Lecture Notes Part I
Uploaded by
nguoideptrynhatthegioiLecture Notes Part I
Uploaded by
nguoideptrynhatthegioiAN INTRODUCTION TO p-ADIC L-FUNCTIONS
by
Joaquín Rodrigues Jacinto & Chris Williams
Contents
General overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1. Motivation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2. The Riemann zeta function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3. p-adic L-functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4. Structure of the course. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5. Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Part I: The KubotaLeopoldt p-adic L-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2. Measures and Iwasawa algebras. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1. The Iwasawa algebra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2. p-adic analysis and Mahler transforms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3. An example: Dirac measures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4. A measure-theoretic toolbox. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3. The Kubota-Leopoldt p-adic L-function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1. The measure µa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2. Restriction to Z×p ................................................................... 19
3.3. Pseudo-measures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4. Interpolation at Dirichlet characters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1. Characters of p-power conductor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2. Non-trivial tame conductors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3. Analytic functions on Zp via the Mellin transform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4. The values at s = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5. The p-adic family of Eisenstein series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Part II: Iwasawa's main conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6. The Coleman map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.1. Notation and Coleman's theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2. Example: cyclotomic units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.3. Proof of Coleman's theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.4. The Coleman map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.5. The fundamental exact sequence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.6. The Kummer sequence, Euler systems and p-adic L-functions. . . . . . . . . . . . . . . . . . . . 45
7. The Main Conjecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.1. The Λ-modules arising from Galois theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.2. Measures on Galois groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.3. The main conjecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
7.4. Cyclotomic units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.5. On a theorem of Iwasawa. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.6. An application of class eld theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.7. Some consequences of Iwasawa theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8. Iwasawa's µ-invariant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8.1. Iwasawa's theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.2. Some consequences of Iwasawa's theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Appendix A. The complex class number formula. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Appendix B. Class eld theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Appendix C. Power series and Iwasawa algebras. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
General overview
In these lectures, we aim to give an introduction to p-adic L-functions and the foun-
dations of Iwasawa theory. Iwasawa's work found overarching structures that explained
previous results of Kummer linking special values of L-functions with the arithmetic of
cyclotomic elds. His methods have been applied with stunning success in other settings,
and much of what we know today about long-standing conjectures such as the Birch and
Swinnerton-Dyer conjecture has its roots in Iwasawa theory.
We will focus mainly on the construction and study of Kubota and Leopoldt's p-adic
Iwasawa
MW84
interpolation of the Riemann zeta function and on the ideas surrounding the
Main conjecture, now a theorem due to Mazur and Wiles (see [ ]). We will describe
some classical results linking L-values and arithmetic data that led to the study of p-adic
L-functions, and give several constructions of the p-adic zeta function. In particular, a
construction due to Coleman using cyclotomic units will naturally lead to the statement of
the Main conjecture, which we will prove when p is a Vandiver prime (which conjecturally,
at least, covers every prime). We will nally see a theorem of Iwasawa describing the growth
of the p-part of the class group of cyclotomic elds.
Recommended reading:
Lan90
The material in Part I of these notes is largely contained
Cyclotomic elds I and II (see [
Col
in Lang's ]), and is heavily based on Colmez's
CS06
beautiful lecture notes on the p-adic Riemann zeta function [ ] (in French). Part II is
based on the book Cyclotomic elds and zeta values by Coates and Sujatha (see [
Was97
]).
Part III is based on Washington's book An introduction to cyclotomic elds (see [ ],
especially 13). These lectures can be regarded as an introduction to the topics treated in
the references mentioned above, which the reader is urged to consult for further details, and
CS06 Lan90
as a prelude to Rubin's proof of the Main Conjecture using the theory of Euler systems, as
described in [ ] and [ , Appendix].
1. Introduction
1.1. Motivation.
1.1.1. Classical L-functions . The study of L-functions and their special values goes back
centuries, and they are central objects of modern number theory. Examples include:
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 3
The famous Riemann zeta function, de ned by
X Y
ζ(s) = n−s = (1 − p−s )−1 , s ∈ C,
n≥1 p
where the last product runs over all prime numbers p and the second equality follows is
a conseqence of the unique factorisation theorem. The sum converges absolutely for the
real part of s greater than 1, makingζ a holomorphic function in a right half-plane. The
expression as a product is called an Euler product.
Let F be a number eld. The zeta function of F is
X Y
ζF (s) ..= N (I)−s = (1 − N (p)−s )−1 ,
0̸=I⊂OF p
where the sum is over all non-zero ideals in the ring of integers (and which again converges
for Re(s) > 1), and the product is over all non-zero prime ideals of K. The existence of the
Euler product again follows from unique factorisation.
Let χ : (Z/N Z)× → C× be a Dirichlet character, and extend it to a function χ:Z→C
by setting it to be 0 at integers not coprime to N . The L-function of χ is
X Y
L(χ, s) ..= χ(n)n−s = (1 − χ(p)p−s )−1 .
n≥1 p
Yet again, this converges in a right half-plane.
Let E/Q be an elliptic curve of conductor N. L-function
One can de ne an
X Y Y
L(E, s) ..= an (E)n−s = (1 − ap (E)p−s + p1−2s )−1 Lp (s),
n≥1 p∤N p|N
where ap (E) = p + 1 − #E(Fp ) and the factors Lp (s) at bad primes are de ned as Lp (s) = 1
(resp.(1 − p−s )−1 , resp. (1 + p−s )−1 ) if E has bad additive (resp. split multiplicative, resp.
non-split multiplicative) reduction at p.
P+∞ n
Let f = n=1 an (p)q ∈ Sk (Γ0 (N ), ωf ) be a (normalised) newform of weight k , level
N and character ωf . The L-function associated to f is given by
X Y Y
L(f, s) ..= an (f )n−s = (1 − ap (f )p−s + ωf (p)pk−1−2s )−1 (1 − ap (f )p−s )−1 .
n≥1 p∤N p|N
Any reasonably behaved L-function should have the following basic properties (which
(1)
may be non-trivial to prove!) :
(1) A meromorphic continuation to the whole complex plane;
(2) A functional equation relating s and k−s for some k ∈ R;
(3) An Euler product.
Remark 1.1 . More generally, let GQ = Gal(Q/Q) denote the absolute Galois group of
Q and let V ∈ RepL GQ be a p-adic Galois representation (i.e. a nite dimensional vector
space over a nite extension L of Qp equipped with a continuous linear action of GQ ). For
ℓ ̸= p a rational prime, one de nes a local factor at ℓ as
Lℓ (V, s) ..= det(Id − Frob−1
ℓ ℓ
−s
|V Iℓ )−1 ,
where Frobℓ denotes the arithmetic Frobenius at ℓ, and Iℓ denotes the inertia group at ℓ.
One de nes a local factor at p as
Lp (V, s) ..= det(Id − φ−1 p−s |Dcris (V ))−1 ,
(1) We will treat the meromorphic continuation of the Riemann zeta function in the sequel.
4 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
where this time Dcris (V ) denotes the crystalline module of V |GQp from p-adic Hodge theory
and φ denotes the crystalline Frobenius. One then de nes the global L-function of V as the
formal product
Y
L(V, s) = Lℓ (V, s).
ℓ
(2)
When V is the representation attached to am arithmetic object the L-function of the
representation is typically equal to the L-function attached to that object; for example,
GQ acting through the character χ
Bel09
taking V = Qp (χ) (that is, V = Qp with via class
eld theory), one recovers the Dirichlet L-functions described above. See [ ] for an
introduction to these topics.
1.1.2. Special values and arithmetic data . Much of the interest of L-functions comes
through their special values. There are deep results and conjectures relating special values
of L-functions to important arithmetic information, of which a prototypical example is the
following:
Theorem 1.2 (Class number formula). Let F be a number eld with r1 real embed-
dings, r2 pairs of complex embeddings, w roots of unity, discriminant D, and regulator R.
The zeta function ζF has a simple pole at s = 1 with residue
2r1 (2π)r2 R
ress=1 ζF (s) = p hF ,
w |D|
where hF is the class number of F .
In theory, this gives a method for calculating the class number of a number eld, although
in general computing the regulator is di cult. In special cases related to cyclotomic elds,
though, it can give an eective computation of the class number.
A second famous example of links between special values of L-functions and arithmetic
information comes in the form of the Birch and SwinnertonDyer conjecture. Let E/Q be
an elliptic curve. The set of rational points E(Q) forms a nitely generated abelian group,
and Birch and SwinnertonDyer predicted that
rankZ E(Q) = ords=1 L(E, s).
Let's say we want to attack the conjecture. There are two natural subquestions:
(a) We could try to prove that rankZ E(Q) ≥ ords=1 L(E, s). A natural approach is to
try to construct rational points on the elliptic curve. The theory of Heegner points is based
StarkHeegner points has been used
Dar01
on such an idea. More recently, the p-adic theory of
with some success (see [ ], where the theory was initiated). These constructions tend
to give points of in nite order on E(Q) if and only if the L-function vanishes to a certain
order (for example, a Heegner point has in nite order if and only if the order of vanishing
is precisely 1).
(b) Conversely, we could try and prove that rank E(Q) ≤ ords=1 L(E, s). In this case we
bound Euler systems
Rub00
want to the number of points. One method for trying to do this uses
(see [ ] for a comprehensive introduction). The primary application of the theory of
Euler systems is in bounding certain Galois cohomology groups, known as Selmer groups,
which are de ned using local behaviour and can be viewed as a cohomological interpretation
of the group of rational points on E . The dierence between the Selmer group and E(Q)
is captured in the TateShafarevich group X(E/Q), which is a torsion abelian group that
is conjecturally nite. If the p-part of X(E/Q) is nite, then the p-Selmer group and
the group E(Q) have the same rank (as abelian groups), so bounding the Selmer group is
equivalent to bounding E(Q).
(2) For example, a number eld, an elliptic curve, a modular form, etc.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 5
The ideas above have led to the only known special cases of the conjecture; in particular,
Kol88 GZ86 MM91
we now know it to be true (under some assumptions) when ords=1 L(E, s) ≤ 1 due to work
of Kolyvagin, GrossZagier and MurtyMurty (see [ ], [ ] and [ ]).
Remark 1.3 . Something that both (a) and (b) have in common is their use of p-adic L-
functions.
MTT86
p-adic version of Birch and Swinnerton-Dyer, due to Mazur, Tate
In fact, there is a
and Teitelbaum (see [ ]), which relates the rank of the p-Selmer group to the order of
vanishing of a p-adic L-function at s = 1. This formulation (which is conjecturally equivalent
) has recently been proved in many cases by SkinnerUrban
SU14 Kat04
(3)
to the standard formulation
(see [ ]), following work of Kato (see [ ]). Their proof uses a version of the
Iwasawa Main Conjecture for elliptic curves.
1.1.3. Iwasawa's main conjecture: a general picture . As it has been mentioned, many
arithmetic objects have associated Galois representations, and these Galois representations
come with L-functions. In general, one might expect that where there is an L-function,
there is a p-adic L-function, and that there is a version of the Main conjecture for this
p-adic L-function. The main conjecture essentially says that the p-adic L-function should
control the size of the Selmer groups of the Galois representation it is attached to (recalling
from above that the Selmer groups are subgroups of the Galois cohomology de ned by local
conditions).
{Galois representations} / {L-functions}
{Galois cohomology} o IMC
{p-adic L-functions}
In the case of elliptic curves, the application of this to Birch and SwinnertonDyer comes
through the links between the Selmer groups and E(Q).
In this lecture course, we will focus on the simplest example of the above picture, namely
the Main conjecture for the Riemann zeta function, as formulated by Iwasawa himself. In
the process, we will construct the p-adic analogue of the zeta function on the way to stating
the main conjecture, which we will prove for a special case. In doing so, we hope to give an
introduction to the rich area of p-adic L-functions and Iwasawa theory.
1.2. The Riemann zeta function. Since the Riemann zeta function will be a central
player in the rest of these lectures, we take a brief detour to describe some of the classical
(4)
theory surrounding it. We start with the following general result
Theorem 1.4 . Let f : R −→ R be a C ∞ -function such that f, f ′ , f ′′ , . . . (i.e. f and
≥0
all of its derivatives) all decay exponentially at in nity, and let
Z ∞
Γ(s) = e−t ts−1 dt.
0
be the usual Gamma function. The function
Z ∞
1
..
L(f, s) = f (t)ts−1 dt, s ∈ C,
Γ(s) 0
(3) Toshow that p-adic Birch and SwinnertonDyer implies classical Birch and SwinnertonDyer, one must
show niteness of X and prove a precise relation between the order of vanishing of the classical and p-adic
L-functions. Both of these are, naturally, extremely di cult open problems.
(4) We thank Keith Conrad for pointing out an error in an earlier draft of these notes. We previously only
assumed f to have exponential decay at in nity, but in the proof, we actually use that f and all of its
derivatives are decay exponentially at in nity.
6 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
which converges to a holomorphic function for Re(s) > 0, has an analytic continuation to
the whole complex plane, and
dn
L(f, −n) = (−1)n f (0).
dtn
We call L(f, s) the Mellin transform of f .
Proof. To show analytic continuation, we claim that when Re(s) > 1, we have
L(f, s) = −L(f ′ , s + 1),
where f ′ = df /dt. This is an easy exercise in integration by parts, using the identity
Γ(s) = (s − 1)Γ(s − 1), and gives the analytic continuation to all of C by iteration. Finally,
iterating the same identity n + 1 times shows that
L(f, −n) = (−1)n+1 L(f (n+1) , 1)
Z ∞
= (−1)n+1 f (n+1) (t)dt
0
n (n)
= (−1) f (0)
from the fundamental theorem of calculus, giving the result.
Now we pick a speci c choice of f, namely, we let
t X tn
f (t) = = Bn ,
et −1 n!
n≥0
the generating function for the Bernoulli numbers Bn .
Remark 1.5 . The Bernoulli numbers are highly combinatorial, and satisfy recurrence
relations that ensures they are rational numbers; for example, the rst few are
1 1 1
B0 = 1, B1 = − , B2 = , B3 = 0, B4 = − , ...
2 6 30
For k≥3 odd, Bk = 0.
(5)
We want to plug this function into Theorem 1.4, and for this, we require :
Lemma 1.6 . The function f (t) and all of its derivatives decay exponentially at in nity.
Proof. For t > 0, we may expand f (t) as a geometric series
f (t) = t(e−t + e−2t + e−3t + · · · )
=: tF (t).
Note that f ′ (t) = F (t) + tF ′ (t), and f ′′ (t) = 2F ′ (t) + tF ′′ (t); arguing inductively we see
f (n) (t) = nF (n−1) (t) + tF (n) (t)
= n(e−t + e−2t + · · · ) + (−1)n t(e−t + 2n e−2t + 3n e−3t + · · · )
∼ (−1)n te−t as t → ∞.
This decays exponentially.
Lemma 1.7 . For the choice of f as above, we have
(s − 1)ζ(s) = L(f, s − 1).
(5) Again, we thank Keith Conrad for pointing out this elegant proof.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 7
Proof. We use the classical formula for Γ(s) above. Substituting t for nt and rearranging,
we obtain Z ∞
1
n−s = e−nt ts−1 dt.
Γ(s) 0
Now, when Re(s) is su ciently large, we can write
1 X ∞ −nt s−1
Z Z ∞ÅX ã
X 1
ζ(s) = n−s = e t dt = e−nt t · ts−2 dt,
Γ(s) 0 Γ(s) 0
n≥1 n≥1 n≥1
and the result now follows from the identity
1 X
= e−nt .
et −1
n≥1
From the theorem above, we immediately obtain:
Corollary 1.8 . For n ≥ 0, we have
Bn+1
ζ(−n) = −
,
n+1
In particular, ζ(−n) ∈ Q for n ≥ 0, and ζ(−n) = 0 if n ≥ 2 is even.
1.3. p-adic L-functions. We have already seen two examples where special values of L-
functions should be able tell us information about arithmetic objects. In fact, there are very
general conjectures (for example, the BlochKato and Beilinson conjectures) predicting that
these links exist on a much wider scale, and despite some partial results in special cases these
conjectures remain deep open problems. Much of what we do know about these conjectures
comes through the theory of p-adic L-functions. In this section, we explain what a p-adic
L-function is and the properties it should satisfy.
1.3.1. p-adic L-functions, a rst idea . The complex ζ -function is a function
ζ : C −→ C
with complex analytic properties which is rational at negative integers. Since Z is a common
subset of both C and Zp ⊆ Cp , and since they are dense in Zp , it is natural to ask if there
exists a function
ζp : Zp −→ Cp
that is `p-adic analytic' (in some sense to be de ned) and which is uniquely characterized
by the property that it agrees with the complex L-function at negative integers in the sense
that
ζp (1 − n) = (∗) · ζ(1 − n),
for some explicit factor (∗). We would say that such a function `p-adically interpolates the
special values of ζ(s)'.
1.3.2. Ideles, measures and Tate's thesis . In practice, there is no
(6)
single analytic func-
tion on Zp that interpolates all of the special values , as we will explain in Section 4.3.
Tat50 Iwa52
Instead, a better way of thinking about L-functions is to use a viewpoint initiated by Tate
in his thesis [ ] (and later independently by Iwasawa; see [ ]). This viewpoint
sees L-functions as measures on ideles, and allows one to package together all Dirichlet
L-functions, including the Riemann zeta function, into a single object. We will give a
brief account of the classical theory here, but for fuller accounts, one should consult the
references above.
We begin with the following observations.
(6) Rather,
there are p − 1 dierent analytic functions ζp,1 , ..., ζp,p−1 on Zp , and ζp,i interpolates only the
values ζ(−k) for which k ≡ i (mod p − 1).
8 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Lemma 1.9 . (i) Let χ : (Z/N Z)× −→ C× be a Dirichlet character. Then χ can be
seen as a function Y
χ: Z× ×
ℓ −→ C .
ℓ prime
There exists an identi cation of C with Homcts (R>0 , C× ) by sending s to the char-
(ii)
acter x 7→ xs .
Proof. To see part (i), suppose that N = ℓn is a power of some prime ℓ; then we can see
×
χ as a function on Zℓ via the identi cation
Z× ∼ n × n
ℓ = (Z/ℓ Z) × (1 + ℓ Zℓ ).
The general case follows from the Chinese remainder theorem.
We turn to part (ii). For s ∈ C, the function x 7→ xs is visibly a continuous character on
R>0 . We want to show that all such characters are of this form. After taking a logarithm,
this is equivalent to showing that all continuous homomorphisms (of additive groups) g :
R → C are of the form g(x) = xg(1), which is easily shown by directly computing the values
of g on Q and extending by continuity.
By the identi cation of C with Homcts (R>0 , C× ) one can view ζ as a function
ζ : Homcts (R>0 , C× ) −→ C
x 7→ xs 7−→ ζ(s).
But we can add in Dirichlet characters using the following.
Proposition 1.10 . Under the identi cations above, each pair (χ, s), where χ is a Dirich-
let character and s ∈ C, corresponds to a (unique) continuous character
Y
κχ,s : R>0 × Z×
ℓ −→ C
×
ℓ prime
(x, y) 7−→ xs χ(y),
where we equip the source with the product topology. All continuous characters on this group
are of this form.
