≡ 0 (mod π
L).
The proof of this Theorem is based on Sinnott’s proof that the Iwasawa µ-invariant vanishes for cyclotomic Z
p-extensions of abelian number fields ([10]) as exposed in Washington’s book([12], paragraph 16.2).
Now, let’s suppose that M
′(T, θ, a) ≡ 0 (mod π
L). We have:
∀n ≥ 0, M (T, θ, a) ≡
pn+1
X
−1k=1,k6≡0 (modp)
a
ka
pn+1− 1 θ(k)ω
−1(k)(1+T )
i(k)(mod ω
n(T )), where i(k) = Log
p(k)/Log
p(1 + p). Thus, for all n ≥ 1, we have:
(1+T )M
′(T, θ, a) ≡
pn+1
X
−1k=1,k6≡0 (modp)
i(k) a
ka
pn+1− 1 θ(k)ω
−1(k)(1+T )
i(k)(mod (p
n, ω
n(T ))).
Therefore, for all n ≥ 1, we get:
pn+1
X
−1k=1,k6≡0 (modp)
i(k) a
ka
pn+1− 1 θ(k)ω
−1(k)(1 + T )
i(k)≡ 0 (mod (π
L, ω
n(T ))).
Recall that i(k) ≡ i(k
′) (mod p
n) if and only if
<k>−1p
≡
<k′p>−1(mod p
n).
Therefore changing i(k) to
<k>−1p
permutes exponents modulo p
nand do not affect divisibility by π
L. Thus:
∀n ≥ 1,
pn+1
X
−1k=1,k6≡0 (modp)
< k > −1 p
a
ka
pn+1− 1 θ(k)ω
−1(k)(1+T )
<k>p−1≡ 0 (mod (π
L, ω
n(T ))).
Let α ∈ µ
p−1. For n ≥ 1, set:
h
αn(t) ≡
pn+1
X
−1k=1,k≡α (modp)
α
−1k − 1 p
a
ka
pn+1− 1 θ(k)ω
−1(k)(1+T )
α−1k−1
p
(mod (p
n, ω
n(T ))).
Note that:
h
αn+1(T ) ≡ h
αn(T ) (mod (p
n, ω
n(T ))).
Now recall that:
Λ
L≃ lim
←
Λ
L(p
n, ω
n(T )) . Therefore, there exists h
α(T ) ∈ Λ
Lsuch that:
∀n ≥ 1, h
α(T ) ≡ h
αn(T ) (mod (p
n, ω
n(T ))).
Thus, we have: X
α∈µp−1
h
α(T ) ≡ 0 (mod π
L).
And also: X
α∈µp−1
(1 + T )h
α((1 + T )
p− 1) ≡ 0 (mod π
L).
Now, note that:
(1 + T )h
αn((1 + T )
p− 1) ≡ f
nα((1 + T )
α−1− 1) (mod (p
n, ω
n(T ))), where
f
nα(T ) ≡
pn+1
X
−1k=1,k≡α (modp)
α
−1k − 1 p
a
ka
pn+1− 1 θ(k)ω
−1(k)(1+T )
k(mod (p
n, ω
n(T ))).
For α ∈ µ
p−1, let s
0(α) ∈ {1, · · · , p − 1} such that α ≡ s
0(α) (mod p). We have:
f
nα(T ) ≡ θ(α)ω
−1(α)
p
X
n−1ℓ=0
α
−1(s
0(α) + pℓ) − 1 p
a
s0(α)+pℓa
pn+1− 1 (1+T )
s0(α)+pℓ(mod (p
n, ω
n(T ))).
But:
pn−1
X
ℓ=0
a
s0(α)+pℓa
pn+1− 1 (1 + T )
s0(α)+pℓ≡ a
s0(α)(1 + T )
s0(α)a
p(1 + T )
p− 1 (mod (p
n, ω
n(T ))),
and
p
X
n−1ℓ=0
ℓ a
s0(α)+pℓa
pn+1− 1 (1+T )
s0(α)+pℓ≡ −a
p(1+T )
pa
s0(α)(1 + T )
s0(α)(a
p(1 + T )
p− 1)
2(mod (p
n, ω
n(T ))).