Proof. The rst assertion is immediate from above. To see the converse, let κ be such
a character. Then we already know that the restriction of κ to R>0 must be of the form
x 7→ xs . Furthermore, we have an isomorphism of topological groups
∼
Y
Z× ×
ℓ = lim(Z/M Z) ,
←−
ℓ prime
where the right hand side is equipped with the pro nite topology, and by taking a su ciently
small open neighbourhood of 1 in C× we see that any continuous character κ′ from this to
×
C must have open kernel. Hence the kernel has nite index, and κ descends to the ( nite)
quotient, which one can check is of the form (Z/N Z)× for some N , giving rise to a Dirichlet
character χ of conductor N. Then κ = κχ,s .
The product space is more usually written as follows.
De nition 1.11 . ideles A×
De ne the
Y′
of Q to be
A× = A× .. ×
Q =R × Q×ℓ
ℓ prime
= (xR , x2 , x3 , x5 , ...) : xℓ ∈ Z×
ℓ for all but nitely many ℓ .
(The prime on the product denotes restricted product, which indicates the almost everywhere
integral property in the de nition). It's a good exercise to check that:
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 9
Proposition 1.12 (Strong approximation). YThere is a decomposition
A× ∼
= Q× × R>0 × Z×
ℓ .
ℓ prime
Hence all continuous characters
Q× \A× −→ C×
are of the form κχ,s as above, where χ is a Dirichlet character and s ∈ C.
Now we can consider all Dirichlet L-functions at once via the function
L : Homcts (Q× \A× , C× ) −→ C
κχ,s 7−→ L(χ, s).
In the framework of Tate, this function L can be viewed as integrating κχ,s against the
Haar measure on Q× \A× . In his thesis, Tate showed that properties such as the analytic
continuation and functional equations of Dirichlet L-functions by using harmonic analysis
on measures. Indeed, the idelic formulation gives a beautiful conceptual explanation for the
appearance of the Γ-functions and powers of 2πi in the functional equation of the zeta func-
tion; these factors are the `Euler factors at the archimedean place'. The measure-theoretic
perspective has proven to be a powerful method of de ning and studying automorphic L-
functions in wide generality.
Remark 1.13 .
IfK is any number eld, one can analogously de ne the ideles A×
K of K as the restricted
×
K × \A× ×
Q
product v Kv over all places v of K . Continuous homomorphisms K → C are
called Hecke characters or Gröÿencharacters. They are examples of automorphic forms for
GL1 /K .
× ×
By class eld theory, the idele class group K \AK injects (with dense image)
into GK
ab ..
= Gal(K ab /K), where K ab denotes the maximal abelian extension of K .
Since any character of GK must factor through GK , under this identi cation we have
ab
×
Homcts (K \AK , C ) = Homcts (GK , C ) = Homcts (GK , GL1 (C)), where GK = Gal(K/K)
× × ×
denotes the absolute Galois group of Q. We can then package Dirichlet L-functions over K
into a complex analytic function on the space of one dimensional complex representations
of the absolute Galois group GK .
1.3.3. p-adic L-functions via measures p-adic version of this picture, a
. To obtain a
natural thing to do is to look at continuous characters from Q× \A× into C× (rather than
×
Q × p
C ). Again, such a function corresponds to a function on R>0 × Zℓ . Since R>0 is
connected and Cp is totally disconnected, the restriction of any such character to R>0 is
Q ×
trivial. Also using topological arguments we nd that the restriction to ℓ̸=p Zℓ factors
through a nite quotient, so gives rise to some Dirichlet character of conductor prime to p.
×
This leaves the restriction to Zp , which is by far the most interesting part.
In the measure-theoretic viewpoint of L-functions, it is then natural to look for an ana-
(7)
lytic function
ζp : Homcts (Z× ×
p , Cp ) −→ Cp
in such a way that
ζp (x 7→ xk ) = (∗) · ζ(1 − k), k≥1
since Z×
(7) Precisely, × ×
p = µp−1 × (1 + pZp ), the space Homcts (Zp , Cp ) can be identi ed with p − 1 copies of
the open unit ball in Cp (see the exercises). It carries the structure of a rigid analytic p-adic space, and a
function on this space is rigid analytic if it can be written as a convergent power series on each ball. Such
an analytic function will be a measure if these coe cients are bounded.
10 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
for an explicit factor (∗), that is, for a function on p-adic characters interpolating the values
ζ(−k) for k ≥ 0. We say such a function is a measure on Z× p . In equivalent and more
elementary terms, a measure on Z×p is an element of the continuous dual of the space of
×
continuous functions on Zp . We will prove:
Theorem 1.14 (Kubota-Leopoldt, Iwasawa). There exists a (pseudo-)measure (8)
ζp on Z×
p such that, for all k > 0,
Z
xk · ζp ..= ζp (x 7→ xk ) = 1 − pk−1 ζ(1 − k).
Z×
.
p
Remark 1.15 Note that we removed the Euler factor at p. This is a general phe-
nomenon appearing in the theory of p-adic L-functions.
From such an object, we can build the (meromorphic) functions on Zp we were initially
looking for. But now, we have much more, and the power of the measure-theoretic approach
becomes obvious:
Theorem 1.16 . Let χ be a Dirichlet character of conductor p , n ≥ 0, viewed as a
n
locally constant character on p.
Then, for all k > 0,
Z×
Z
χ(x)xk · ζp = 1 − χ(p)pk−1 L(χ, 1 − k).
Z×
p
In other words, when viewed as a measure the KubotaLeopoldt p-adic L-function is a
single p-adic gadget that encodes the special values not only of the Riemann zeta function,
but also of all of its twists by characters of p-power conductor. This is pretty magic!
Indeed, even though one only uses the values ζ(−k) to construct the measure ζp , Theorem
1.16 a rms that its values at in nitely many dierent points are still related to the complex
L-function. We will also see a formula of Leopoldt showing another striking resemblance
when evaluating at the character x 7→ χ(x).
To complete the image given in 1.3.2, one can take into account Dirichlet characters of
conductor prime to p. The ideas that go into the proof of Theorem 1.16 can also be used to
show:
Theorem 1.17 .
Let D > 1 be any integer coprime to p, and let η denote a (primitive)
Dirichlet character of conductor D. There exists a unique measure ζη on Z× p with the
following interpolation property: for all primitive Dirichlet characters χ with conductor pn
for some n ≥ 0, we have, for all k > 0,
Z
χ(x)xk · ζη = 1 − χη(p)pk−1 L(χη, 1 − k).
Z×
.
p
Remark 1.18 Let (Z/DZ)×∧ (Z/DZ)× . The mea-
denote the space of characters on
× × ×∧
sures given by Theorem 1.17 can be seen as functions on Homcts (Zp , Cp ) × (Z/DZ) and
×∧ ×∧
they are compatible with respect to the natural maps (Z/DZ) → (Z/EZ) for E|D .
This shows that they de ne a function on
×∧ Y ∧
Homcts (Z× ×
p , Cp ) × lim Z/DZ = Homcts (Z× ×
p , Cp ) × Z×
←− ℓ
(D,p)=1 ℓ̸=p
= Homcts (Q ×
\A× ×
Q , Cp ).
In other words, they give a measure on the idele class group of Q.
(8) Pseudo-measures will be de ned in Section 3. Roughly speaking, such an object is a measure that is
allowed to have simple poles.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 11
Remark 1.19 . The measure-theoretic interpretation of p-adic L-functions also allows
us to generalise to number elds in a clean and conceptual way, as we elaborate in this
remark.
Let F∞ = Q(µp∞ ) denote the eld extension of Q obtained by adjoining all p-power
roots of unity. This is a Galois extension of Q with Gal(F∞ /Q) ∼
= Z×
p via the cyclotomic
character (see, for example, the notation at the start of Part II). Under this isomorphism,
we can see ζp as a pseudo-measure on Gal(F∞ /Q).
Note that F∞ is the maximal abelian extension of Q that is unrami ed outside p.
Indeed, the Kronecker-Weber theorem states that if K/Q is abelian, then K ⊂ Q(µm ) for
some minimal m. By the rami cation properties of cyclotomic elds, if a prime ℓ rami es
in K , then ℓ|m, and hence if K is unrami ed outside p, there exists some n such that
K ⊂ Q(µpn ) = Fn ⊂ F∞ .
Now let K/Q be a number eld; then the p-adic analogue of the zeta function ζK (s)
ab,p
should be a pseudo-measure on Gal(K /K), where K ab,p is the maximal abelian extension
of K unrami ed outside primes above p. This is also the natural setting for the construction
of p-adic L-functions of other arithmetic objects, such as elliptic curves or modular forms over
K . It is possible to translate measures on this Galois group into measures on (OK ⊗ Zp )×
or analytic functions on OK ⊗ Zp , but this is not as clean; over Q, things work nicely
since the class number is 1 and the totally positive global units are trivial. For a general
number eld K, the strong approximation theorem takes a more complicated form, and we
Loe14
end up with a collection of measures/analytic functions indexed by a class group. For an
BSW19
example of the theory for modular forms over imaginary quadratic elds, see [ ] (for
measures/distributions) or [ , 3] (for analytic functions).
1.4. Structure of the course. The course will be split into three major parts. In Part
I, we construct the p-adic analogue of the Riemann zeta function, called the Kubota-Leopoldt
p-adic L-function, L-functions. In
and prove that it interpolates special values of Dirichlet
Part II, we will look at the Main conjecture. In particular, we'll explore some of the deeper
underlying structure discovered by Iwasawa, proving a connection between certain units in
cyclotomic elds and the p-adic L-function. This will naturally lead to the statement of the
Main conjecture, which we will prove in a special case. In Part III, we will look at further
topics in Iwasawa theory, including the λ- and µ-invariants of a Zp -extension. The latter
theory allows precise control of the p-part of the ideal class group in a tower of p-extensions
of number elds.
1.5. Acknowledgements. We are grateful to Keith Conrad and Luis Santiago for
pointing out a number of corrections on an earlier draft of these notes. Further comments
and corrections are welcome!
12 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Part I: The KubotaLeopoldt p-adic L-function
In this part, we give a construction of the KubotaLeopoldt p-adic L-function and the p-
adic L-functions of Dirichlet characters. In Section 2, we introduce the necessary formalism
of p-adic measures and Iwasawa algebras, and show that there is an isomorphism from the
Iwasawa algebra of Zp to the space Zp [[T ]]Zp , given by the Mahler
of power series over
×
transform. In Section 3, we construct a pseudo-measure on Zp that interpolates the values
of the Riemann zeta function at negative integers. In Section 4, we show moreover that
this pseudo-measure interpolates the values L(χ, −k) for χ a Dirichlet character of p-power
conductor. Further, if η is a Dirichlet character of conductor prime to p, we construct a
measure on Zp that interpolates the values L(χη, −k) as χ runs over Dirichlet characters of
p-power conductor. Finally, in Section 4.3 we rephrase the construction in terms of analytic
functions on Zp via the Mellin transform.
2. Measures and Iwasawa algebras
In the introduction, we explained that a natural way to construct p-adic L-functions is
to construct suitable p-adic measures on Z×
p. In this section, we introduce the formalism of
the theory of p-adic analysis that we will be using in the sequel. Whilst some of the results
of this section may appear a little dry in isolation, uency in the measure-theoretic language
will greatly help us simplify calculations that would otherwise be very technical.
2.1. The Iwasawa algebra. We x a nite extension L of Qp , equipped with the
p-adic valuation normalized such that vp (p) = 1; this will serve as the coe cient eld. We
write OL for its ring of integers. Let G be a pro nite abelian group (e.g. G = Zp or G = Zp ,
×
which are the examples of most interest to us).
We denote by C (G, L) the space of continuous functions ϕ : G → L, equipped with the
valuation vC (ϕ) = inf x∈G vp (ϕ(x)) (giving rise to the sup norm). This valuation makes
C (G, L) into an L-Banach space, i.e. a complete topological L-vector space whose topology
is de ned by a valuation vC satisfying
(i) vC (f ) = +∞ if and only if f = 0;
(ii) vC (f + g) ≥ min(vC (f ), vC (g)) for all f, g ∈ C (G, L);
(iii) and vC (λf ) = vp (λ) + vC (f ) for all λ ∈ L, f ∈ C (G, L).
De nition 2.1 .
We de ne the space M (G, L) of L-valued measures on G as the dual
Homcts (C (G, L), L) equipped with the strong topology. If ϕ ∈ C (G, L) and µ ∈ M (G, L),
the evaluation of µ at ϕ will be denoted by
Z
ϕ(x) · µ(x),
G
R
or by
G
ϕ·µ if the variable of integration is clear from the context (in the literature, this
µ ∈ M (G, L)
R
is sometimes written alternatively as
G
ϕ · dµ). We say that an element
is an OL -valued measure, and write µ ∈ M (G, OL ), if µ takes values inOL . Since G is
compact and measures are continuous (or, equivalently, bounded), we have that M (G, L) =
M (G, OL ) ⊗OL L. We will be mainly concerned with OL -valued functions and measures.
Remark 2.2 . We can think of measures as additive functions
µ : {compact open subsets of G} −→ OL .
Indeed, let µ be such a function and let ϕ ∈ C (G, OL ). We will see how to integrate ϕ
against µ. Assume rst that ϕ is locally constant. Then there is some open subgroup H of
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 13
G such that ϕ can be viewed as a function on G/H . We de ne the integral of ϕ against µ
to be
Z X
ϕ · µ ..= ϕ(a)µ(aH).
G [a]∈G/H
In the general case, we can write ϕ = limn→∞ ϕn , where each ϕn is locally constant. Then
we can de ne
Z Z
..
ϕ · µ = lim ϕn · µ,
G n→∞ G
which exists and is independent of the choice of ϕn we take. This de nes an element in
M (G, OL ).
Conversely, ifµ ∈ M (G, OL ) and U ⊂ G is an open compact set, one de nes µ(U ) ..=
R
1
G U
(x) · µ(x) , where 1U (x) denotes the characteristic function of U .
Proposition 2.3 . We have an isomorphism
M (G, OL ) ∼
= lim O [G/H],
←− L
H
where the limit is over all open subgroups of G.
Proof. Let µ be a measure, and let H be an open subgroup of G. We de ne an element
λH of OL [G/H] by setting
X
λH ..= µ(aH)[a].
[a]∈G/H
By the additivity property of µ, we see that (λH )H ∈ lim OL [G/H], so we have a map from
←−
measures to this inverse limit.
Conversely, given such an element λ of the inverse limit, write λH for its image in OL [G/H]
under the natural projection. Then
X
λH = ca [a].
[a]∈G/H
We de ne
µ(aH) = ca .
Since the λH are compatible under projection maps, this de nes an additive function on the
open compact subgroups of G, i.e. an element µ ∈ M (G, OL ).
De nition 2.4 . We de ne the Iwasawa algebra of G to be
Λ(G) ..= lim
←− L
O [G/H].
H
(Note that we suppress L from the notation).
Remark 2.5 . The Iwasawa algebra Λ(Zp )OL -algebra structure, and
has a natural
hence by transport of structure we obtain such a structure on M (Zp , OL ). It turns out that
the algebra structure can be described directly via convolution of measures. For general G,
given two measures µ, λ ∈ M (G, OL ), one de nes their convolution µ ∗ λ to be
Z Z ÅZ ã
ϕ · (µ ∗ λ) = ϕ(x + y) · λ(y) · µ(x).
G G G
One checks that this does give an algebra structure and that the isomorphism above is an
isomorphism of OL -algebras.
14 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
2.2. p-adic analysis and Mahler transforms. In this section we establish a link
between p Z
-adic measures on p and power series.
De nition 2.6. x∈Z Forn ≥ 1, p and let
Ç å
x ..
x(x − 1) · · · (x − n + 1)
= ,
n n!
Ç å
x
and let = 1.
0
Remark 2.7
x
. One easily checks that x 7→ x
n
de nes an element in C (Zp , Zp ) of valu-
ation vC n = 0.
x (9)
The following theorem says that the functions
n form an orthonormal basis for the
space C (Zp , L) and is fundamental in everything that follows.
Theorem 2.8 (Mahler). Let ϕ : Z p → L be a continuous function. There exists a
unique expansion Ç å
X x
ϕ(x) = an (ϕ) ,
n
n≥0
where an (ϕ) ∈ L and an → 0 as n → ∞. Moreover, vC (ϕ) = inf n∈N vp (an (ϕ)).
Proof. Col10a
See [ , Théorème 1.2.3.].
Remark 2.9. The coe cients an (ϕ) are called the Mahler coe cients of ϕ. One can
write down the Mahler coe cients of ϕ very simply; we de ne the discrete derivatives of ϕ
by
ϕ[0] = ϕ, ϕ[k+1] (x) = ϕ[k] (x + 1) − ϕ[k] (x),
and then an (ϕ) = ϕ[n] (0).
De nition 2.10 . Let µ ∈ Λ(Zp ) be a p-adic measure on Zp . De ne the Mahler trans-
form (or Amice transform ) of µ to be
Z ñ Ç å ô
X Z x
Aµ (T ) =
.. x
(1 + T ) · µ(x) = · µ T n ∈ OL [[T ]].
Zp Zp n
. The Mahler transform gives an O -algebra isomorphism
n≥0
Theorem 2.11 L
Λ(Zp ) −∼→ OL [[T ]].
Proof. We can explicitly de ne an inverse to the transform. Let g(T ) = n≥0 cn T n ∈
P
OL [[T ]]. Let H ⊂ Zp be an open subgroup, and for each [a] ∈ Zp /H let 1aH denote the
characteristic function of the coset aH ⊂ Zp . This is a continuous function on Zp , and hence
has a Mahler expansion
Ç å
X x
1aH (x) = a[a]
n ,
n
n≥0
[a]
with an ∈ OL . Then de ne
X
µ[a] ..= a[a]
n cn ,
n≥0
and
X
µH = µ[a] [a].
[a]∈Zp /H
(9) If
B is an L-Banach space, an orthonormal basis of B is a collection (ei )i∈I such that (ai )i∈I 7→ i∈I ai ei
P
de nes an isometry between ℓ0∞ (I, L) and B , where ℓ0∞ (I, L) is the set of sequences in L indexed by I that
tend to 0 (in a sense that depends on I ). One can show that every L-Banach space B with valuation vB
such that vB (B) = vp (L) admits an orthonormal basis.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 15
It is an easy check that (µH )H is an element of the Iwasawa algebra and the resulting
function OL [[T ]] → Λ(Zp ) is an inverse to the Mahler transform.
De nition 2.12 . If g ∈ OL [[T ]], we write µg ∈ Λ(Zp ) for the corresponding (OL -valued)
measure on Zp (so that Aµg = g).
Remark 2.13 . Let g ∈ OL [[T ]] with associated measure µg . From the de nitions, it is
evident that
Z
µg = g(0).
Zp
2.3. An example: Dirac measures. In this section, we give an example of the above
theory in action via Dirac measures.
De nition 2.14 . Let a ∈ Zp . The Dirac measure δa ∈ M (Zp , OL ) is the linear
functional `evaluation at a', that is, the measure de ned by
δa : C (Zp , OL ) −→ OL
ϕ 7−→ ϕ(a).
Under the identi cation of measures with additive functions on open compact subsets of
Zp , we nd that this corresponds to the function
®
1 if a∈X
δea (X) =
0 if a∈
/ X,
as can be seen directly from the proof of the identi cation.
As an element of the Iwasawa algebra, again from the proof we nd that at nite level
δa corresponds to the basis element [a + pn Zp ] ∈ OL [Zp /pn Zp ], which in the inverse limit
we denote by [a].
Finally, we compute the Mahler transform of δa . If a ∈ Z×
p then, by de nition, this is
Ç å
X a
Aδa (T ) = T n = (1 + T )a .
n
n≥0
2.4. A measure-theoretic toolbox. There are natural operations one might consider
on measures, and via the Mellin transform these give rise to operators on power series. As
such, the following operations can be considered as a `toolbox' for working with measures
and power series, and as we shall see in the sequel, the ability to manipulate measures in
Col10a
this way has important consequences. For further details (and even more operations), see
[ ].
2.4.1. Multiplication by x . Given a measure µ on Zp , it's natural to wish to evaluate it
at the function xk for a positive integer k. To allow us to do that, we de ne xµ to be the
measure de ned by
Z Z
f (x) · xµ = xf (x) · µ.
Zp Zp
We can ask what this operation does on Mahler transforms; we nd:
Lemma 2.15 . We have
Axµ = ∂Aµ ,
where ∂ denotes the dierential operator (1 + T ) dT
d
.
16 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. We have
Ç å Ç å Ç å Ç å Ç å
x x x x x
x = (x − n) +n = (n + 1) +n .
n n n n+1 n
The result follows directly.
From the above lemma and Remark 2.13, we immediately obtain:
Corollary 2.16 . For µ ∈ Λ(ZZ ), we havep
xk · µ = ∂ k Aµ (0).
Zp
2.4.2. Multiplication by z
x
x
. Let z be such that |z − 1| < 1. Then the Mahler transform
of z µ is
Azx µ (T ) = Aµ (1 + T )z − 1 .
Indeed, from the de nition of the Mahler transform, we see that
Z
x
Aµ (1 + T )z − 1 =
(1 + T )z · µ,
Zp
and this is precisely the Mahler transform of zxµ (one has to be slightly careful about
convergence issues).
2.4.3. Restriction to open compact subsets . Consider an open compact subset X ⊂ Zp .
If we de ne 1X to be the characteristic function of this subset, we can consider the restriction
ResX (µ) of µ to X de ned by
Z Z
f · ResX (µ) ..= f 1X · µ.
X Zp
In the case X = b + pn Zp , we can write this characteristic function explicitly as
1 X x−b
1b+pn Zp (x) = ζ ,
pn
ζ∈µp
n
and then using the above, we calculate the Mahler transform of Resb+pn Zp (µ) to be
1 X −b
AResb+pn Zp (µ) (T ) = ζ Aµ (1 + T )ζ − 1 .
pn
2.4.4. Restriction to Z
ζ∈µp
n
×
p . From the above applied to b=0 and n = 1, it is immediate
that
1 X
ARes (T ) = Aµ (T ) − Aµ (1 + T )ζ − 1 .
× (µ)
(2.1)
Zp p
2.4.5. The action of Z , φ and ψ
ζ∈µp
×
p . We introduce an action of Z×
p that serves as a
×
precursor to a Galois action later on. Let a∈ Zp . We can de ne a measure σa (µ) by
Z Z
f (x) · σa (µ) = f (ax) · µ.
Zp Zp
This has Mahler transform
Aσa (µ) = Aµ (1 + T )a − 1 .
In a similar manner, we can de ne an operator φ acting as `σp ' by
Z Z
f (x) · φ(µ) = f (px) · µ,
Zp Zp
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 17
Aφ(µ) = φ(Aµ ) ..= Aµ (1 + T )p − 1 .
and this corresponds to Finally, we also de ne the
analogous operator for p−1 ; we de ne a measure ψ(µ) on Zp by de ning
Z Z
f (x) · ψ(µ) = f (p−1 x) · µ.
Zp pZp
Note that ψ ◦ φ = id, whilst
φ ◦ ψ(µ) = RespZp (µ).
In particular, we have
ResZ×
p
(µ) = (1 − φ ◦ ψ)(µ). (2.2)
The operator ψ also gives an operator on OL [[T ]] under the Amice transform, and using the
restriction formula above, we see that it is the unique operator satisfying
1 X
φ ◦ ψ(F )(T ) = F (1 + T )ζ − 1 .
p
ζ∈µp
The following result will be useful in Part II.
Corollary 2.17 . Let µ ∈ Λ(Z ) be a measure. Then µ is supported on Z
p
×
p if and only
if ψ(Aµ ) = 0.
Proof. The operator φ is easily seen to be injective. We have an injection ι : Λ(Z×
p ) ,→
Λ(Zp ) given by
Z Z
ϕ · ι(µ) = ϕ Z×
· µ,
p
Zp Z×
p
and as ResZ×p
◦ ι is the identity on Λ(Z× ×
p ), we can identify Λ(Zp ) with its image as a subset
×
of Λ(Zp ). If µ ∈ Λ(Zp ), then µ ∈ Λ(Zp ) if and only if ResZ× (µ) = µ, or equivalently if and
p
only if Aµ = Aµ − φ ◦ ψ(Aµ ), which happens if and only if ψ(Aµ ) = 0.
Remark 2.18 . Whilst we identify Λ(Z×
p) with a subset of Λ(Zp ), it is important to
remark that it is not a subalgebra. Indeed, convolution of measures on Z×
p uses the mul-
tiplicative group structure whilst convolution of measures on Zp uses the additive group
structure, so if λ and µ are two measures on Z×
p, we do not have µ ∗Z× λ = µ ∗Zp λ.
p
Remark 2.19 . Power series rings have been generalized to what now are called Fontaine
rings. It turns out that Galois representations are connected to certain modules over these
rings called (φ, Γ)-modules. The operations described above are examples of the basic op-
erations we have on (φ, Γ)-modules and their interpretation with p-adic analysis inspired
Col10b
the proof of the p-adic Langlands correspondence for GL2 (Qp ). For further details, see
[ ].
Remark 2.20 . In this section, we have seen one of the major bene ts of passing to
inverse limits, and considering measures on Zp (rather than just Z/pn Z): namely, the con-
nection to power series rings. This has profound consequences later in the story, when
considering modules over the algebra Λ(Zp ). In particular, passing to the inverse limit
`rigidi es' the picture considerably, and as a consequence, there is a structure theorem for
nitely generated Λ(Zp )-modules similar to that for modules over a principal ideal domain.
This structure theorem does not hold for the nite level algebras Λ(Z/pn Z), and it is used
crucially in even stating the Iwasawa main conjecture. We describe the structure theorem
in C.
18 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
3. The Kubota-Leopoldt p-adic L-function
3.1. The measure µ . a Recall the results of the introduction: we can write the Riemann
zeta function in the form
Z ∞
1
(s − 1)ζ(s) = f (t)ts−2 dt,
Γ(s − 1) 0
where f (t) = t/(et − 1), and that ζ(−k) = (dk f /dtk )(0) = (−1)k Bk+1 /(k + 1). We want
to remove the smoothing factor at s = 1. For this, let a be an integer coprime to p and
consider the related function
1 a
fa (t) = − .
et − 1 eat − 1
This is also C∞ and has exponential decay at in nity, so we can consider the function
L(fa , s) as in the introduction. The presence of a removes the factor of s − 1, at the cost of
introducing a dierent smoothing factor:
Lemma 3.1 . We have
L(fa , s) = (1 − a1−s )ζ(s),
which has an analytic continuation to C, and
fa(k) (0) = (−1)k (1 − a1+k )ζ(−k).
We now introduce the p-adic theory from above. Note the following very simple observa-
tion.
Lemma 3.2 . (i) Under the substitution et = T + 1, the derivative d/dt becomes the
operator ∂ = (1 + . d
T ) dT
(ii) In particular, if we de ne
1 a
Fa (T ) ..= − ,
T (1 + T )a − 1
we have
fa(k) (0) = ∂ k Fa (0).
The right hand side in (ii) should look familiar as the expression in Corollary 2.16, which
expressed the integral of the function xk over Zp in terms of its Mahler transform. So, if
Fa (T ) can be written as an element of Zp [[T ]], then we will get a measure on Zp that sees
values of the Riemann zeta function.
Proposition 3.3 . The function F (T ) is an element of Z [[T ]].
a p
Proof. We can expand
Ç å
a
X a
T n = aT 1 + T g(T ) ,
(1 + T ) − 1 =
n
n≥1
1 a n−2
P
where g(T ) = n≥2 a n T has coe cients in Zp since we have chosen a coprime to p.
Hence, expanding the geometric series, we nd
1 a 1 X
− = (−T )n g(T )n ,
T (1 + T )a − 1 T
n≥1
which is visibly an element of Zp [[T ]].
De nition 3.4 . Let µa denote the measure on Zp corresponding to Fa (T ) under the
Mahler transform.
We have proved:
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 19
Proposition 3.5 . For Zn ≥ 0, we have
xk · µa = (−1)k (1 − ak+1 )ζ(−k).
Zp
3.2. Restriction to Z . ×
p Recall from the introduction that we want the
×
p-adic analogue
of the Riemann zeta function to be a measure on Zp , not all of Zp . We have already de ned
a restriction operator in equation (2.2), which on Mahler transforms acts as 1 − φ ◦ ψ.
Proposition 3.6 . ZWe have
xk · µa = (−1)k (1 − pk )(1 − ak+1 )ζ(−k).
Z×
p
(In other words, restricting to Z×
p removes the Euler factor at p).
Proof. We rst show that ψ(µa ) = µa by consider the action on power series. Indeed, we
have by de nition
Å ã
1 X 1
(φ ◦ ψ) = p−1
T (1 + T )ζ − 1
ζ p =1
Å ã
1 1
= p
= φ ,
(1 + T ) − 1 T
as can be seen by calculating the partial fraction expansion. By injectivity of φ, we deduce
that ψ( T1 ) = 1
T and hence ψ(µa ) = µa since ψ commutes with the action of Z×
p.
Since ResZ×
p
= 1 − φ ◦ ψ , we deduce that
Z Z Z Z
xk · µ a = xk · (1 − φ ◦ ψ)µa = xk · (1 − φ)µa = (1 − pk ) xk · µ a ,
Z×
p Zp Zp Zp
as required.
3.3. Pseudo-measures. It remains to remove the dependence on a. To do so, we need
to introduce the notion of a pseudo-measure. The Riemann zeta function has a simple pole
at s = 1, and pseudo-measures let us take this into account on the p-adic side. (Thus far,
the presence of a has acted as a `smoothing factor' which removes this pole).
De nition 3.7 . Let G be an abelian pro nite group, and let Q(G) denote the fraction
eld of the Iwasawa algebra Λ(G). A pseudo-measure on G is an element λ ∈ Q(G) such
that
([g] − [1])λ ∈ Λ(G)
for all g ∈ G.
The following lemma shows that a pseudo-measure µ on Z×
p is uniquely determined by
k
R
the values
Z×
x ·µ for k > 0.
.
p
Lemma 3.8 (i) Let µ ∈ Λ(Z×
p ) such that
Z
xk · µ = 0
Z×
p
for all k > 0. Then µ = 0.
(ii) Let µ ∈ Λ(Zp ) such that
×
Z
xk · µ ̸= 0
Z×
p
for all k > 0. Then µ is not a zero divisor in Λ(Z× p ).
(iii) Part (i) holds if, more generally, µ is a pseudo-measure.
20 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. To prove part (i), note that the vanishing condition forces the Mahler transform
of µ to be constant, since each non-trivial binomial polynomial is divisible by x. But as µ
×
is a measure on Zp , it vanishes under ψ , so must be zero.
For part (ii), suppose there exists a measure λ such that µλ = 0. Then
Z ÅZ ãÅ Z ã
0= xk · (µλ) = xk · µ xk · λ ,
Z×
p Z×
p Z×
p
which forces λ=0 by part (i).
Finally, let µ be a pseudo-measure satisfying the vanishing condition. Let a ̸= 1 be an
integer prime to p; then there is a natural measure
[a] − [1] ∈ Λ(Z×
p ),
λ = ([a] − [1])µ ∈ Λ(Z×
R
with
Z×
p
f · ([a] − [1]) = f (a) − f (1). Consider the measure p ). Then
λ satis es the condition of part (i), so λ = 0. But [a] − [1] satis es the condition of part (ii),
so it is not a zero-divisor, and this forces µ = 0, as required.
De nition 3.9
×
.
Let a be an integer that is prime to p, and let θa denote the element
of Λ(Zp ) corresponding to [a] − [1]. Note that
Z
xk · θ a = a k − 1
Z×
p
from the de nitions. There is a discrepancy between this and the power of a that appears
in the interpolation formula; to account for this, note that on Z×
p, we have a well-de ned
operation `multiplication by x−1 ' given by
Z Z
f (x) · x−1 µ ..= x−1 f (x) · µ,
Z×
p Z×
p
and that Z
xk · xµa = (−1)k (ak − 1)(1 − pk−1 )ζ(1 − k).
Z×
p
Finally de ne
x−1 ResZ× µa
..
ζp = p
∈ Q(Z×
p ).
θa
Proposition 3.10 . Suppose a p−1
̸= 1 (mod p2 ). Then the element ζp is a well-de ned
pseudo-measure that is:
(i) independent of the choice of a,
(ii) and satis es the interpolation property that for all k > 0, we have
Z
xk · ζp = (−1)k (1 − pk−1 )ζ(1 − k).
Z×
p
Proof. The element θa is not a zero-divisor by Lemma 3.8, so ζp is well-de ned. To prove
independence, if a
b are two integers coprime to p, then
and
Z Z
xk ·(θa x−1 ResZ×
p
µ b ) = xk ·(θb x−1 ResZ×
p
µa ) = (−1)k (1−ak )(1−bk )(1−pk−1 )ζ(1−k)
Z×
p Z×
p
for all k > 0, so that
θa µb = θb µa ,
by Lemma 3.8, giving the required independence. The interpolation property is a formal
consequence of the de nition and the algebra structure on Λ(Z×
p) (see Remark below).
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 21
It remains to prove that ζp augmentation ideal I(Z×
is a pseudo-measure. De ne the p)⊂
Λ(Zp ) to be the kernel of the natural `degree' map Λ(Zp ) → OL [Zp /Zp ] = OL . We have
× × × ×
analogous augmentation ideals I((Zp /p ) ) ⊂ OL [(Zp /p ) ], and an isomorphism
n × n ×
I(Z× ∼ I((Zp /pn )× ).
p ) = lim
←−
The condition on a ensures it is a topological generator of Z×
p. Since (Zp /pn )× is cyclic,
n
generated by the reduction a of a modulo p , we have
I((Zp /pn )× ) = ([a] − [1])OL [(Zp /pn )× ],
and in the inverse limit we see that
I(Z× ×
p ) = ([a] − [1])Λ(Zp ).
Thus if g ∈ Λ(Z×
p ), we have [g] − [1] ∈ I(Z×
p ), and we must have
[g] − [1] = ν([a] − [1])
for some ν ∈ Λ(Z×
p ). Then
([g] − [1])ζp = ν([a] − [1])ζp = ν · x−1 ResZ×
p
µa ∈ Λ(Z×
p ),
that is, ζp is a pseudo-measure.
Remark 3.11 . One must take care in the above when discussing products of measures.
As remarked in Remark 2.18, whilst we identify Λ(Z×
p) as a subset of Λ(Zp ), it is not a
subalgebra; the convolution of measures over Zp uses the additive group structure, and the
convolution of measures over Z×
p the multiplicative group structure. Writing µ ∗Z×
p
λ for the
convolution over Z×
p, we have
Z Z ÅZ ã ÅZ ãÅ Z ã
k k k k
x · (µ ∗Z×
p
λ) = (xy) · µ(x) · λ(y) = x ·µ x ·λ ,
Z×
p Z×
p Z×
p Z×
p Z×
p
justifying the calculations above.
We've removed the dependence on a. We summarise the main result:
Theorem 3.12 . There is a unique pseudo-measure ζ p on Z×
p such that for all k > 0,
we have Z
xk · ζp = (1 − pk−1 )ζ(1 − k).
Z×
p
(Note that we can remove the (−1)k from the interpolation formula as ζ(1 − k) ̸= 0 if
and only if k is even).
De nition 3.13 . We call ζp the KubotaLeopoldt p-adic L-function (or the p-adic zeta
function ).
See Remark 6.19 for further comments on the multiplication by x−1 introduced in the
de nition of ζp .
4. Interpolation at Dirichlet characters
4.1. Characters of p-power conductor. Throughout the construction of the
KubotaLeopoldt p-adic L-function, we've kept half an eye on the interpolation property
and links to the values of the Riemann zeta function, so the interpolation of these values
should not have come as a surprise. However, now some real magic happens. Since the
introduction, we've failed to mention Dirichlet L-functions once but, miraculously, the
KubotaLeopoldt p-adic L-function also interpolates Dirichlet L-values as well.
22 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Theorem 4.1 . Let χ be a (primitive) Dirichlet character of conductor p for some
n
integer n ≥ 1 (seen as a locally constant character of p ).
Z× Then, for k > 0, we have
Z
χ(x)xk · ζp = L(χ, 1 − k).
Z×
p
The rest of this subsection will contain the proof of this result. The proof is somewhat
calculation-heavy, but given familiarity with the dictionary between measures and power
series is not conceptually di cult.
In particular: the Riemann zeta function was the complex Mellin transform of a real
analytic function, which via Theorem 1.4 gave us a formula for its special values. Under
the transformation et = T + 1, we obtained a p-adic power series; and under the measures
power series correspondence given by the Mahler transform, this gave us a measure on Zp ,
from which we constructed ζp . To obtain interpolation at Dirichlet characters, we pursue
this in reverse, as summarised in the following diagram:
(1 − a1−s )ζ(s) (1 − χ(a)a1−s )L(χ, s)
O O
Mellin Mellin
fa (t) o / Fa (T ) ∈ OL [[T ]] fa,χ (t) o / Fa,χ (T ) ∈ OL [[T ]]
et =T +1 O et =T +1 O
twist by χ
µa ∈ Λ(Zp ) o / µa,χ ∈ Λ(Zp )
ζp
Firstly, we introduce a twisting operation on measures. If µ is a measure on Zp , we de ne
a measure µχ on Zp by
Z Z
f (x) · µχ = χ(x)f (x) · µ.
Zp Zp
In particular, under this we have
Z Z
χ(x)xk · ζp = xk · (ζp )χ
Z×
p Z×
p
= ∂ k A(ζp )χ (0),
where the last equality is Corollary 2.16. Thus we want to determine the Mahler transform
of µχ in terms of Aµ , for which we use our measure-theoretic toolkit. This requires a classical
de nition.
De nition 4.2 . Let χ be a primitive Dirichlet character of conductor pn , n ≥ 1. De ne
the Gauss sum of χ as
X
G(χ) ..= χ(c)εcpn ,
c∈(Z/pn Z)×
where (εpn )n∈N denotes a system of primitive pn th roots of unity in Qp such that εppn+1 = εpn
all n ≥ 0 (if we x an isomorphism Qp ∼
n
for = C, then one can take εpn ..= e2πi/p ).
Remark 4.3 . We will make constant use of the following basic properties of Gauss
sums, whose proofs are left as exercises:
G(χ)G(χ−1 ) = χ(−1)pn .