Set:
f
α(T ) = θ(α)ω
−1(α) a
s0(α)(1 + T )
s0(α)a
p(1 + T )
p− 1 ( α
−1s
0(α) − 1
p − α
−1a
p(1 + T )
pa
p(1 + T )
p− 1 ).
Then:
∀n ≥ 1, f
nα(T ) ≡ f
α(T ) (mod (p
n, ω
n(T ))).
We have obtained:
X
α∈µp−1
f
α((1 + T )
α−1− 1) ≡ 0 (mod π
L).
Observe that for all α ∈ µ
p−1, f
α(T ) ∈ Λ
L∩L(T ). Thus, by Sinnott’s Lemma ([12], Lemma 16.9), for all α ∈ µ
p−1there exists c
α∈ O
Lsuch that:
f
α(T ) + f
−α((1 + T )
−1− 1) ≡ c
α(mod π
L).
Observe that:
f
α(T ) ≡ θ(α)ω
−1(α) a
s0(α)a
p− 1 ( α
−1s
0(α) − 1
p −α
−1a
pa
p− 1 )(1+T )
s0(α)(mod (π
L, T
p)), and
f
−α((1 + T )
−1− 1) ≡ θ(−α)ω
−1(−α)
apa−ps−10(α)(
−α−1(p−sp0(α))−1+ α
−1apa−1p)(1 + T )
s0(α)(mod (π
L, T
p)).
Thus, for all α ∈ µ
p−1, we must have:
a
s0(α)(
α−1s0p(α)−1−α
−1apa−1p) ≡ θ(−1)a
p−s0(α)(
−α−1(p−sp0(α))−1+α
−1apa−1p) (mod π
L).
For α = 1, we get:
a a
pa
p− 1 ≡ −θ(−1)a
p−1( a
pa
p− 1 − 1) (mod π
L).
Thus:
a
2≡ −θ(−1) (mod π
L).
We obtain a ≡ −1 (mod π
L) and θ is odd or θ is even and a
2≡ −1
(mod π
L). For the second case, if we consider all the equations obtained
when α runs through µ
p−1, we obtain that for all b ∈ {1, · · · , p − 1}, b even, we must have:
b ≡ ω(b) + p (mod p
2), and for all b ∈ {1, · · · , p − 1}, b odd, we must have:
b ≡ ω(b) (mod p
2),
This leads to a contradiction and the Theorem is proved. ♦ We will need the following Lemma:
lemma 3.4 There exists α ∈ µ
p−1such that for all prime numbers ℓ, ℓ ≡ α (mod p) and ℓ ≥ p
2, we have:
T r
Q(ζℓ)/Q( ζ
ℓp+1+ ζ
ℓp−1(ζ
ℓp− 1)
2) 6≡ 0 (mod p), where ζ
ℓis a primitive ℓth root of unity.
Proof For a ∈ Z, let [a]
ℓ∈ {0, · · · , ℓ − 1}, such that a ≡ [a]
ℓ(mod ℓ). Let Φ
ℓ(X) be the ℓth cyclotomic polynomial, then:
Φ
ℓ(1) = ℓ, Φ
′ℓ(1) = ℓ(ℓ − 1)
2 , and
Φ
′′ℓ(1) = ℓ(ℓ − 1)(ℓ − 2)
3 .
Now: Φ
′ℓ(X)
Φ
ℓ(X) = X
ρ∈µℓ\{1}
1 X − ρ ,
and Φ
′′ℓ(X)
Φ
ℓ(X) − ( Φ
′ℓ(X)
Φ
ℓ(X) )
2= − X
ρ∈µℓ\{1}
1 (X − ρ)
2. Therefore:
T r
Q(ζℓ)/Q( 1
ζ
ℓ− 1 ) = 1 − ℓ 2 , and
T r
Q(ζℓ)/Q( 1
(ζ
ℓ− 1)
2) = (ℓ − 1)
24 − (ℓ − 1)(ℓ − 2)
3 .
Furthermore, if a ∈ Z, a 6≡ 0 (mod ℓ), we have:
T r
Q(ζℓ)/Q( ζ
ℓaζ
ℓ− 1 ) = ℓ + 1
2 − [a]
ℓ. Set:
S = T r
Q(ζℓ)/Q( ζ
ℓp+1+ ζ
ℓp−1(ζ
ℓp− 1)
2).
Let m be the order of p modulo ℓ, set:
a = [1 + p
m−1]
ℓand b = [1 − p
m−1]
ℓ. We have:
S = T r
Q(ζℓ)/Q( ζ
ℓa+ ζ
ℓb(ζ
ℓ− 1)
2).
Since ℓ ≥ p
2, we have b ≥ 3, and thus:
a = ℓ + 2 − b.
Now:
S = T r
Q(ζℓ)/Q( ζ
ℓa− 1
(ζ
ℓp− 1)
2) + T r
Q(ζℓ)/Q( ζ
ℓb− 1
(ζ
ℓp− 1)
2) + 2T r
Q(ζℓ)/Q( 1 (ζ
ℓp− 1)
2).
We obtain:
S = −b
2+ b(ℓ + 2) − ℓ
2+ 6ℓ + 5
6 .
Therefore, S ≡ 0 (mod p) implies that:
ℓ
2+ 2
3 ∈ (F
p)
2. Now observe that the function field F
p(T,
q
T2+23
) has genus zero and thus has exactly p + 1 places of degree one. The Lemma follows. ♦
Theorem 3.5 Let θ ∈ ∆, θ b even. There exists α ∈ µ
p−1such that for all prime numbers ℓ, ℓ ≡ α (mod p) and ℓ ≥ p
2, we have:
X
ρ∈µℓ\{1}
M
′(T, θ, ρ) 6≡ 0 (mod p).
Proof
Let ℓ be a prime number, ℓ 6= p. Suppose that we have:
X
ρ∈µℓ\{1}
M
′(T, θ, ρ) ≡ 0 (mod p).
For α ∈ µ
p−1, set:
f
α(T ) = X
ρ∈µℓ\{1}
θ(α)ω
−1(α) ρ
s0(α)(1 + T )
s0(α)ρ
p(1 + T )
p− 1 ( α
−1s
0(α) − 1
p −α
−1ρ
p(1 + T )
pρ
p(1 + T )
p− 1 ).
By the proof of Theorem 3.3, we get:
X
α∈µp−1
f
α((1 + T )
α−1− 1) ≡ 0 (mod p).
Therefore by [12], Lemma 16.9, for all α ∈ µ
p−1there exists c
α∈ Z
psuch that:
f
α(T ) + f
−α((1 + T )
−1− 1) ≡ c
α(mod p).
But:
f
1(T ) ≡ −(1 + T )T r
Q(ζℓ)/Q( ζ
ℓp+1(ζ
ℓp− 1)
2) (mod (p, T
p)), and
f
−1((1 + T )
−1− 1) ≡ −(1 + T )T r
Q(ζℓ)/Q( ζ
ℓp−1(ζ
ℓp− 1)
2) (mod (p, T
p)).
Thus we get:
T r
Q(ζℓ)/Q( ζ
ℓp+1+ ζ
ℓp−1(ζ
ℓp− 1)
2) ≡ 0 (mod p).
It remains to apply Lemma 3.4. ♦
4 p-adic L-functions
Let θ ∈ ∆. b We set:
- f (T, θ) = 0 if θ is odd,
- if θ is even, f (T, θ) is the Iwasawa power series associated to the p-adic L-function L
p(s, θ) (see [12], paragraph 7.2.).
Recall that if θ 6= 1, f (T, θ) ∈ Λ, anf if θ is the trivial character then:
(1 − 1 + p
1 + T )f (T, θ) ∈ Λ
∗. For α ∈ µ
p−1, we set:
ρ
n(α) = α − ζ
pn+1ω(α − 1) ∈ U
n, and
ρ
∞(α) = (ρ
n(α))
n≥0∈ U
∞. We set:
η
n= Y
α∈µp−1\{1}
ρ
n(α)
α−1−1∈ U
n, and
η
∞= (η
n)
n≥0∈ U
∞. Lemma 4.1 Let θ ∈ ∆, θ b 6= 1, ω. Then:
U
∞eθ= (η
e∞θ)
Λ. Proof By the proof Theorem 2.3:
φ
θn(ρ
n(α)
eθ) ≡ M
n(θ, α
−1)e
θω−1T
n(mod p
n+1).