G(χ) = χ(a) c∈(Z/pn Z)× χ(c)εac / Z×
P
pn for any a ∈ Zp . Note that, if a ∈ p, both sides
vanish.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 23
Lemma 4.4 . The Mahler transform of µ χ is
1 X
Aµχ (T ) = χ(c)−1 Aµ (1 + T )εcpn − 1 .
G(χ−1 )
c∈(Z/pn Z)×
Proof. Since χ is constant (mod pn ), the measure µχ is simply
X
µχ = χ(c)Resc+pn Zp (µ).
c∈(Z/pn Z)×
Using this expression and the formula for the Mahler transform of the restriction of a
measure to c + pn Zp , we nd that
1 X X
Aµχ = ζ −b Aµ (1 + T )ζ − 1 .
χ(b)
pn
b∈(Z/pn Z)× ζ∈µpn
Writing µpn = {εcpn : c = 0, ..., pn − 1}, and rearranging the sums, we have
1 X X
Aµχ = n pn Aµ (1 + T )εpn − 1
χ(b)ε−bc c
p
c (mod p ) b∈(Z/p Z)×
n n
1 X
G(χ)χ(−c)−1 Aµ (1 + T )εcpn − 1 ,
=
pn
c∈(Z/pn Z)×
where the last equality follows from the standard identity
X
χ(b)ε−bc
pn = χ(−c)
−1
G(χ)
b∈(Z/pn Z)×
of Gauss sums. We complete the proof by using the identity G(χ)G(χ−1 ) = χ(−1)pn .
Now consider the case where µ = µa , the measure from which we built the Kubota
Leopoldt p-adic L-function, and which has Mahler transform
1 a
Aµa (T ) = − .
T (1 + T )a − 1
Applying the above transformation, we obtain a measure µχ,a with Mahler transform
ñ ô
1 X
−1 1 a
Fχ,a (T ) = χ(c) − .
G(χ−1 ) (1 + T )εcpn − 1 (1 + T )a εac
pn − 1
c∈(Z/pn Z)×
Via the standard substitution et = T + 1 , this motivates the study of the function
ñ ô
1 X 1 a
fχ,a (t) = χ(c)−1 − ,
G(χ−1 ) et εcpn − 1 eat εac
pn − 1
c∈(Z/pn Z)×
by way of analogy with the case of the Riemann zeta function.
Lemma 4.5 . We have
L(fχ,a , s) = χ(−1) 1 − χ(a)a1−s L(χ, s),
where L(fχ,a , s) is as de ned in Theorem 1.4. Hence, for k ≥ 0, we have
(k)
(0) = (−1)k χ(−1) 1 − χ(a)ak+1 L(χ, −k)
fχ,a
= − 1 − χ(a)ak+1 L(χ, −k).
Proof. We follow a similar strategy as in the case of the Riemann zeta function. In
particular, we can expand as a geometric series, obtaining
1 X
= e−kt ε−kc
pn .
et εcpn −1
k≥1
24 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Then we have
Z ∞ XÅ ã
1 X
L(fχ,a , s) = χ(c)−1 e−kt ε−kc
p n − e −akt −akc
ε p n ts−1 dt.
Γ(s)G(χ−1 ) 0 c∈(Z/pn Z)× k≥1
Note that
X
χ(c)−1 ε−akc
pn = χ(−ak)G(χ−1 ),
c∈(Z/pn Z)×
and similarly for the rst term, so that the expression collapses to
Z ∞
1 X
χ(−k) e−kt − χ(a)e−akt ts−1 dt.
L(fχ,a , s) =
Γ(s) 0 k≥1
For Re(s) ≫ 0, we can rearrange the sum and the integral, and then we can evaluate the
k th term of the sum easily to (1 − χ(a)a1−s )k −s , giving
X
L(fχ,a , s) = χ(−1) 1 − χ(a)a1−s χ(−k)k −s ,
k≥1
showing the equality of L-functions. To prove the nal statement about special values,
observe that a simple computation shows that fχ,a (−t) = −χ(−1)fχ,a (t), which implies
(k)
(looking at the series expansions) that fχ,a (0) = 0 unless χ(−1)(−1)k = −1. This concludes
the proof.
Note that, in the proof of Lemma 4.5, we have also shown the following useful fact.
Lemma 4.6 .If χ is an even character, that is if χ(−1) = 1, then L(χ, −k) = 0 if k is
even. If χ is an odd character, then L(χ, −k) = 0 if k is odd.
We can now prove Theorem 4.1.
Proof. (Theorem 4.1). Since χ is 0 on pZp , and by the above, we have
Z Z Z
χ(x)xk · µa = χ(x)xk · µa = xk · µχ,a ,
Z×
p Zp Zp
where µχ,a is the twist of µa by χ. We know this integral to be
∂ k Fχ,a (0) = fχ,a
(k)
(0),
under the standard transform et = T + 1 . Hence, by Lemma 4.5, we nd
Z
χ(x)xk · µa = −(1 − χ(a)ak+1 )L(χ, −k),
Z×
p
so that
Z
χ(x)xk · x−1 µa = −(1 − χ(a)ak )L(χ, 1 − k),
Z×
p
By de nition, we have
Z
χ(x)xk · θa = −(1 − χ(a)ak ),
Z×
p
and hence we nd
Z
χ(x)xk · ζp = L(χ, 1 − k),
Z×
p
as was to be proved.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 25
4.2. Non-trivial tame conductors. We can go even further. The theorem above
deals with the case of `tame conductor 1', in that we have constructed a p-adic measure that
interpolates all of the L-values L(χ, 1 − k) for k>0 and cond(χ) = pn with n ≥ 0 (where
trivial conductor corresponds to the Riemann zeta function). More generally:
Theorem 4.7 .
Let D > 1 be any integer coprime to p, and let η denote a (primitive)
Dirichlet character of conductor D. There exists a unique measure ζη ∈ Λ(Z× p ) such that,
for all primitive Dirichlet characters χ with conductor pn , n ≥ 0, and for all k > 0, we have
Z
χ(x)xk · ζη = 1 − χη(p)pk−1 L(χη, 1 − k).
Z×
.
p
Remark 4.8 (i) In this case, we obtain a genuine measure rather than a pseudo-
measure. As L-functions of non-trivial Dirichlet characters are everywhere holomorphic,
there is no need for the smoothing factor involving a.
(ii) Implicit in this theorem is the fact that the relevant Iwasawa algebra is de ned over
a ( xed) nite extension L/Qp containing the values of η.
Since many of the ideas involved in proving the above theorem are present in the case of
trivial tame conductor, the proof of Theorem 4.7 is a good exercise. As such, we give only the
main ideas involved in the proof. Note rst that the calculation relating L(fχ,a , s) to L(χ, s)
above was entirely classical, in the sense that p did not appear anywhere; accordingly, we can
perform a similar calculation in the general case. Since there is no need for the smoothing
factor a, we can then consider the function
−1 X η(c)−1
fη (t) = .
G(η −1 ) et εcD − 1
c∈(Z/D)×
(This scaling by −1 also appears in the trivial tame conductor situation, but it is incorpo-
rated into θa ). De ning Fη (T ) by substituting T +1 for et and expanding the geometric
series, we nd
−1 X X εkc
Fη (T ) = η(c)−1 D
T k.
G(η −1 ) (εcD − 1)k+1
c∈(Z/D)× k≥0
This is an element of OL [[T ]] for some su ciently large nite extension L of Qp , since the
−1
Gauss sum is a p-adic unit (using G(η)G(η ) = η(−1)D, and the fact that D is coprime to
p) and εcD −1 ∈ OL× (since it has norm dividing D). There is therefore a measure µη ∈ Λ(Zp ),
the Iwasawa algebra over OL , corresponding to Fη under the Mahler transform.
Lemma 4.9 . We have L(f , s)Z= −η(−1)L(η, s). Hence
η
xk · µη = L(η, −k)
Zp
for k ≥ 0.
Proof. This is proved in a similar manner to above, equating ∂ with d/dt and using the
general theory described in Theorem 1.4.
Lemma 4.10 . We haveZψ(F ) = η(p)F . Hence
η η
xk · µη = 1 − η(p)pk L(η, −k).
Z×
p
Proof. We show rst that
1 X 1 1
= . (4.1)
p c
(1 + T )ζεD − 1 (1 + T )p εpc
D −1
ζ∈µp
26 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Expanding each summand as a geometric series, the left hand side is
−1 X X X
(1 + T )n εnc n
Dζ =− (1 + T )pn εpcn
D ,
p
ζ∈µp n≥0 n≥0
and summing the geometric series now gives the right hand side of (4.1). It follows that
−1 X X η(c)−1
(φ ◦ ψ)(Fη ) =
pG(η)−1 (1 + T )ζεcD − 1
ζ∈µp c∈(Z/D)×
−1 X η(c)−1
=
G(η −1 ) (1 + T )p εpc
D −1
c∈(Z/D)×
= η(p)φ(Fη ).
The rst claim now follows by the injectivity of φ. For the second, we note that
ResZ×
p
(µη ) = (1 − φ ◦ ψ)(µη )
= µη − η(p)φ(µη ),
and
Z Z
xk · φ(µη ) = pk xk · µ η .
Zp Zp
The result now follows simply from Lemma 4.9.
Now let χ be a Dirichlet character of conductor pn for some n ≥ 0, and let θ ..= χη the
n
product (a Dirichlet character of conductor Dp ). Using Lemma 4.4, we nd easily that:
Lemma 4.11 . The Mahler transform of µ θ
.
. = (µη )χ is
−1 X θ(c)−1
Fθ (T ) ..= Aµθ (T ) = .
G(θ−1 ) (1 + T )εcDpn − 1
c∈(Z/Dpn )×
Via a calculation essentially identical to the cases already seen, we can prove
Z Z
k
χ(x)x · µη = xk · µ θ
Zp Zp
= L(θ, −k),
that
ResZ×
p
(µθ ) = 1 − θ(p)φ (µθ ),
and that accordingly
Z
χ(x)xk · µη = 1 − θ(p)pk L(θ, −k).
Z×
p
Finally, to complete the proof of Theorem 4.7 and to ensure compatibility with the con-
struction of ζp , we introduce a shift by 1. The following is directly analogous to the con-
struction of ζp ; note again that ζη is truly a measure, not a pseudo-measure.
De nition 4.12 . De ne ζη ..= x−1 ResZ×
p
(µη ).
We see that
Z
χ(x)xk · ζη = 1 − θ(p)pk−1 L(θ, 1 − k).
Z×
p
which completes the proof of Theorem 4.7.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 27
4.3. Analytic functions on Z via the Mellin transform.
p The reader should
hopefully now be convinced that the language of measures is a natural one in which to
discuss p-adic L-functions. In this subsection, we use this (more powerful) language to
answer the question we originally posed in the introduction: namely, we de ne analytic
functions on Zp that interpolate the values ζ(1 − k) for k > 0. In passing from measures
to analytic functions on Zp , we lose the clean interpolation statements. In particular, there
is no single analytic function on Zp interpolating the values ζ(1 − k) for all k > 0; rather,
there are p − 1 dierent `branches' of the KubotaLeopoldt p-adic L-function on Zp , each
interpolating a dierent range.
The reason we cannot de ne a single p-adic L-function on Zp is down to the following
like x−s · ζp
R
technicality. We'd to be able to de ne ζp (s) = Z× p
for s ∈ Zp . The natural
s
way to de ne the exponential x 7→ x is as
xs = exp(s · log(x)),
but unfortunately in the p-adic world the exponential map does not converge on all of Zp ,
so this is not well-de ned for general x ∈ Z×
p. Instead:
Lemma 4.13 .
The p-adic exponential map converges on pZp . Hence, for any s ∈ Zp ,
the function 1 + pZp → Zp given by x 7→ xs ..= exp(s · log(x)) is well-de ned.
Proof. This is a standard result in the theory of local elds. See, for example, [ Cas86 ].
De nition 4.14
∼
. Recall that we assume p to be odd and that we have a decomposition
Z×
p = µp−1 × (1 + pZp ). Let
ω : Z×
p −→ µp−1 ,
⟨·⟩ : Z×
p −→ 1 + pZp ,
where ω(x) = Teichmüller lift of the reduction modulo p of x and ⟨x⟩ = ω −1 (x)x denote the
×
projections to the rst and second factors respectively. Note that if x ∈ Zp , then we can
write x = ω(x)⟨x⟩.
Hence the function ⟨x⟩s is well-de ned. When p is odd, for each i = 1, .., p − 1 we can
de ne an injection
Zp ,−→ Homcts (Z× ×
p , Cp )
s 7−→ x 7→ ω(x)i ⟨x⟩s ,
and hence we can de ne an analytic function
ζp,i : Zp −→ Cp
Z
s 7−→ ω(x)i ⟨x⟩1−s · ζp .
Z×
p
This function does not interpolate as wide a range of values as the measure ζp , since the
character xk can be written in the form ω(x)i ⟨x⟩k if and only if k ≡ i (mod p − 1), and in
this case xk is the value of ω(x)i ⟨x⟩1−s at the value s = 1 − k . Then we have:
Theorem 4.15 . For all k ≥ 0 with k ≡ i (mod p − 1), we have
ζp,i (1 − k) = (1 − pk−1 )ζ(1 − k).
More generally:
28 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
De nition 4.16 . Letθ = χη be a Dirichlet character, where η
n
has conductor D prime
to p and χ has conductor p for n ≥ 0. De ne
Z
Lp (θ, s) ..= χ(x)⟨x⟩1−s · ζη , s ∈ Zp .
Z×
.
p
Remark 4.17
An equivalent de nition is
Z
Lp (θ, s) = χω −1 (x)⟨x⟩−s · µη . (4.2)
Z×
Was97
p
In [ ], the analytic functions Lp (θ, s) are constructed directly without using measures,
and the more direct approach diers from the one obtained using our measure-theoretic
approach by precisely this factor of ω. This twist by 1 also appears naturally when we study
the Iwasawa main conjecture.
Directly from the de nitions, we have ζp,i (s) = Lp (ω i , s). Hence for arbitrary k > 0,
we have
ζp,i (1 − k) = (1 − ω i−k (p)pk−1 )L(ω i−k , 1 − k).
Of course, ω i−k is just the trivial character when i ≡ k (mod p − 1), so we recover Theorem
4.15.
Theorem 4.18 . For all k ≥ 1, we have
Lp (θ, 1 − k) = 1 − θω −k (p)pk−1 L(θω −k , 1 − k).
Proof. We use the description of (4.2). From the de nitions, we have χω −1 (x)⟨x⟩k−1 =
χω −k (x) · ω k−1 (x)⟨x⟩k−1 = χω −k (x)xk−1 , so that
Z Z
χ(x)⟨x⟩k−1 · µη = χω −k (x)xk−1 · µη
Z×
p Z×
p
= 1 − θω −k (p)pk−1 L(θω −k , 1 − k),
as required.
More generally, for any measure µ Z×
onp we de ne
Z
Melµ,i (s) = ω(x)i ⟨x⟩s · µ,
Z×
p
the Mellin transform of µ at i. So ζp,i (s) = Melζp ,i (1 − s).
4.4. The values at s = 1. We end Part I with an example of further remarkable links
between the classical and p-adic zeta functions. Let θ be a non-trivial Dirichlet character,
which as usual we write in the form χη , where χ has conductor pn and η has conductor D
prime to p. As explained above, we have
Z
χ(x)xk · ζη = L(θ, 1 − k)
Z×
p
for k > 0; we say that the range of interpolation is {0, −1, −2, −3, ...}. It's natural to ask
what happens outside this range of interpolation. In particular, what happens when we take
k = 0? Since this isoutside the range of interpolation this value may have a priori nothing
to do with classical L-values. Indeed, the classical value L(θ, 1) is transcendental(10) , and if
it is transcendental one cannot see it as a p-adic number in a natural way. However, just
because we cannot directly equate the two values does not mean there is no relationship
between them; there is a formula for the p-adic L-function at s = 1 which is strikingly
similar to its classical analogue.
(10) This follows from Baker's theorem and Theorem 4.19, part (i).
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 29
Theorem 4.19 . Let θ be a non-trivial even Dirichlet character of conductor N , and let
ζ denote a primitive N th root of unity. Then:
(i) (Classical value at s = 1). We have
1 X
θ−1 (a) log 1 − ζ a .
L(θ, 1) = −
G(θ−1 )
a∈(Z/N Z)×
(ii) (p-adic value at s = 1). We have
1 X
Lp (θ, 1) = − 1 − θ(p)p−1 θ−1 (a) logp 1 − ζ a ).
G(θ−1 )
a∈(Z/N Z)×
If θ is an odd character, both sides of the p-adic formula vanish. In any case, the formulae
are identical up to replacing log with its p-adic avatar and, as usual, deleting the Euler factor
at p. This result can be used to prove a p-adic analogue of the class number formula. For
completeness, we prove these results below.
4.4.1. The complex value at s = 1 .
Proof. (Theorem 4.19, classical value). Write
X X
L(θ, 1) = θ(a) n−s .
a∈(Z/N Z)× n≡a (mod D)
Using the fact that
®
1 X
(a−n)c 0 if n ̸≡ a mod N
ζ =
N 1 if n ≡ a mod N ,
c∈(Z/N Z)
we show that the above formula equals
Ñ é
X 1 X X 1 X X X
θ(a) ζ (a−n)c n−s = θ(a)ζ ac ζ −nc n−s
×
N N
a∈(Z/N Z) n≥1 c∈(Z/N Z) c∈(Z/N Z) a∈(Z/N Z)× n≥1
G(θ) X X
= θ−1 (c) ζ −nc n−s ,
N
c∈(Z/N Z) n≥1
the last equality following from one of the standard identities for Gauss sums. Evaluating
this expression at s = 1 (checking that there is no convergence problem) and using the
Taylor series expansion of the logarithm, and applying the other standard identity of Gauss
sums, we obtain the result.
Remark 4.20 . We see from the formula above that the parity of the character θ plays
an important role on the behaviour of the zeta function at s = 1. Making some elementary
calculations we can deduce that, if θ is even, then
1 X
L(θ, 1) = − θ−1 (c) log |1 − ζ c |.
G(θ−1 )
c∈(Z/N Z)×
If θ is odd, we can use the functional equation to obtain
1
L(θ) = iπ B −1 ,
G(θ−1 ) 1,θ
where Bk,θ denotes the k th twisted Bernoulli number (see [ Was97 , Chapter 4]).
30 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
4.4.2. The p-adic value at s = 1 . Recall the power series
−1 X θ(c)−1
Fθ (T ) = ,
G(θ−1 ) (1 + T )ζ c − 1
c∈(Z/N )×
which gives rise to a measure µθ on Zp that interpolates the special values of L(θ, s).
Accordingly, by the measure-theoretic arguments we've employed repeatedly above, we have
Z
Lp (θ, 1) = ..
x−1 · µθ
Z×
p
= Ax−1 Res × (µθ )
(0).
Zp
We rst compute Ax−1 µθ .
Lemma 4.21 . There exists a constant C such that
1 X
Ax−1 µθ (T ) = − θ−1 (c) log (1 + T )ζ c − 1 + C.
G(θ−1 )
c∈(Z/N Z)×
Proof. This follows immediately from the formula
(1 + T )ζ c 1
∂ log (1 + T )ζ c − 1 =
c
=1+
(1 + T )ζ − 1 (1 + T )ζ c − 1
θ−1 (c) = 0.
P
and the fact that
. We have
c∈(Z/N Z)
Lemma 4.22
ARes (T ) = Ax−1 µθ (T ) − θ(p)p−1 Ax−1 µθ (1 + T )p − 1 .
× (µθ )
Zp
Proof. This is immediate from the formula
ARes × (µχ )
(T ) = (1 − φ ◦ ψ)Aµχ (T )
Zp
and the fact that
ψ(x−1 µθ ) = p−1 x−1 ψ(µθ )
= χ(p)p−1 x−1 µθ .
We can now complete the proof of Theorem 4.19.
Proof. (Theorem 4.19, p-adic value). Evaluating at T =0 the formula of Lemma 4.22
and using Lemma 4.21 we obtain
Lp (θ, 1) = −(1 − θ(p)p−1 )Ax−1 µθ (0)
1 X
= −(1 − θ(p)p−1 ) θ−1 (c) logp (ζ c − 1),
G(θ−1 )
c∈(Z/N Z)×
as required.
5. The p-adic family of Eisenstein series
We now take a brief detour to illustrate another example of p-adic variation in number
theory, namely the p-adic variation of modular forms. In constructing the KubotaLeopoldt
p-adic L-function, we have seen many of the key ideas that go into the simplest example
of this, namely the p-adic family of Eisenstein series, which we will illustrate below. For
simplicity, in this section we'll take p an odd prime.
Let k≥4 be an even integer. The Eisenstein series of level k , de ned as
X 1
Gk (z) ..= , z ∈ H ..= {z ∈ C : Im(z) > 0}
(cz + d)k
c,d∈Z
(c,d)̸=(0,0)
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 31
can be viewed as a two-dimensional analogue of the zeta value ζ(k). It is an example of
a modular form of weight k . In the classical theory of modular forms, one computes the
normalised Fourier expansion of such an object to be
Gk (z)(k − 1)! ζ(1 − k) X
Ek (z) ..= = + σk−1 (n)q n ,
2 · (2πi)k 2
n≥1
DS05
k−1 2iπz
P
where σk−1 (n) =0<d|n d and q = e . In particular, it is a power series with rational
coe cients. (This is a standard exercise; see [ , Chapter 1.1] for details).
From the de nition, we see the KubotaLeopoldt p-adic L-function as a pseudo-measure
that, when evaluated at xk with k≥4 even, gives the constant coe cient of the Eisenstein
series of weight k. The idea now is to nd measures giving similar interpolations of the other
coe cients. Fortunately, these are much easier to deal with. We want interpolations of the
functionsd 7→ dk−1 , where k is varying p-adically. When d is coprime to p, we can de ne this
measure simply to be δd , the Dirac measure at d (recalling this is de ned by evaluation at d).
When d p, however, we run into an immutable obstacle. There is no Dirac
is divisible by
measure on Z×
p corresponding / Z×
to evaluation at p, since p ∈ p . Moreover, the function
p 7→ p can never be interpolated continuously p-adically; it simply behaves too badly for
k
this to be possible. Suppose there was indeed a measure θp with
Z
xk · θ p = p k ,
Z×
p
and then suppose kn is a strictly increasing sequence of integers p-adically tending to k.
Then Z Z
pkn = xkn · θp −→ xk · θp = pk ,
Z×
p Z×
p
which is clearly impossible since pkn tends to 0. We get around this issue by taking p-
stabilisations to kill the coe cients at p.
De nition 5.1 . We de ne the p-stabilisation of Ek to be
(p)
Ek (z) ..= Ek (z) − pk−1 Ek (pz).
An easy check shows that
(p) (1 − pk−1 )ζ(1 − k) X p
Ek = + σk−1 (n)q n ,
2
n≥1
where
p
X
σk−1 (n) = dk−1 .
0<d|n
p∤d
(p)
(The series Ek de nes a modular form of weight k and level Γ0 (p)).
We've done all the work in proving:
Theorem 5.2 . There exists a powerXseries
E(z) = An q n ∈ Q(Z×
p )[[q]]
n≥0
such that:
(a) A0 is a pseudo-measure, and An ∈ Λ(Z×
p ) for all n ≥ 1;
32 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
(b) For all even k ≥ 4, we have
Z XÅZ ã
k−1 (p)
x ..
· E(z) = xk−1 · An q n = Ek (z).
Z×
p n≥0 Z×
p
Proof. Clearly, A0 is simply the pseudo-measure xζp /2 (shifting by 1 again in the opposite
direction). We then de ne
X
An = δd ∈ Λ(Z×
p ).
0<d|n
p∤d
By the interpolation property of the KubotaLeopoldt p-adic L-function, A0 interpolates
the constant term of the Eisenstein series. We also have
Z X Z
xk−1 · An = xk−1 · δd
Z×
p 0<d|n Z×
p
p∤d
(p)
X
= dk−1 = σk−1 (n),
0<d|n
p∤d
so we get the required interpolation property.
Remark .
These results are often presented in a dierent (equivalent) way. One de nes the weight
space
W(Cp ) = Homcts (Z× ×
p , Cp )
and shows that, topologically, it is the union of p − 1 open unit balls in Cp (centered around
the (p − 1)th roots of unity). W(Cp ) via the maps
The integers are naturally a subset of
x 7→ xk , and two integers k, k ′ ′
lie in the same unit ball if and only if k ≡ k (mod p − 1). This
space can be given more structure; there is a rigid analytic space W such that the elements of
W(Cp ) are the Cp -points of W . Giving a measure on Z× p is equivalent to giving a bounded
rigid analytic function on W . De ning O(W) to be the space of rigid analytic functions on
W , we can view E as a power series in O(W)[[q]]. We see it as a p-adic interpolation of the
Eisenstein series over the weight space.
The power series E(z) is an example of a Λ-adic modular form. In particular, it can be
colloquially described as the statement:
Eisenstein series vary p-adically continuously as you change the
weight; if k and k′ are closep-adically, then the Fourier expansions of
Ek and Ek ′ are close p-adically.
The theory of p-adic modular forms, and in particular the construction and study p-adic
families of Eisenstein series, was introduced by Serre to give a new construction of the p-
adic zeta function of a totally real number eld. Pioneering work of Hida went much further
than this, showing that similar families (known as Hida families ) exist for far more general
Hid93
modular forms, and his work has been vastly generalised to the theory of Coleman families
and eigenvarieties. For a avour of Hida's work, see his book [ ].
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 33
Part II: Iwasawa's main conjecture
This part is devoted to the formulation and study of Iwasawa's Main Conjecture. We start
by studying cyclotomic units, which will play a central role in the sequel, and by showing
their relation to class numbers. We then state and prove the existence of Coleman's inter-
polating power series. These interpolating power series and the study of their logarithmic
derivatives will lead to an alternative construction of the p-adic zeta function. Although a
priori more obscure, this new construction will stablish a tangible connection between units
in the cyclotomic tower and the p-adic zeta function. Finally, using class eld theory, we
will naturally arrive at the formulation and proof of (a special case of ) the Main Conjecture.
Notation . Our study of the Iwasawa main conjecture requires a lot of notation, which
we introduce straight away for convenience. The following should be used as an index of
the key notation, and the reader is urged to consult the de nition of new objects as they
appear in the text.
Let p be an odd prime. Throughout this section, we work with coe cient eld L = Qp .
For n ∈ N, write
Fn ..= Q(µpn ), Fn+ ..= Q(µpn )+ ;
Vn ..= OF×n , Vn+ ..= OF×+ ;
n
Kn ..= Qp (µpn ), Kn+ ..= Qp (µpn )+ ;
× ×
Un ..= OK n
, Un+ ..= OK +.
n
The extensions Fn /Q, Kn /Qp , Fn+ /Q and Kn+ /Qp are Galois and totally rami ed (the rst
n−1 p−1 n−1
two of degree (p − 1)p and the last two of degree
2 p ) and we denote pn the unique
prime ideal above the rational prime p. We note
[
+ .. +
F∞ = Q(µp∞ ) = Fn , F∞ = F∞ ,
n≥1
.. + +..
and G = Gal(F∞ /Q), G = Gal(F∞ /Q) = G/⟨c⟩, where c denotes the complex conjugation.
n n
Since Gal(Fn /Q) sends a primitive p th root of unity to a primitive p th root of unity, one
deduces an isomorphism
χn : Gal(Fn /Q) −∼→ (Z/pn Z)×
given by
σ(ζ) = ζ χn (σ) ,
for σ ∈ Gal(Fn /Q) and ζ ∈ µpn any primitive pn th root of unity. By in nite Galois theory,
we then see that
G = Gal(F∞ /Q) ..= lim Gal(Fn /Q) = lim(Z/pn Z)× ∼
= Z×
p,
←− ←−
n n
cyclotomic character χ ..= lim χn .
via the Observe that χ induces an isomorphism
+ ∼ ×
←−
G = Zp /{±1}.
We also de ne
Un,1 ..= {u ∈ Un : u ≡ 1 (mod pn )}, Un,1
+ ..
= Un,1 ∩ Un+ . (5.1)
The subsets Un,1 and Un,1
+
are important as they have the structure of Zp -modules (indeed,
if u ∈ Un,1 or Un,1 and a ∈ Zp , then u =
+ × a a k
P
k≥0 k (u − 1) converges). By contrast, the
full local units Un and Un are only Z-modules.
+
In general, our notations satisfy the following logic: if Xn is any subgroup of Un , then
+
we let Xn = Xn ∩ Un+ , Xn,1 = Xn ∩ Un,1 and Xn,1
+
= Xn+ ∩ Un,1
+
. Observe that, since
34 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Vn ⊆ Un , the same applies for any subgroup Xn of Vn .
It will be essential for our constructions and methods to consider these modules at all
levels simultaneously. In that spirit, we de ne
U∞ ..= lim Un , U∞,1 ..= lim Un,1 ;
←− ←−
n n
U∞
+ ..
= lim Un+ , U∞,1
+ ..
= lim Un,1
+
;
← −
n
← −
n
where all limits are taken with respect to the norm maps. All of these in nite level
modules are compact Zp -modules (since they are inverse limits of compact Zp -modules)
and moreover they are all endowed with natural continuous actions of G = Gal(F∞ /Q) or
G + = Gal(F∞
+
/Q) = G {±1} . Accordingly, they are endowed with continuous actions of the
+
Iwasawa algebras Λ(G) or Λ(G ) (which is the primary reason for passing to in nite level
objects).
We x once for all a compatible system of roots of unity (ζpn )n∈N , that is, a sequence
n p
where ζpn is a primitive p th root of unity such that ζ n+1
p = ζpn for all n ∈ N. We let
πn = ζpn − 1, which is a uniformiser of Kn .
6. The Coleman map
In 6.4, we use the identi cations above to construct the Coleman map, which can be seen
as a machine for constructing p-adic L-functions from a compatible system of units. We also
explain (very!) brie y how this map can be generalised to a machine conjecturally capable
of constructing p-adic L-functions for a very large class of p-adic Galois representations.
Coleman's work is important as it puts the construction of the p-adic zeta function into a
larger and more conceptual framework. Even more importantly, it provides an important
bridge between analytic objects (p-adic L-functions) and algebraic structures (the arithmetic
of cyclotomic elds), and will serve as the key step in our formulation of the main conjecture
in the next section.
6.1. Notation and Coleman's theorem. In the notational introduction to Part II,
we de ned
Kn = Qp (µpn ), K∞ = Qp (µp∞ )
to be the local versions of Fn and F∞ from the previous section. We also de ned
×
Un = OKn
to be the module of local units at level n, took a compatible system (ζpn ) of primitive pn th
roots of unity, and de ned πn ..= ζpn − 1, a uniformiser for Kn .
Proposition 6.1 . Let u ∈ Un be a local unit at level n. There exists a power series
f ∈ Zp [[T ]] such that f (πn ) = u.
Proof. This is essentially immediate from the fact that πn is a uniformiser. Indeed, Kn
is totally rami ed, so one can choose some a0 ∈ Zp such that
a0 ≡ u (mod πn ),
and then a1 ∈ Zp such that
u − a0
a1 ≡ (mod πn ),
πn
f (T ) = n an T n . By construction,
P
and so on, and then de ne this satis es the required
property.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 35
The problem with this proposition is that such a power series f is far from unique, since
we had an abundance of choice at each coe cient. In the usual spirit of Iwasawa theory,
Coleman realised it was possible to solve this problem by passing to the in nite tower K∞ .
Recall that we de ned
U∞ ..= lim Un ,
←−
n
where the projective limit is taken with respect to the norm maps Kn → Kn−1 . Coleman's
theorem says that for each u ∈ U∞ , there is a unique power series fu satisfying the condition
of the above proposition for all n. A more useful formulation of this statement is:
Theorem 6.2 (Coleman). There exists a unique injective multiplicative map
U∞ → Zp [[T ]]
u 7→ fu
such that fu (πn ) = un for all u ∈ U∞ and n ≥ 1.
We prove this theorem in 6.3 below.
6.2. Example: cyclotomic units. This theorem allows an alternative construction of
the KubotaLeopoldt p-adic L-function. In particular, let a∈Z prime to p, and de ne
ζpan −1
cn (a) ..= ∈ Un .
ζ pn −1
Lemma 6.3 . We have c(a) = ((c (a))..
n n ∈ U∞ .
Proof. This is equivalent to proving that Nn,n−1 (cn (a)) = cn−1 (a). Since the minimal
polynomial of ζ pn over Kn−1 is X p − ζpn−1 , we see that
Y
Nn,n−1 (ζpan − 1) = (ζpan η − 1)
η∈µp
= ζpap
n − 1 = ζpan−1 − 1,
Xp − 1 =
Q
where in the penultimate equality we have used the identity η∈µp (Xη − 1). The
result now follows since the norm is multiplicative.
It is possible to write down fc(a) ∈ Zp [[T ]] directly by inspection; indeed, we see that
(1 + T )a − 1
fc(a) (T ) =
T
satis es the required property. (Indeed, fc(a) is even a polynomial).
Proposition 6.4 . We have
∂ log fc(a) = a − 1 − Fa (T ),
where Fa (T ) is the power series de ned in Lemma 3.2.
Proof. We compute directly that
∂ log fc(a) = ∂ log (1 + T )a − 1 − ∂ log(T )
a(1 + T )a T +1
= a
−
(1 + T ) − 1 T
a 1
=a−1+ −
(1 + T )a − 1 T
= a − 1 − Fa (T ),
as required.
36 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Remark 6.5 . In particular, we nd that
(1 − φ ◦ ψ)∂ log fc(a) = −(1 − φ ◦ ψ)Fa ,
since 1 − φ ◦ ψ kills the constant term. In terms of measures, this means that the measure
ResZ× p
(µ∂ log fc(a) ) is equal to the measure ResZ×
p
(µa ) which we used in the construction of
ζp . Hence Coleman's isomorphism allows us to study ζp , and in particular the structure of
Λ(G)/ζp I(G), using local units. We will see more about the units cn (a), and in particular
the module in Un generated by them, in the next section.
6.3. Proof of Coleman's theorem. The proof of Theorem 6.2 will occupy the rest
of this section. We begin by showing the injectivity of the map u 7→ fu .
Lemma 6.6 . Suppose u = (u ) ∈ U n ∞ and f, g ∈ Zp [[T ]] both satisfy
f (πn ) = g(πn ) = un
for all n ≥ 1. Then f = g .
Proof. The Weierstrass preparation theorem (Proposition C.2 of the appendix) says that
we can write any non-zero power series h(T ) ∈ Zp [[T ]] in the form pm u(T )r(T ), where u(T )
is a unit and r(T ) h(T ) converges to a function on
is a distinguished polynomial. Any such
Qp , and since u(T ) cannot have zeros, we deduce
the maximal ideal in the ring of integers of
that h(T ) has a nite number of zeros in this maximal ideal. Now (πn )n≥1 is an in nite
sequence of elements in this maximal ideal, so the fact that (f − g)(πn ) = 0 for all n ≥ 1
implies that f = g , as required.
We now move to showing the existence of such a series fu . Lemma 6.7 and Proposition
6.8 below will show the existence of a norm operator on power series, and then translate
the norm compatibility condition of units into norm invariance of power series; Lemma 6.9
will show certain continuity properties of this norm operator, which will allow us to prove
the theorem by a standard diagonal argument.
Recall that the action of φ on Zp [[T ]] is de ned by φ(f )(T ) = f ((1 + T )p − 1), f (T ) ∈
Zp [[T ]], and that this action is injective. Importantly, we also have
φ(f )(πn+1 ) = f ((πn+1 + 1)p − 1) = f (ζppn+1 − 1) = f (πn ).
From our work with measures, we have also seen the existence of an additive operator ψ
with the property that
1 X
(φ ◦ ψ)(f )(T ) = f (ζ(1 + T ) − 1),
p
ζ∈µp
and that we henceforth call the trace operator (this terminology will become clear in the
proof of Lemma 6.7). We now de ne a multiplicative version of this operator.
Lemma 6.7 . There exists a unique multiplicative operator N , the norm operator, such
that
Y
(φ ◦ N )(f )(T ) = f (ζ(1 + T ) − 1).
ζ∈µp
Proof. The ring B = Zp [[T ]] is an extension of A = Zp [[φ(T )]] = φ(Zp [[T ]]) of degree
p, the former being obtained by adjoining a pth root of (1 + T )p to the latter. Each au-
tomorphism of B over A is given by T →
7 (1 + T )ζ − 1 for some ζ ∈ µp . There is a norm
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 37
map
NB/A : Zp [[T ]] −→ φ(Zp [[T ]])
Y
f (T ) 7−→ f ((1 + T )ζ − 1).
ζ∈µp
The norm operator N is then de ned to be φ−1 ◦ NB/A , recalling that φ is injective. (The
−1 −1
trace operator is similarly equal to p φ ◦ TrB/A , where TrB/A is the trace operator for
the same extension).
Let f ∈ Zp [[T ]]× ; then f (πn ) ∈ Un for all n. Suppose additionally that N (f ) = f , that
is, f is invariant under the norm map. Then:
Proposition 6.8 . For f as above, we have
Nn+1,n f (πn+1 ) = f (πn ),
so that (f (πn ))n ∈ U∞ is a norm compatible system of local units.
Proof. Since the minimal equation of ζn+1 over Kn is X p − ζn = 0, we can write the
norm as
Y
Nn+1,n f (ζn+1 − 1) = f (νζn+1 − 1).
ν∈µp
Q
By the de nition of N, since N (f ) = f , we have φ(f )(T ) = ν∈µp f (ν(1 + T ) − 1), so that
Y
φ(f )(πn+1 ) = f (νζn+1 − 1).
ν∈µp
Since φ(f )(πn+1 ) = f (πn ), we are done.
Let W denote the subspace of Zp [[T ]]× on which N acts as the identity. Since the Coleman
power series attached to a system of units is unique, we have an injection W ,→ U∞ given
by evaluation at (πn )n≥1 . To prove Theorem 6.2 it su ces to prove that this map is also
surjective. We need the following lemma on the behaviour of N modulo powers of p.
Lemma 6.9 . Let f (T ) ∈ Z [[T ]]. We have
p
(i) If φ(f )(T ) ≡ 1 (mod pk ) for some k ≥ 0, then f (T ) ≡ 1 (mod pk ).