Thus:
φ
θn(η
neθ) ≡ ( X
α∈µp−1\{1}
(α − 1)M
n(θ, α))e
θω−1T
n(mod p
n+1).
But: X
α∈µp−1\{1}
(α − 1)M
0(θ, α) = (p − 1)θ(−1)ω
−1(−1) ∈ Z
∗p. The Lemma follows. ♦
Let θ ∈ ∆, θ b 6= 1, ω. We denote by G(T, θ) ∈ Λ
∗the power series that corresponds to (( P
α∈µp−1\{1}
(α − 1)M
n(θ, α)))
n≥0. Observe that:
X
α∈µp−1\{1}
(α−1)M
n(θ, α) = (p−1)θ(−1)ω
−1(−1)γ
n(p−1)
pn
X
ℓ=1,ℓ6≡0 (modp)
θ(ℓ)ω
−1(ℓ)γ
n(ℓ).
Let θ ∈ ∆, θ b 6= 1, θ even. By π
neθwe mean the unique (p − 1)
2th root in U
nwhich is congruent to 1 modulo π
nof:
p−1
Y
a=1
( ζ
pω(a)n+1− 1
ζ
pn+1− 1 )
(p−1)θ−1(a). We set:
π
∞eθ= (π
neθ)
n≥0∈ U
∞. Theorem 4.2
i) Let θ ∈ ∆, θ b even and non-trivial. Then:
(π
∞eθ)
−G(1+T1+p−1,θ)= (η
∞eθ)
f(1+T1+p−1,θ). ii) Let θ ∈ ∆, θ b 6= 1, ω. Then:
∀α ∈ µ
p−1\ {1}, (ρ
∞(α)
eθ)
G(1+T1+p−1,θ)= (η
∞eθ)
M(1+T1+p−1,θ,α−1). Proof Recall that:
φ
θn(ρ
n(α)
eθ) ≡ M
n(θ, α
−1)e
θω−1T
n(mod p
n+1), and
φ
θn(η
neθ) ≡ ( X
α∈µp−1\{1}
(α − 1)M
n(θ, α))e
θω−1T
n(mod p
n+1).
Thus ii) follows from Lemma 4.1, Lemma 2.4, and Proposition 2.2. Note that:
φ
θn(π
neθ) ≡ e
θω−11
ζ
pn+1− 1 (mod p
n+1).
Let f(X) =
XpnX+1−1−1, then:
(X − 1)Xf
′(X) + Xf (X) = p
n+1X
pn+1. Therefore:
1
ζ
pn+1− 1 = 1 p
n+1pn+1
X
−1k=1
kζ
pkn+1.
Let θ ∈ ∆, b for θ 6= 1, set:
v
n(θ) = − 1 p
n+1pn+1−1
X
k=1,k6≡0 (modp)
kθ(k)ω
−1(k)γ
n(k),
and if θ = 1, set:
v
n(θ) = −(1 − (1 + p)γ
n(1 + p)) 1 p
n+1pn+1−1
X
k=1,k6≡0 (modp)
kω
−1(k)γ
n(k).
Then by [12], paragraph 7.2, (v
n(θ))
n≥0corresponds to f (
1+T1−1, θ) if θ 6= 1, and to (1 − (1 + p)(1 + T ))f (
1+T1− 1, θ) if θ is trivial. Observe that if θ is even and non-trivial:
e
θω−11
ζ
pn+1− 1 = −v
n(θ)e
θω−1T
n, and if θ = 1 :
(1 − (1 + p)γ
n(1 + p))e
θω−11
ζ
pn+1− 1 = −v
n(θ)e
θω−1T
n. The Theorem follows. ♦
Theorem 4.3 Let d be an integer, d ≥ 2, d 6≡ 0 (mod p). Let θ ∈ ∆. b Then:
X
ρ∈µd, o(ρ)=d
M (T, θ, ρ) = −f ( 1
1 + T −1, θ)( X
ℓdivides d
ℓµ( d
ℓ )θ(ℓ)ω
−1(ℓ)(1+T )
Logp(ℓ) Logp(1+p)
),
wher µ(.) is the M¨obius function.