(ii) For f ∈ Zp [[T ]] , we have
×
N (f ) ≡ f (mod p).
(iii) For f as above, if f ≡ 1 (mod pk ) with k ≥ 1, then
N (f ) ≡ 1 (mod pk+1 ).
(iv) If f ∈ Zp [[T ]]× , and k2 ≥ k1 ≥ 0, then
N k2 (f ) ≡ N k1 (f ) (mod pk1 +1 ).
Proof. We leave parts (i) and (ii) as an exercise. To see part (iii), suppose that f ≡
1 (mod pk ) with k ≥ 1, and let p denote the maximal ideal of K1 = Qp (µp ). For each ζ ∈ µp ,
as (ζ − 1)(1 + T ) ∈ pZp [[T ]], we have
ζ(1 + T ) − 1 ≡ T (mod pZp [[T ]]),
so that
f ζ(1 + T ) − 1 ≡ f (T ) (mod ppk Zp [[T ]])
by considering each term seperately. It follows that
Y
φ ◦ N (f )(T ) = f ζ(1 + T ) − 1
ζ∈µp
≡ f (T )p (mod ppk Zp [[T ]]),
38 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
but since both φ ◦ N (f ) and f (T )p are elements of Zp [[T ]], this is actually an equivalence
modulo ppk ∩ Zp = pk+1 . If f (T ) ≡ 1 (mod pk ), then f (T )p ≡ 1 (mod pk+1 ), and then the
proof follows from part (i).
To see part (iv), from part (ii) we see that
N k2 −k1 f
≡ 1 (mod p).
f
Then iterating N and using part (iii) k1 times, we obtain the result.
With this in hand, we can complete the proof.
Proof of Theorem 6.2. Uniqueness and injectivity of the map was proved in Lemma 6.6.
We now prove that the map W ,→ U∞ is surjective. Let u = (un )n≥1 ∈ U∞ . For each n,
choose fn ∈ Zp [[T ]] such that
fn (πn ) = un
2n
and de ne gn = N fn . One can use Lemma 6.9 to show that
gm (πn ) ≡ un (mod pm+1 ),
so that limm→+∞ gm (πn ) = un . So it su ces to nd a convergent subsequence of (gm ); but
such a subsequence exists, since Zp [[T ]] is compact. If we let fu denote the limit of this
subsequence, then we have fu (πn ) = un for all n, proving the required existence.
In fact, since, by construction, N (fu ) = fu in the above proof, we have proved:
Theorem 6.10 . The association u 7→ f u induces an isomorphism
U∞ −∼→ (Zp [[T ]]× )N =1 .
6.4. The Coleman map. The Coleman map is motivated by the example of 6.2,
where we saw that a distinguished family of local units the cyclotomic units are strongly
linked to the KubotaLeopoldt p-adic L-function. In particular, in Remark 6.5, we saw that
ζp can be de ned by:
considering the tower c(a) of cyclotomic units,
taking its Coleman power series fc(a) ,
applying ∂ log,
×
restricting to Zp via (1 − φ ◦ ψ),
−1
multiplying by x (which corresponds to ∂ −1 on power series),
passing to the corresponding measure on Z×
p by inverting the Amice transform,
and nally dividing by θa .
We are led to consider the following.
De nition 6.11 . Let
u7→fu (T ) ∂ log 1−φ◦ψ ∂ −1 A −1
Col : U∞ −−−−−−→ (Zp [[T ]]× )N =1 −−−→ Zp [[T ]] −−−−→ Zp [[T ]]ψ=0 −−→ Zp [[T ]]ψ=0 −−−→ Λ(Z×
p ),
where the rst map is the Coleman isomorphism, the second is the logarithmic derivative
appearing in 6.2, the third corresponds on measures to restricting from Zp to Z×
p, the
−1
fourth is multiplication by x , and the last is the usual Amice correspondence between
power series and measures.
In particular, the results of 6.2 show that we have
Col(c(a))
ζp = − .
θa
The Coleman map ultimately puts this connection between cyclotomic units and the
KubotaLeopoldt p-adic L-function into a much larger Galois-theoretic framework.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 39
We have already hinted, in Remark 2.20, that the structure theorem for Λ(Z×
p )-modules
stated in Appendix C is crucial in studying the Iwasawa Main Conjecture. It is natural to
try and equip U∞ with a Λ(Z× ×
p )-module structure such that Col becomes a map of Λ(Zp )-
× ×
modules. As Λ(Zp ) is the completed group ring of Zp over Zp , this amounts to equipping
×
it with compatible actions of Zp and Zp . For the latter, we may use the natural Galois
action of G ∼
= Z×p on the local units. For the former, however, we are stuck: whilst there is
a natural action of Z on U∞ by u 7→ u for an integer a, this does not extend to an action
a
of Zp .
6.4.1. The action of Z p . To x the absence of a Zp -action on local units, we recall the
de nition of the subgroup U∞,1 ⊂ U∞ introduced in the discussion following (5.1). In
particular, we showed there that the action of Z does extend to Zp on U∞,1 .
Proposition 6.12 . The map Col induces a map
U∞,1 −→ Λ(Z×
p)
of Zp -modules, which we also denote by Col.
Proof. The map is obtained by restricting to U∞,1 . We check that it is Zp -equivariant
by showing this for each map in the composition.
a ∈ Zp on U∞,1 is by u 7→ ua . u ∈ U∞,1 , and write fu = ak (u)T k ;
P
The action of Let k≥1
then
a0 (u) ≡ 1 (mod p).
Indeed, by de nition fu (πn ) = un ≡ 1 (mod pn ) for each n, and as πn is a uniformiser for
Kn , we see that
X
fu (πn ) = a0 (u) + ak (u)πnk ∈ a0 (u) + pn ,
k≥1
from which we see that a0 (u) ≡ 1 (mod pn ). But it lies in Zp , hence the claim.
Thus fu (T ) − 1 ∈ (p, T ). As Zp [[T ]] is complete in the (p, T )-adic topology,
Ç å
X a
a
fu (T ) = (fu (T ) − 1)j
j
j≥0
a
converges to a power series fu (T ) ∈ Zp [[T ]]. Since by construction fua (πn ) = uan , we have
fua = fua ∈ W = (Zp [[T ]]× )N =1 .
As a result, we have equipped the image of U∞,1 in W with a Zp -action such that the
restriction of the Coleman isomorphism is Zp -equivariant.
We now compute that ∂ log(fua ) = a∂ log(fu ), so ∂ log is equivariant for the natural
Zp -action on Zp [[T ]]. Finally the maps (1 − φ ◦ ψ), ∂ −1 and A −1 are Zp -equivariant by
de nition.
Finally, we note that in passing to this restriction, we have not lost any information.
Lemma 6.13 . We have U ∞ = µp−1 × U∞,1 .
Proof. We start at nite level n. As p is totally rami ed in Kn for all n, there is a unique
prime pn of Kn above p, and reduction modulo pn gives a short exact sequence
1 → Un,1 → Un → µp−1 → 1,
which is clearly split; thus Un = µp−1 × Un,1 . The result follows in the inverse limit.
Lemma 6.14 . The subgroup µ p−1 of U∞ is killed by Col. In particular, no information
is lost when restricting to U∞,1 .
40 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. Note µp−1 ⊂ Z×
p. The rst map u 7→ fu is an isomorphism that sends v ∈ µp−1 to
∂ log
the constant power series fv (T ) = v . But the kernel of the second map Zp [[T ]] −−−→ Zp [[T ]]
is comprised of constant power series, and in particular, it kills fv . Thus µp−1 is mapped to
zero under the composition, and hence under Col.
Remark 6.15
p
. f ∈ Zp [[T ]] is constant and invariant under N , then this forces
In fact, if
f = f; thus the kernel of the composition of the rst two maps is exactly µp−1 .
6.4.2. The action of Galois . From now on, we identify Z×
p with the Galois group G=
Gal(F∞ /Q) via the cyclotomic character. Under this identi cation, we consider Col to have
image in Λ(G).
Proposition 6.16 . The group U ∞,1 is a Λ(G)-module.
Proof. The Galois group G is naturally isomorphic to Gal(K∞ /Qp ), since p is totally
rami ed in F∞ ; in particular, it acts on U∞ . This action xes 1 ∈ µp−1 and hence stabilises
the subspace U∞ . Moreover it commutes with the Zp -action, hence the result.
Notation . For a ∈ Z×
p, let σa ∈ G be the corresponding element of G with χ(σa ) = a,
recalling that χ is the cyclotomic character.
Proposition 6.17 . For a ∈ Z , and u ∈ U
×
p ∞, we have
Col(σa (u)) = σa (Col(u)).
That is, Col is equivariant for the G -action.
Proof. This is easy to check if we understand how G∼
= Z×
p acts on each of the modules
involved. If u = (un )n≥1 ∈ U∞ , then
σa (u) = (σa (un ))n≥1 ∈ U∞ ,
and if f (T ) ∈ Zp [[T ]], then
σa (f )(T ) = f (1 + T )a − 1 .
Then:
We have
(σa fu )(πn ) = fu ((1 + πn )a − 1)
= fu (ζpan − 1)
= fu (σa (ζpn − 1))
= σa (un ),
so that u 7→ fu (T ) is G -equivariant.
If f (T ) ∈ Zp [[T ]]× , then an easy calculation on power series shows that
∂ log(σa (f )) = aσa (∂ log(f )).
Z×
On measures, restriction to p is G -equivariant since the action of σa is by multiplying
× ×
the variable by a ∈ Zp , which obviously stabilises both Zp and pZp .
ψ=0
As operations on Zp [[T ]] , we have
∂ −1 ◦ σa = a−1 σa ◦ ∂ −1 ,
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 41
as is easily checked on measures; indeed,
Z Z
f (x)
f (x) · ∂ −1 σa µ = · σa µ
Z×
p Z×
p
x
Z
f (ax)
= ·µ
ax
Z×
p
Z
= a−1 f (ax) · ∂ −1 µ
Z×
p
Z
= a−1 f (x) · σa ∂ −1 µ.
Z×
p
By de nition of the action, the inverse Amice transform A −1 is equivariant under σa .
Putting all that together, the result follows.
The results of this subsection can be summarised as follows.
. The map Col restricts to a map U
Corollary 6.18 ∞,1 → Λ(G) of Λ(G)-modules.
Remark 6.19. There is a conceptual explanation for why ∂
−1
is required here, and why
) in the de nition of ζp : the Galois action on U∞ is, by de nition, given
−1
it appeared (via x
by the cyclotomic character, so the left-hand side corresponds to the Galois representation
Qp (1). The measure µa , as de ned in Part I, interpolates L-values for the trivial Galois
representation Qp . The ∂ −1 normalises away that dierence. We return to this in Remark
7.2.
6.5. The fundamental exact sequence. The Iwasawa Main Conjecture puts the
relationship between cyclotomic units and ζp into a wider framework. For this, it is important
to measure the kernel and cokernel of the Coleman map which relates them.
6.5.1. The logarithmic derivative . We start with a more careful study of the logarithmic
derivative ∂ log. In this section, we show that any element xed by ψ is the logarithmic
CS06
derivative of an element xed by the norm operator. For convenience of notation, and
consistency with [ ], we make the following de nition.
De nition 6.20. For f (T ) ∈ Zp [[T ]]× , de ne its logarithmic derivative as
∆(f ) ..= ∂ log f
∂f (T ) f ′ (T )
= = (1 + T ) .
f (T ) f (T )
The main result of this section is the following.
Theorem 6.21 . The logarithmic derivative induces a surjection
N =1
Zp [[T ]]× → Zp [[T ]]ψ=1
with kernel µp−1 .
The di culty of the proof of Theorem 6.21 lies in the fact that, in general, the module
Zp [[T ]]ψ=1 admits no simple description. We rst prove (Lemma 6.22) that the image of
the logarithmic derivative is contained Zp [[T ]]ψ=1 and calculate its kernel. We then reduce
the proof, via Lemma 6.23, to the analogous result modulo p. Finally, in Lemma 6.24 and
Lemma 6.25 we calculate the reduction modulo p of both spaces. Recall that we de ned
N =1
W = Zp [[T ]]× .
Lemma 6.22 . We have ∆(W ) ⊆ Z [[T ]] p
ψ=1
and the kernel of ∆ on W is µp−1 .
42 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. We described the kernel of ∆ in Remark 6.15 above. To see the rst part, if
f ∈ W , then
Y
φ(f ) = (φ ◦ N )(f ) = f ((1 + T )ζ − 1).
ζ∈µp
(11)
Applying ∆ to the above equality and using the fact that ∆ ◦ φ = p φ ◦ ∆, we obtain
X
−1
(φ ◦ ∆)(f ) = p ∆(f )((1 + T )ζ − 1) = (φ ◦ ψ)(∆(f )),
ζ∈µp
which shows that ψ(∆(f )) = ∆(f ) by injectivity of φ.
We move now to the proof of surjectivity. In the following, let
A = ∆(W ) ⊆ Fp [[T ]]; B = Zp [[T ]]ψ=1 ⊆ Fp [[T ]]
be the reduction modulo p of the modules we need to compare.
Lemma 6.23 . If A = B, then ∆(W ) = Z [[T ]] p
ψ=1
.
Proof. Let f0 ∈ Zp [[T ]]ψ=1 . By hypothesis, there exists a g1 ∈ W such that ∆(g1 ) − f0 =
pf1 for some f1 ∈ Zp [[T ]]. Since ∆(W ) ⊆ Zp [[T ]]ψ=1 by Lemma 6.22, we deduce that
ψ(f1 ) = f1 and hence there exists some g2 ∈ W such that ∆(g2 ) − f1 = pf2 for some
f2 ∈ Zp [[T ]]. We deduce by induction the existence of gi ∈ W and fi ∈ Zp [[T ]]ψ=1 , i ≥ 1,
such that
∆(gi ) − fi−1 = pfi .
Since ∆(a) = 0 for any a ∈ Z×
p and since ψ is linear, we can assume that gi (0) ≡ 0 (mod p)
for all i ≥ 1. If we let
n
Y k
hn = (−1)k−1 (gk )p ∈ W ,
k=1
then one easily checks that ∆(hn ) = f0 + (−1)n−1 pn+1 fn . By compactness, the sequence
(hn )n≥1 admits a convergent subsequence converging to an element h ∈ W satisfying ∆(h) =
f0 , which shows the result.
The following lemma calculates the reduction modulo p of W.
Lemma 6.24 . We have W = F [[T ]] . p
×
Proof. One inclusion is obvious. Conversely, for any element f ∈ Fp [[T ]]× , lift it to an
element f˜0 ∈ Zp [[T ]]× (ii) and (iv) of Lemma 6.9, the sequence N k (f˜0 )
and, by points
˜
converges to an element f that is invariant under N and whose reduction modulo p is f .
As we pointed out, the delicate and technical part of the proof of Theorem 6.21 is con-
tained in the following two lemmas describing the reduction of Zp [[T ]]ψ=1 modulo p.
Lemma 6.25 . We have B = ∆(F [[T ]] ). p
×
Proof. We have ∆(W ) ⊆ Zp [[T ]]ψ=1 by Lemma 6.22, so the inclusion ∆(Fp [[T ]]× ) ⊂ B
is clear using Lemma 6.24. For the other inclusion, take any f ∈B and use Lemma 6.26
below to write
f = ∆(a) + b
P+∞
for some a ∈ Fp [[T ]]× and b= T +1 pm
m=1 dm T T . Since ψ(f ) = f and ψ(∆(a)) = ∆(a) (by
a slight abuse of notation, as f and ∆(a) are actually the reduction modulo p of elements
xed by ψ ), we deduce that ψ(b) = b. But we can explicitly calculate the action of ψ on
(11) It is easy to see on power series using the de nitions.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 43
(12)
b. Using the identity ψ(g · φ(f )) = ψ(g)f , the identity T pm = φ(T m ) in Fp [[T ]] and the
T +1
fact that ψ xes
T , we deduce that
+∞
X T +1 m
ψ(b) = dm T ,
m=1
T
which immediately implies b=0 and concludes the proof.
Lemma 6.26 . We have
T +1
Fp [[T ]] = ∆(Fp [[T ]]× ) + C,
T
where C = { +∞ pn
} ⊆ Fp [[T ]].
P
n=1 an T
P+∞
Proof. One inclusion is clear. Take g ∈ Fp [[T ]] and write
T
T +1 g = n=1 an T n . De ne
+∞ +∞
X X k
h= am T mp .
m=1 k=0
(m,p)=1
T T +1 ×
Clearly
T +1 g − h ∈ C , so it su ces to show that T h ∈ ∆(Fp [[T ]] ). Indeed, we will show
by induction that, for every m ≥ 1, there exists αi ∈ Fp for 1 ≤ i < m such that
Å m−1 ã
T +1 X
hm ..= h− ∆(1 − αi T i ) ∈ T m Fp [[T ]].
T i=1
The case m=1 is empty. Suppose that the claim is true for m and that α1 , . . . , αm−1 have
been chosen. Observe rst that
+∞
T + 1 X k ik
∆(1 − αi T i ) = − iαi T ,
T
k=1
so we can write
+∞
T +1 X
hm = dk T k .
T
k=m
Observe that, by construction of h and hm , we have dn = dnp for all n. If dm = 0 then we
set αm = 0. If dm ̸= 0 then, by what we have just remarked, m must be prime to p, hence
invertible in Fp , and we set αm = − dmm . One can then check that
+∞
Y
g= (1 − αn T n ) ∈ Fp [[T ]]
n=1
T +1
satis es ∆(g) = T h, which concludes the proof.
Corollary 6.27 . The map ∆ : W → Z [[T ]] p
ψ=1
is surjective.
Proof. By Lemma 6.23, it su ces to prove that A = B, which follows directly from
Lemma 6.24 and Lemma 6.25.
The above corollary nishes the proof of Theorem 6.21.
(12) Again, this can be easily checked on measures.
44 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
6.5.2. Proof of the fundamental exact sequence . Finally, we give the fundamental exact
sequence describing the kernel and cokernel of the Coleman map.
De nition 6.28. Z (1)Let p Z
be the module G
p equipped with the action of by σ·x =
χ(σ)x.
Theorem 6.29. The Coleman map induces an exact sequence of G -modules
Col
0 → µp−1 × Zp (1) −→ U∞ −−→ Λ(G) −→ Zp (1) → 0,
where the last map sends µ ∈ Λ(G) to G χ · µ. In particular, it induces an exact sequence
R
Col
0 → Zp (1) −→ U∞,1 −−→ Λ(G) −→ Zp (1) → 0
of Λ(G)-modules.
Proof. The rst map in the composition de ning Col is an isomorphism by Theorem 6.10.
By Theorem 6.21, the second map surjects onto Zp [[T ]]ψ=1 with kernel µp−1 . By Lemma
a
6.30 below, the third map has kernel Zp , which is the image of (1 + T ) for a ∈ Zp , under
a
∆. This is the power series interpolating the sequence (ζpn )n≥1 . Accordingly, when we pull
this back to U∞ , we get the factor
(13)
Zp (1) = {ζpan : a ∈ Zp } ⊂ U∞ .
Finally, the rst two maps in Col are surjective and the third map has cokernel Zp by Lemma
6.30, showing the exactness of the sequence.