Proof If θ is odd, the result is clear by Lemma 3.2. Thus we assume that θ is even. Let Φ
d(X) be the dth cyclotomic polynomial, i.e.
Φ
d(X) = Y
ρ∈µd, o(ρ)=d
(X − ρ) ∈ Z[X].
Then:
Φ
d(X) = Y
ℓdivides d
(X
ℓ− 1)
µ(d/ℓ).
If we take logarithmic derivatives, we get:
Φ
′d(X)
Φ
d(X) = X
ℓdivides d
ℓµ( d
ℓ ) X
ℓ−1X
ℓ− 1 . It follows that:
X
ρ∈µd, o(ρ)=d
ζ
pn+1ζ
pn+1− ρ = X
ℓdivides d
ℓµ( d
ℓ ) ζ
pℓn+1ζ
pℓn+1− 1 . Thus:
X
ρ∈µd, o(ρ)=d
1
ρζ
pn+1− 1 = ( X
ℓdivides d
ℓµ( d
ℓ )σ
n(ℓ)) 1 ζ
pn+1− 1 . Now:
ρ
pn+1− 1
ρζ
pn+1− 1 − 1 =
pn+1−1
X
k=1
ρ
kζ
pkn+1. Thus:
1
ρζ
pn+1− 1 − 1 ρ
pn+1− 1 =
pn+1−1
X
k=1
ρ
kρ
pn+1− 1 ζ
pkn+1.
We are working in the extension Q
p(µ
d, ζ
pn+1)/ Q
p(µ
d) and we identify Gal( Q
p(µ
d, ζ
pn+1)/ Q
p(µ
d)) with Γ
n× ∆. Therefore:
e
θω−11
ρζ
pn+1− 1 = M
n(θ, ρ)e
θω−1ζ
pn+1+ e
θω−1(
p
X
n−1k=1
(ρ
p)
k(ρ
p)
pn− 1 ζ
pkn).
Now observe that the map µ
d→ µ
d, ρ 7→ ρ
p, is an isomorphism. Thus:
e
θω−1( X
ρ∈µd, o(ρ)=d
1
ρζ
pn+1− 1 ) = ( X
ρ∈µd, o(ρ)=d
M
n(θ, ρ))e
θω−1T
n.
But T
ngenerates a normal basis for the field extension K
n/Q
p, thus:
X
ρ∈µd, o(ρ)=d
M
n(θ, ρ) = −( X
ℓdivides d
ℓµ( d
ℓ )θ(ℓ)ω
−1(ℓ)γ
n(ℓ))v
n(θ),
where v
n(θ) is as in the proof of Theorem 4.2. The Theorem follows. ♦
Corollary 4.4 Let θ ∈ ∆, θ b even and non-trivial. Then:
f
′(T, θ) 6≡ 0 (mod p).
Proof Let α ∈ µ
p−1as in Theorem 3.5. Let ℓ be a prime number such that ℓ ≥ p
2and ℓ ≡ α (mod p
2) (note that there exist infinitely many such primes). We know that:
X
ρ∈µℓ\{1}
M
′(T, θ, ρ) 6≡ 0 (mod p).
But by Theorem 4.3:
X
ρ∈µℓ\{1}
M (T, θ, ρ) = −f( 1
1 + T − 1, θ)(ℓθ(ℓ)ω
−1(ℓ)(1 + T )
Logp(ℓ)
Logp(1+p)
− 1).
But since ℓ
p−1≡ 1 (mod p
2), we have:
Log
p(ℓ)
Log
p(1 + p) ≡ 0 (mod p).
Thus, if we take derivatives and reduce modulo p, we get:
X
ρ∈µℓ\{1}
M
′(T, θ, ρ) ≡ 1
(1 + T )
2f
′( 1
1 + T −1, θ)(ℓθ(ℓ)ω
−1(ℓ)(1+T )
Logp(ℓ)
Logp(1+p)
−1) (mod p).
The Corollary follows. ♦
The case of the trivial character is treated in the last section.