Finally, we turn to the G -equivariance. The arguments above show that the map µp−1 ×
Zp (1) → U∞ is G -equivariant. That Col is G -equivariant was Corollary 6.18. The last map
is G -equivariant since
Z Z
χ(x) · σµ(x) = χ(σx) · µ(x)
G G
Z
= χ(σ) χ · µ,
G
and the action of G on Zp (1) is by χ.
Lemma 6.30 . There is an exact sequence
1−φ
0 → Zp → Zp [[T ]]ψ=1 −−−→ Zp [[T ]]ψ=0 → Zp → 0,
where the rst map is the natural inclusion and the last map is evaluation at T = 0.
Proof. Injectivity of the rst map is trivial. Surjectivity of the last map follows, for
example, from the fact that ψ(1 + T ) = 0, since
X
(φ ◦ ψ)(1 + T ) = p−1 ζ(1 + T ) = 0.
ζp
Let f (T ) ∈ Zp [[T ]]ψ=0 be in the kernel of the last map, that is be such that f (0) = 0.
n (14) n
P
Then φ (f ) goes to zero (in the weak topology ) and hence n≥0 φ (f ) converges to an
element g(T ) whose image under (1 − φ) is f (T ). Since ψ ◦ φ = id, we also have
X X
ψ(g) = ψ ◦ φn (f ) = ψ(f ) + φn−1 (f ) = g,
n≥0 n≥1
as ψ(f ) = 0, which shows that
f ∈ (1 − φ) Zp [[T ]]ψ=1
(13) Z (1) ..=
p lim µpn is a free Zp -module of rank 1 on which the absolute Galois group GQ acts by the
←−
cyclotomic character. It is an integral version of Qp (1).
(14) Recall: the weak topology corresponds to the (p, T )-adic topology on Z [[T ]].
p
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 45
and hence that the sequence is exact at Zp [[T ]]ψ=0 . Finally, if f (T ) ∈ Zp [[T ]] is not constant,
thenf (T ) = a0 + ar T r + . . . for some ar ̸= 0 and φ(f )(T ) = a0 + par T r + . . . ̸= f (T ), which
shows that ker(1 − φ) = Zp and nishes the proof.
We conclude this section by the following digression on the generalisations of Coleman's
map that lead to a conjectural construction (under the assumption of the existence of certain
global cohomological elements) of p-adic L-functions of more general motives.
6.6. The Kummer sequence, Euler systems and p-adic L-functions. We end
this section with a disgression about how the picture described here generalizes to more
general contexts. Let GQ be the absolute Galois group of Q and consider, for m ≥ 1, the
Kummer exact sequence
m
x7→xp
0 → µpm → Gm −−−−−→ Gm → 0. (6.1)
Evaluating at Q, this short exact sequence induces, for any number eld F, a long exact
sequence on cohomology
m
x7→xp ×
0 → µpm (F ) → F × −−−−−→ F × → H 1 (F, µpm ) → H 1 (F, Q ).
Here, for any topological GF -module A, H 1 (F, A) ..= H 1 (GF , A) for the Galois
we write
cohomology, that is the continuous group cohomology of GF . By Hilbert 90, we have
H 1 (F, F × ) = 0. Taking inverse limits, which is exact, over m ≥ 1, we obtain
F × ⊗ Zp ∼
= H 1 (F, Zp (1)).
Explicitly, at each nite level, the isomorphism
n
F × ⊗ Z/pn Z = F × /(F × )p −∼→ H 1 (GF , µpn )
× n
is given as follows. Take a ∈ F× and take any b∈Q bp = a. Then ca : σ 7→ σ(b)
such that
b
de nes a 1-coycle on GF and it
n
is a coboundary if and only if a is a p −th root of unity in
×
F , which shows that the map sending the class of a to the class of ca is well de ned.
Let m = Dpn , n ≥ 1, and de ne
−1
ζm −1 ×
cm ..= ∈ OQ(µm)
,
ζm − 1
which is a global analogue and a generalization of the cyclotomic units cn (−1) considered in
Example 6.2. One can show that these elements satisfy the following relations with respect
to the norm maps:
(
cm if ℓ|m
NQ(µmℓ )/Q(µm ) (cmℓ ) =
(1 − ℓ−1 )cm if ℓ ∤ m.
Using the Kummer map described below, we get elements zm ..= ∂(cm ) ∈ H 1 (Q(µm ), Zp (1))
satisfying
(
zm if ℓ|m
coresQ(µmℓ )/Q(µm ) (zmℓ ) = −1
(1 − ℓ )zm if ℓ ∤ m,
where we have used that Frobℓ acts on Zp (1) simply by multiplication by ℓ. Observe also
that (1 − ℓ−1 ) is the Euler factor at ℓ of the Riemann zeta function.
De nition 6.31 .
V ∈ RepL GQ be a global p-adic Galois representation, which is
Let
unrami ed outside a nite set Σ of primes and let T ⊆ V be an OL -lattice stable by GQ .
An Euler system for (V, T, Σ) is a collection of classes
zm ∈ H 1 (Q(µm ), T ), (m, Σ) = {p}
46 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
satisfying
(
zm if ℓ|m
coresQ(µmℓ )/Q(µm ) (zmℓ ) =
Pℓ (V ∗
(1), σℓ−1 )zm if ℓ ∤ m,
I
where
∗
Pℓ (V (1), X) = det(1 − Frobℓ−1 X|V ∗ (1) ℓ ) is the Euler factor at ℓ of the L-function
associated to V ∗ (1) and σℓ denotes the image of Frobℓ in Gal(Q(µm )/Q).
By what we have mentioned before, cyclotomic units form an Euler system for the
representation Zp (1). These elements are at the base of Rubin's proof of the main conjec-
ture. In general, constructing Euler systems for a Galois representation is a very di cult
task, and very few examples exist at the moment. Moreover, there is no actual axiomatic
study of Euler systems allowing us to study the few examples known under the same setting.
In exactly the same way, replacing Q by Qp and F by a nite extension K of Qp , and
observing that K × ⊗ Zp = K × since K × is already p-adically complete, we obtain from
Kummer's exact sequence (6.1) an isomorphism
K× ∼
= H 1 (K, Zp (1)).
Taking K = Kn forn ≥ 1 in the last isomorphism of the above paragraph, and considering
the inverse limit over all n, we see that there is a map
U∞ −→ lim H 1 (Kn , Zp (1)),
←−
n≥1
where the inverse limit is taken with respect to corestriction maps in Galois cohomology.
We de ne Iwasawa cohomology groups by
1
HIw (Qp , Qp (1)) = lim H 1 (Kn , Zp (1)) ⊗Zp Qp .
..
←−
n≥1
The remarks made so far allow one to reinterpret the Coleman map as a map
1
Col : HIw (Qp , Qp (1)) → M (G, Qp ),
where we recall that M (G, Qp ) = Λ(G) ⊗Zp Qp is the space of Qp -valued measures on
G. The cyclotomic units cn (a) we saw earlier form what is known as an Euler system, a
system of (global) Galois cohomology classes that are compatible under corestriction maps.
By localising, the cyclotomic units give rise to an element of the Iwasawa cohomology.
By combining the above with Proposition 6.4, we see that the p-adic zeta function can
be obtained by evaluating Col at this Iwasawa cohomology class (and, as usual, dividing
through by the measure θa to account for the pole).
Let now V ∈ RepL GQp be any p-adic representation of GQp , i.e a nite dimensional L-
vector space V equipped with a continuous linear action of GQp . As before, we de ne its
Iwasawa cohomology groups as
1
HIw (Qp , V ) ..= lim H 1 (Kn , T ) ⊗OL L,
←−
n≥1
where T ⊆V denotes any OL -lattice of V stable under the action of the Galois group GQp ,
and where as before the inverse limit is taken with respect to the corestriction maps in
cohomology. Morally, Iwasawa cohomology groups are the groups where the local parts at p
PR95
of Euler systems of a global p-adic representation live. Assuming that the representation is
(15)
crystalline , the Coleman map has been generalized by Perrin-Riou [ ]. Under some
choices, she constructed big logarithm maps
1
LogV : HIw (Qp , V ) → D(G, L),
(15) Loosely,a p-adic representation of GQp being crystalline is a condition from p-adic Hodge theory that
is equivalent to an ℓ-adic representation of GQp (with ℓ ̸= p) being unrami ed.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 47
where D(G, L) denotes the space of L-valued distributions on G (16)
. The map LogV satis es
certain interpolation properties expressed in terms of Bloch-Kato's exponential and dual
exponential maps and, for V = Qp (1), we reobtain the Coleman map.
The general idea is that, given an Euler system for a global p-adic Galois representation,
localizing it at the place p and applying Perrin-Riou's machine, one can construct a p-adic
L-function for V. In a diagram:
locp 1 LogV
Euler systems −−→ HIw (Qp , V ) −−−→ p-adic L-functions .
See [ Col00 ] for further references on this subject.
7. The Main Conjecture
In this section we will continue moving from the analytic picture we developed in Part I to
a more arithmetic setting. We have already seen that cyclotomic units are intimately related
to the p-adic zeta function, and in this section we will study further properties of the module
generated by the cyclotomic units. In particular, we will consider the p-adic closure of the
cyclotomic units inside the local units, and show a theorem of Iwasawa calculating the exact
image of this closure under the Coleman map (in terms of the p-adic zeta function). This
endows the connection between cyclotomic units and the p-adic zeta function with a deeper
algebraic structure. We will next use class eld theory to relate these modules to Galois
groups, and hence state Iwasawa's Main Conjecture. Under certain assumptions that will
be proved later and, most importantly, under a crucial assumption on the prime p (which
conjecturally always holds), we will provide a proof of the main conjecture.
7.1. The Λ-modules arising from Galois theory. As in the last section, we intro-
duce the Λ-modules that will be the protagonists of the Galois side of the main conjecture,
and we urge the reader to refer back to this as these objects appear in the text. De ne
Mn ..= maximal abelian p-extension of Fn unrami ed outside the unique prime of Fn over p,
Mn+ ..= maximal abelian p-extension of Fn+ unrami ed outside the unique prime of Fn+ over p,
Ln ..= maximal unrami ed abelian p-extension of Fn ,
Ln+ ..
= maximal unrami ed abelian p-extension of Fn+ ,
and set
M∞ ..= ∪n≥1 Mn = maximal abelian p-extension of F∞ unrami ed outside p,
M∞
+ ..
= ∪n≥1 Mn+ = maximal abelian p-extension of
+
F∞ unrami ed outside p,
L∞ ..= ∪n≥1 Ln = maximal unrami ed abelian p-extension of F∞ ,
L∞
+ ..
= ∪n≥1 Ln+ = maximal unrami ed abelian p-extension of
+
F∞ .
Finally, de ne
X∞ ..= Gal(M∞ /F∞ ), X∞
+ +
= Gal(M∞ +
/F∞ ),
Y∞ ..= Gal(L∞ /F∞ ), Y∞
+ +
= Gal(L∞ +
/F∞ ).
These modules t into the following diagram of eld extensions:
(16) Recallthat measures were interpreted as bounded rigid analytic functions on the p-adic weight space.
The space D(G, L) is precisely de ned as (not necessarily bounded) rigid analytic functions. Equivalently,
in terms of p-adic functional analysis, it is the continuous dual of the space of locally analytic functions (i.e
continuous functions that locally admit a power series expansion).
48 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
M∞
X∞
L∞
Y∞
F∞ Mn
Ln
G
Fn
Gn
Q
(There is an identical diagram for the totally real objects, with everything adorned by a
superscript +).
We defer to the appendix an interpretation of these modules in terms of ideles and ideal
class groups, as given by class eld theory.
Remark 7.1 . Recall that G = Gal(F∞ /Q) was de ned at the start of the previous
section. The advantage of considering the whole cyclotomic tower instead of considering
each level individually is that we get in this fashion modules over the Iwasawa algebras
Λ(G) = OL [[G]] and Λ(G + ) = OL [[G + ]], whose structure is simpler than that of modules
over OL [Gn ] (resp. OL [Gn ]). We describe this action: take elements x ∈ X∞ , σ ∈ G and
+
choose any lifting σ̃ ∈ Gal(M∞ /Q) of σ , then
σ · x ..= σ̃xσ̃ −1
gives a well de ned action of G on X∞ which extends by linearity and continuity to an
action of Λ(G) on X∞ (exercise: check these claims). In exactly the same way we de ne
actions of Λ(G) on Y∞ and of Λ(G ) on X∞ and Y∞ . For further details on the structure
+ + +
of Λ-modules, where Λ is one of the Iwasawa algebras above, see the appendix.
7.2. Measures on Galois groups. In the previous section, the fact that we can use
the cyclotomic character to see the p-adic zeta function in terms of measures on G was
heavily trailed, and we even took the Coleman map to have values in Λ(G). In the process,
we gave a conceptual reason for the twist by 1 introduced in the de nition of ζp , which
ensured that the Coleman map was G -equivariant. We now elaborate on this and further
conceptually explain why this twist is introduced in the context of the main conjecture.
Recall that F∞ = ∪n≥1 Q(µpn ), that G = Gal(F∞ /Qp ) ∼
= Z×
p via the cyclotomic charac-
ter, and that this isomorphism induces an identi cation of measures on Z×
p and measures on
the Galois group G , as has already been used in the de nition of the Coleman map. Also,
G+ = Gal(F∞ /Q) = G/⟨c⟩ is identi ed through the cyclotomic character with Z×
+
p /{±1}.
Remark 7.2 .
Let GQ = Gal(Q/Q) denote the absolute Galois group of Q. There is a natural projec-
tion GQ → G given by restriction to F∞ , and if we compose χ with this projection, we get
a map
χ : GQ −→ Z×
p
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 49
that we continue to call the cyclotomic character. This allows us to de ne a Galois repre-
sentation
χ : GQ −→ GL(V ),
where V is a 1-dimensional Qp -vector space, under Z× ×
p ⊂ Qp = GL(V ). We write V =
Qp (1) for this Galois representation. Recall from the introduction that, whenever we have
a global Galois representation, we can construct a complex L-function de ned as an Euler
product, and note that
L(Qp (1), s) = ζ(s + 1),
so rescaling the p-adic or complex zeta function corresponds to twisting the Galois repre-
sentation.
Note that ζp is precisely the p-adic L-function of the Galois representation Qp (1), and
this twist by 1 corresponds to the fact that we get ζ(s + 1), not ζ(s), hence the interpolation
of the values ζ(1 − k). The main conjecture (as we will state it) can be viewed as a precise
relation between the Selmer group and the p-adic L-function of Qp (1), so it is more natural
in this context to include the twist by 1.
7.2.1. Passing to G +
. Now observe that ζp , which ostensibly is an element of Q(G),
vanishes at the characters χk , for any odd integer k > 1. We will use this fact to show that
ζp actually descends to a pseudo-measure on G+.
Lemma 7.3 . Let c ∈ G denote the action of complex conjugation. Let R be a ring in
which 2 is invertible and M an R-module with a continuous action of G . Then M decomposes
as
M∼ = M + ⊕ M −,
where c acts as +1 on M + and as −1 on M − .
Proof. We prove this directly by using the idempotents
1+c 1−c
2 and 2 , which act as pro-
+ −
jectors to the corresponding M and M .
Since we are assuming that p is odd, we see that Λ(G) ∼
= Λ(G)+ ⊕ Λ(G)− (as Λ(G)-
+
modules). In fact, the module Λ(G) admits a description solely in terms of the quotient
G+.
Lemma 7.4 . There is a natural isomorphism
Λ(G)+ ∼
= Λ(G + ).
Proof. We work at nite level. Let Gn ..= Gal(Fn /Q), and Gn+ ..= Gal(Fn+ /Q). Then
there is a natural surjection
Zp [Gn ] → Zp [Gn+ ]
induced by the natural quotient map on Galois groups. Since this must necessarily map
− + +
Zp [Gn ] to 0, this induces a map Zp [Gn ] → Zp [Gn ]. The result now follows at nite level
n−1
by a dimension count (as both are free Zp -modules of rank (p − 1)p /2). We obtain the
required result by passing to the inverse limit.
We henceforth freely identify Λ(G + ) with the submodule Λ(G)+ of Λ(G).
Lemma 7.5 . Let µ ∈ Λ(G). ThenZµ ∈ Λ(G +
) if and only if
χ(x)k · µ = 0
G
for all odd k ≥ 1.
50 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. By Lemma 7.3, we can write µ = µ+ + µ− , where µ± = 1±ι 2 µ. We want to show
− k
R
that µ = 0 if and only if
G
χ(x) · µ = 0 for all odd k ≥ 1. Since χ(c) = −1, we have
Z ÅZ Z ã
1
χ(x)k · µ− = χk · µ − (−1)k χk · µ .
G 2 G G
If k is even, the above expression vanishes. The result follows then by Lemma 3.8.
Corollary 7.6 . The p-adic zeta function is a pseudo-measure on G +
.
Proof. This follows directly from the interpolation property, as ζ(1 − k) = 0 precisely
when k≥2 is odd.
7.3. The main conjecture. It is natural to ask about the zeroes of the p-adic zeta
function. Since the zeroes are not modi ed if we multiply by a unit, studying the zeroes of
a measure is equivalent to studying the ideal generated by it. Recall that, even if ζp is not
a measure but a pseudo-measure, the elements ([σ] − 1)ζp , σ ∈ G , belong to the Iwasawa
algebra Λ(G).
De nition 7.7 . Let I(G) denote the augmentation ideal of Λ(G), that is, the ideal
I(G) = OL -span of {[σ] − 1 : σ ∈ G}.
Similarly, de ne I(G + ) to be the OL -span of [σ] − 1 for σ ∈ G+.
Remark 7.8 . The Iwasawa algebra Λ(G) is the completed group ring
P
OL [[G]], and
elements can be written as `power series' of the form g∈G cg [g]. There is a natural degree
map
deg : Λ(G) −→ OL ,
X X
cg [g] 7−→ cg ,
g∈G g∈G
and I(G) is simply the kernel of this map.
Proposition 7.9 . The module I(G)ζ p is an ideal in Λ(G). Similarly, the module I(G + )ζp
is an ideal in Λ(G ). +
Proof. Since ζp is a psuedo-measure, we know ([g] − [1])ζp ∈ Λ(G) for all g ∈ G . Hence
the result follows from the de nition of I(G). The same argument holds for I(G + )ζp .
Recall that M∞
+
p-extension of F∞
denotes the maximal abelian
+
which is unrami ed
outside the unique prime of F∞ above p and X∞ = Gal(M∞ /F∞ ) is endowed with an
+ + + +
+ +
action of Λ(G ). Recall the de nition of the characteristic ideal chΛ(G + ) (X) of a Λ(G )-
module X, as given in the appendix. We have then:
Theorem 7.10 (Iwasawa Main Conjecture). The module X +
∞ is a nitely generated
torsion Λ(G )-module and we have
+
+
chΛ(G + ) (X∞ ) = I(G + )ζp .
Remark 7.11 . It is usual in the literature to formulate Iwasawa main conjecture in
terms of an even Dirichlet character of Gal(Q(µp )/Q). As one can already observe from the
behaviour of the Bernoulli numbers, there exists a certain dichotomy involving the parity of
this character which makes the formulation of the main conjecture dierent in the even and
HK03
odd cases. The above formulation takes into account every such even Dirichlet character.
For a formulation of the main conjecture for odd Dirichlet characters, see [ ].