5 Other results
Let θ be an even Dirichlet character of conductor d or pd, where d ≥ 1, d 6≡ 0 (mod p). For all n ≥ 0, set: q
n= p
n+1d. Let g(T, θ) be the power series introduced in [12], paragraph 7.2, i.e.
g(T, θ) = T − q
01 + T f (T, θ),
where f (T, θ) is the power series associated to the p-adic L-function L
p(s, θ) (see [12], Theorem 7.10). Set L = Q
p(θ), and let π
Lbe a prime of L. Then, the Ferrero-Washington Theorem states (see [12], paragraph 16.2):
g(T, θ) 6≡ 0 (mod π
L).
We have:
Theorem 5.1
g
′(T, θ) 6≡ 0 (mod π
L).
Proof For y ∈ Q , set:
B(y) = (1 + q
0){y} − {(1 + q
0)y} − q
02 ∈ Z
p,
where {y} is the fractional part of y. recall that ([12], proof of Theorem 7.10):
∀n ≥ 0, g(T, θ) ≡
qn
X
a=1,(a,q0)=1
B( a
q
n) θω
−1(a) (1 + T )
−i(a)−1(mod ω
n(T )), where i(a) =
LogLogp(a)p(1+q0)
. Let’s suppose that g
′(T, θ) ≡ 0 (mod p). We have, for all n ≥ 1 :
(1+T )
2g
′(T, θ) ≡ − P
qna=1,(a,q0)=1
(i(a)+1) B(
qan
) θω
−1(a) (1+T )
−i(a)(mod (p
n, ω
n(T ))).
Therefore, we get:
∀n ≥ 1,
qn
X
a=1,(a,q0)=1
(i(a)+1) B( a q
n) θω
−1(a) (1+T )
i(a)≡ 0 (mod (π
L, ω
n(T ))).
Now, changing i(a) to (1 + q
0)
<a>−1p
permutes exponents modulo p
nand does not affect divisibility by π
L. Thus, for all n ≥ 1 :
P
qna=1,(a,q0)=1
((1+q
0)
<a>−1p
+1) B(
qan
) θω
−1(a) (1+T )
(1+q0)<a>−p 1≡ 0 (mod (π
L, ω
n(T ))).
For n ≥ 1 and for α ∈ µ
p−1, set:
H
αn(T ) ≡ P
a≡α (modp)
((1 + q
0)
α−1pa−1+ 1) B(
qan) θω
−1(a) (1 + T )
(1+q0)α−1a−1 p
(mod (p
n, ω
n(T ))).
Note that:
H
αn+1(T ) ≡ H
αn(T ) (mod (p
n, ω
n(T ))).
Thus, there exists H
α(T ) ∈ Λ
Lsuch that:
H
α(T ) ≡ H
αn(T ) (mod (p
n, ω
n(T ))).
We get: X
α∈µp−1
H
α(T ) ≡ 0 (mod π
L).
Therefore:
X
α∈µp−1
(1 + T )
1+q0H
α((1 + T )
p− 1) ≡ 0 (mod π
L).
Let f
α(T ) ∈ Λ
Las in [12], Lemma 16.8. Set:
F
α(T ) = α
−1p (1 + T )f
α′(T ) + (1 − 1 + q
0p )f
α(T ).
It is not difficult to see that F
α(T ) ∈ Λ
L∩ L(T ). By [12], Lemma 16.8, we have:
F
α(T ) ≡ P
a≡α (modp)
((1+q
0)
α−1pa−1+1) B(
qan) θω
−1(a) (1+T )
(1+q0)a(mod (p
n, ω
n(T ))).
Therefore, we get:
X
α∈µp−1
F
α((1 + T )
α−1− 1) ≡ 0 (mod π
L).
Now apply [12], Lemma 16.9, and we get that for all α ∈ µ
p−1, there exists b
α∈ O
Lsuch that:
F
α(T ) + F
−α((1 + T )
−1− 1) ≡ b
α(mod π
L).
But observe that:
F
α(T ) = F
−α((1 + T )
−1− 1).
Thus, for all α ∈ µ
p−1, there exists c
α∈ O
Lsuch that:
α
−1p (1 + T )f
α′(T ) + (1 − 1 + q
0p )f
α(T ) ≡ c
α(mod π
L).
Now we take α = 1, and we set:
G
1(T ) = (1+q
0) P
0<a<q0,a≡1 (modp)
θω
−1(a) (1+T )
a(1+q0)− P
0<a<q20+q0,a≡1 (modp)