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 51
We intend to provide a proof of the main conjecture for primes not dividing the class
number of the eld Q(µp )+ . The arguments of this section form the origins of Iwasawa's
formulation of the main conjecture, and should also be seen as motivation for it. Its complete
proof involves much more sophisticated techniques: there are at least two proofs of it, each
CS06
one showing one divisibility between the two modules and invoking the analytic class number
formula to deduce the other one. One of them (explained in [ ]) uses the theory of Euler
systems, and we hope that this text will facilitate the reader's eventual study of this method.
7.4. Cyclotomic units. We now return to cyclotomic units, and in particular study
the subgroup generated by them in the (local and global) unit groups. In the global case,
this subgroup has nite index in the whole unit group Vn . Since the determination of the
units of a number eld is in general a di cult problem, and cyclotomic units provide a
partial answer in the case of cyclotomic elds, they are objects of classical interest and have
been extensively studied.
De nition 7.12 . For n ≥ 1, we de ne the group Dn of cyclotomic units of Fn to be
the intersection of OF×n and the multiplicative subgroup of Fn× generated by {±ζpn , ζpan − 1 :
1 ≤ a ≤ pn − 1}. We set Dn+ = Dn ∩ Fn+ .
Recall we de ned
ζpan − 1
cn (a) ..= ∈ Dn ,
ζpn − 1
and note that
(1−a)/2
ξn,a ..= ζpn cn (a)
is an element of Dn . In fact:
+
Lemma 7.13 . Let n ≥ 1. Then
(1) The group Dn+ is generated by −1 and
pn
ß ™
ξn,a : 1 < a < , (a, p) = 1 .
2
(2) The group Dn is generated by ζpn and Dn+ .
Proof. We rst show that we need only consider those elements ζpan − 1 with a prime to
p. Indeed, this follows from the identity
m
pY −1
m n−k
ζpbpn = (ζpb+jp
n − 1),
j=0
where (b, p) = 1 and k ≥ 1, and noting that b + jpn−k is prime to p. Also, since
ζpan − 1 = −ζpan (ζp−a
n − 1), we can restrict to considering 1 ≤ a ≤
1 n
2p .
So suppose that
Y
ξ = ±ζpdn (ζpan − 1)ea ∈ Dn ,
n
1≤a< 12 p
(a,p)=1
for some integersd and ea . Since vp (ζpdn ) = 0 and all the p-adic valuations of ζpan − 1 coincide
a 1
P
(namely, vp (ζpn− 1) = (p−1)p n−1 ), we deduce that a ea = 0. Therefore we can write
Y ζpan − 1 ea
Å ã Y
ξ = ±ζpdn = ±ζpen ea
ξn,a ,
ζ p n − 1
a a
1
P
where e = d +
2 a ea (a − 1). This shows the second point and the rst point follows by
of the product is real, so ξ ∈ Dn if and only if e = 0.
ea +
observing that every term ξn,a
Corollary 7.14 . Let a be a generator of (Z/p Z) . Then ξ n ×
a generates Dn+ as a Z[Gn+ ]-
module.
52 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof. Indeed, for any integer 1 ≤ b < pn prime to p, b ≡ ar (mod p) for some r ≥ 0,
and hence
r−1
Y i
ξn,b = (ξn,b )σb .
i=0
As we have already suggested, the group Dn (resp. Dn+ ) of cyclotomic units is of nite
index in the group of units Vn (resp. Vn ) of
+
Fn (resp. Fn+ ), and this index turns out to be
a class number.
De nition 7.15 . For n ≥ 1, we write
h+ . +
n .= #Cl(Fn )
for the class number of Fn+ .
Proposition 7.16 . Let n ≥ 1. The group D n (resp. Dn+ ) is of nite index in the group
of units of Fn (resp. Fn+ ) and we have
n = [Vn : Dn ] = [Vn : Dn ]
h+ + +
Proof. The result follows by showing that the regulator of cyclotomic units is given in
Was97
terms of special L-values at s=1 of Dirichlet L-functions and by the class number formula.
See [ , Theorem 8.2].
7.5. On a theorem of Iwasawa. We will calculate the image under the Coleman map
of the p-adic closure of the module of cyclotomic units inside local units.
De nition 7.17 . For any n ≥ 1, de ne Cn as the p-adic closure of Dn inside the local
units Un , let Cn+ ..= Cn ∩ Un+ , and let
Cn,1 ..= Cn ∩ Un,1 , Cn,1
+ ..
= Cn+ ∩ Un,1 ;
C∞,1 ..= lim Cn,1 , C∞,1
+ ..
= lim Cn,1
+
.
←− ←−
n≥1 n≥1
Remark 7.18 . As we have pointed out whilst de ning notation at the beginning of the
previous chapter, the process of considering elements congruent to 1 modulo the maximal
ideal and taking the p-adic closure allows us to consider the modules C∞,1 and C∞,1 as Λ(G)
+
+
and Λ(G )-modules respectively.
Observe that, as a consequence of Corollary 7.14, we easily deduce the following result.
Lemma 7.19 . The module C
is a cyclic Λ(G + )-module generated by (uξn,a )n≥1 ,
+
∞,1
where a ∈ Z is a topological generator of Z×p (for example, take a to be a primitive root
modulo p such that ap−1 ̸= 1 (mod p)) and u ∈ µp−1 is such that au ≡ 1 (mod p).
Here is the main result of this section, putting the link between cyclotomic units and the
p-adic zeta function into a deeper algebraic structure.
Theorem 7.20 . The Coleman map induces:
(1) An isomorphism of Λ(G + )-modules
∼
U∞,1
+ +
/C∞,1 → Λ(G + )/I(G + )ζp .
−
(2) A short exact sequence of Λ(G)-modules
0 → U∞,1 /C∞,1 → Λ(G)/I(G)ζp → Zp (1) → 0.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 53
Proof. Consider the exact sequence of G -modules of Theorem 6.29:
Col
0 → µp−1 × Zp (1) −→ U∞ −−→ Λ(G) −→ Zp (1) → 0.
Since U∞ = µp−1 × U∞,1 , we can rewrite the above as
Col
0 → Zp (1) −→ U∞,1 −−→ Λ(G) −→ Zp (1) → 0.
Every module in this sequence is now a Zp -module, and moreover Col is Zp -equivariant.
Thus this is an exact sequence of Λ(G)-modules. The theorem will follow by calculating the
image of the modules C∞,1 and C∞,1
+
under the Coleman map.
By Lemma 7.13, it su ces to calculate the image under Col of an element (ζpbn ξn,a )n≥1 ∈
U∞,1 , for a, b ∈ Z× p . But this has already been done: by Proposition 6.4, and the fact that
ζpbn lies in the kernel of the Coleman map, we know that
Ä (1−a)/2 ä
Col (ζpbn ξn,a )n≥1 = Col ζpn
(ξn,a )n≥1 = Col c(a) = ([σa ] − 1)ζp ,
where as usual σa denotes an element of G such that χ(σa ) = a. Since a ∈ Z×
p was arbitrary,
we conclude that the image of C∞,1 (resp. C∞,1 ) under
+
Col is I(G)ζp (resp. I(G + )ζp ). We
deduce an exact sequence
0 → U∞,1 /C∞,1 −→ Λ(G)/I(G)ζp −→ Zp (1) → 0.
This shows the second point. Since p is odd, taking invariants under the group ⟨c⟩ ⊂ G of
order two generated by complex conjugation is exact. As c acts on Zp (1) by −1, we see that
Zp (1)⟨c⟩ = 0, which shows the rst point and concludes the proof of the theorem.
Remark 7.21 . For the purposes of the main conjecture, we will be restricting to the
rst point of Theorem 7.20 above.
7.6. An application of class eld theory. We next use class eld theory to reinter-
pret Theorem 7.20 in terms of some modules arising from Galois theory.
De nition 7.22. n≥1For any E , de ne n as the p-adic closure of Vn inside the local
units Un , let En+ ..= En ∩ Un+ , and let
En,1 ..= En ∩ Un,1 , En,1
+ ..
= En+ ∩ Un,1 ;
E∞,1 ..= lim En,1 , E∞,1
+ ..
= lim En,1
+
.
←− ←−
n≥1 n≥1
We have the following result, connecting units in the cyclotomic tower and modules
coming from Galois theory.
Proposition 7.23 . There is an exact sequence of Λ(G +
)-modules
0→ E∞,1
+
→ U∞,1
+
→ +
Gal(M∞ +
/L∞ ) → 0.
Proof. By Proposition B.5 of the appendix, we know that, if Mn+ (resp. Ln+ ) denotes
+
the maximal abelian p-extension of Kn unrami ed outside p (resp. everywhere), then
Gal(Mn /Ln ) = Un,1 /En,1 . This gives an exact sequence
+ +
0 → En,1
+
→ Un,1
+
→ Gal(Mn+ /Ln+ ) → 0.
By taking inverse limits, which is exact since all modules in the short exact sequence above
are nitely generated Zp -modules (and hence satisfy the Mittag-Le er condition), we deduce
the result.
+
Having in mind that the Coleman map induces an isomorphism of Λ(G )-modules between
U∞,1 /C∞,1 and Λ(G )/I(G )ζp , we rewrite the above result as follows.
+ + + +
54 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Corollary 7.24 . We have an exact sequence of Λ(G)-modules
0 → E∞,1
+ +
/C∞,1 → U∞,1
+ +
/C∞,1 → X∞
+
→ Y∞
+
→ 0.
Proof. This is an immediate consequence of Proposition 7.23.
7.7. Some consequences of Iwasawa theory. We will now state some classical re-
sults from Iwasawa theory that will be proved later and show how we can deduce from them
the main conjecture assuming that the prime p does not divide the class number of the eld
Q(µp )+ .
Proposition 7.25 . If p ∤ h +
1 n for any n ≥ 1.
then p ∤ h+
Proof. Recall that Cl(F1+ )⊗Z Zp = Gal(L1+ /F1+ ) and p ∤ #Cl(F1+ ) translates into L1+ =
F1+ . By the control result of lemma 8.7 of the next section, we have, for all n ≥ 0,
+
(Y∞ )Gn+ = Gal(Ln+ /Fn+ ),
where we recall that Gn+ = Gal(F∞
+
/Fn+ ). We deduce that, if p ∤ h+ +
1 , then (Y∞ )G0 = 0. By
that Y∞ = 0. We conclude that
+
Nakayama's lemma (cf. Lemma 8.8 below), this implies
Gal(Ln+ /Fn+ ) = 0 for all n ̸= 1, i.e., p ∤ h+
n, which nishes the proof.
Corollary 7.26 . Assume p ∤ h . Then +
1
Y∞
+
= E∞,1
+ +
/C∞,1 = 0.
Proof. We have already seen in the proof of Proposition 7.25 that Y∞ +
= 0. We now
show that E∞,1 In Proposition 7.16 we saw that [Vn : Dn ] = hn , which is
C∞,1
+ + + + +
= .
prime to p by the previous proposition. Note now that Dn,1 and Vn,1 are the kernels of the
+ +
reduction maps modulo p; moreover, the image of Dn (mod p) is contained inside the image
+
of Vn (mod p) ⊂ Fp , so we conclude that the index of Dn,1 inside Vn,1 divides (p − 1)hn .
+ × + + +
Hence there is an exact sequence
0 → Dn,1
+
→ Vn,1
+
→ Wn → 0,
where Wn is a nite group of order prime to p by hypothesis. Tensoring the above exact
sequence with Zp , we get
Dn,1
+
⊗Z Zp ∼
= Vn,1
+
⊗Z Z p .
Recall now that En,1
+
(resp. Cn,1
+
) is by de nition the p-adic closure of Vn,1
+
(resp. Dn,1 )
+
insideUn,1
+
, and that Cn,1
+
⊆ En,1
+
. Since we have
natural surjections Dn,1
+
⊗Z Zp → Cn,1
+
and Vn,1 ⊗Z Zp → En,1 , we conclude that the inclusion Cn,1 → En,1 is a surjection, which
+ + + +
nishes the proof.
We can now easily nish the proof of Iwasawa Main Conjecture when p ∤ h+
1.
Theorem 7.27 . if p ∤ h , we have an isomorphism of Λ(G
+
1
+
)-modules
X∞
+ ∼ +
= Λ(G )/I(G )ζp . +
In particular, Iwasawa's main conjecture holds.
Proof. By Corollary 7.24 and Corollary 7.26, we have
+ ∼
X∞ = U∞,1
+ +
/C∞,1 .
+ ∼
By Theorem 7.20, we have U∞,1
+
/C∞,1 = Λ(G + )/I(G + )ζp , and we deduce
+ ∼
X∞ = Λ(G + )/I(G + )ζp .
In particular,
+
) = chΛ(G + ) Λ(G + )/I(G + )ζp = I(G + )ζp ,
chΛ(G + ) (X∞
which nishes the proof.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 55
Remark 7.28 . A prime p such that p ∤ h+
1 is called a Vandiver prime. Conjecturally,
every prime is a Vandiver prime, and under this conjecture we have proved the full Iwasawa
main conjecture. (Contrast to the less strong condition of being a regular prime, that is,
a prime such that p ∤ h1 . Whilst irregular primes are somewhat rare, they do exist, and
moreover there are in nitely many of them).
MW84
The conditional proof above was due to Iwasawa himself. The rst unconditional proof
of the Iwasawa main conjecture was given by MazurWiles [ ]. For a description of a
CS06
dierent unconditional proof, using Euler systems and due to Kolyvagin, Rubin and Thaine,
see [ ].
8. Iwasawa's µ-invariant
We turn nally to the study of some classical Iwasawa theory. We introduce the µ and
λ-invariants of a Zp -extension. In proving Iwasawa's theorem on the µ and λ-invariants,
we develop techniques that can be used to show that the modules appearing in the exact
sequence of Corollary 7.24 are nitely generated torsion modules over the Iwasawa algebra.
This completes the proof of the main conjecture given in the last section for a Van-
diver prime. (Other than this rather peripheral appearence, however, the main conjecture
does not appear again in this section, which is largely independent of the rest of these notes).
The following results will hold for an arbitrary Zp -extension of number elds, although
we will only prove them under some hypotheses that slightly simplify the proofs.
De nition 8.1 . Let F be a number eld. A Zp -extension F∞ of F is a a Galois
extension such that Gal(F∞ /F ) ∼
= Zp .
If F∞ /F Zp -extension, we denote Fn the sub-extension xed by the unique subgroup
is a
n
of Γ with quotient Z/p /Z . Recall rst that any number eld has at least one Zp -extension,
the cyclotomic extension. Indeed, consider the elds F (µpn ), and let
[
F (µp∞ ) = F (µpn ).
n≥1
By Galois theoryGal(F (µp∞ )/F ) is an open subgroup of Gal(Q(µp∞ )/Q) ∼ = Z× p , and hence
contains a maximal quotient isomorphic to Zp (speci cally, the quotient by the nite torsion
subgroup). The corresponding eld (under the fundamental theorem of Galois theory) is
the cyclotomic Zp -extension.
De nition 8.2 . Let F∞ /F be a Zp -extension. For each n, let Fn be the unique subex-
tension of F∞ /F such that
Gal(Fn /F ) ∼
= Z/pn Z.
Example 8.3 . Let F = Q(µp ). Then F∞ = Q(µp∞ ) is the cyclotomic Zp -extension of
F, and
Fn = Q(µpn+1 ).
(Note that earlier we denoted this eld Fn+1 ). The cyclotomic Zp -extension of Q is the eld
µ
F∞p−1 , the xed eld in F∞ of the torsion subgroup µp−1 ⊂ Gal(F∞ /Q).
Leopoldt's conjecture states that the number of independent Zp -extensions of a number
eld F is exactly r2 + 1 , where r2 is the number of complex embeddings of F . In particular,
the conjecture predicts that any totally real number eld possesses a unique Zp -extension
(the cyclotomic one). Whilst the conjecture remains open for general number elds, it is
NSW99
known in the case that F is an abelian extension of Q or and abelian extension of an
imaginary quadratic eld (See [ , Theorem 10.3.16]).
56 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
8.1. Iwasawa's
∼
theorem. Let F F∞ /F
Zp -extension, Γ = ΓF =
be a number eld, a
Gal(F∞ /F ) = Zp and γ0 a topological generator of ΓF . Using this choice of γ0 , we identify
Λ(Γ) with Λ ..= Zp [[T ]] by sending γ0 to T + 1 (when γ0 is sent to 1 by the isomorphism
Γ∼= Zp , this is simply the Mahler transform, but this identi cation holds for any γ0 ). Let
Ln (resp. L∞ ) be the maximal unrami ed abelian p-extension of Fn (resp. F∞ ), write
YF,n = Yn ..= Gal(Ln /Fn ) = Cl(Fn ) ⊗ Zp ,
which is the p-Sylow subgroup of the ideal class group of Fn . Set
Y∞ = YF,∞ ..= lim YF,n .
← −
n
Write en = vp (#Yn ) for the exponent of p in the class number of Fn . The following theorem
is the main result we intend to show in this section.
Theorem 8.4 (Iwasawa). There exists an integer n 0 and integers λ ≥ 0, µ ≥ 0, ν ≥ 0,
all independent of n, such that, for all n ≥ n0 , we have
en = µpn + λn + ν.
Remark 8.5 .
This is another typical example of the power of Iwasawa theory, in which we derive
information at nite levels by considering all levels simultaneously. There are two basic steps
on the proof of Theorem 8.4. We rst show that the module YF,∞ is a nitely generated
torsion Λ(Γ)-module. Using the structure theorem of Λ(Γ)-modules (in the appendix), we
study the situation at in nite level, and then we transfer the result back to nite level to
get the result.
We will only describe the proof for the case where the extension F∞ /F satis es the
following hypothesis: there is only one prime p of F above p, and it rami es completely
Was97
in F∞ . The reduction of the general case to this case is not di cult, and is contained
in [ , 13]. This assumption covers our cases of interest; in particular, it applies if
F = Q(µpm ) or F = Q(µpm )+ for some m≥0 and F∞ /F is the cyclotomic Zp -extension.
8.1.1. First step . The rst step of the proof of Theorem 8.4 consists in showing (Propo-
sition 8.9) that the module Y∞ is a nitely generated Λ(Γ)-module. Then Lemma 8.7
will allow us to recover each Yn from the whole tower Y∞ . We then use a variation of
Nakayama's lemma to conclude.
Since p is totally rami ed in F∞ , and Ln is unrami ed over Fn , we deduce that Fn+1 ∩
Ln = Fn and hence
Yn = Gal(Ln /Fn ) = Gal(Ln Fn+1 /Fn+1 ) = Yn+1 /Gal(Ln+1 /Ln Fn+1 ),
showing that Yn+1 surjects onto Yn . The module Y∞ is equipped with the natural Galois
action of Λ = Λ(Γ), and under the identi cation Λ ∼
= Zp [[T ]], the polynomial 1 + T ∈ Λ
acts as γ0 ∈ Γ.
Let p̃ be the prime of L∞ above p, and write
I ⊆ G ..= Gal(L∞ /F )
L∞ /F∞ is unrami ed, all of the inertia occurs in the subextension
for its inertia group. Since
F∞ /F . Accordingly I ∩ Y∞ = 1 and since F∞ /F is totally rami ed at p, the inclusion
I ,→ G/Y∞ ∼= Γ is surjective, and hence bijective. We deduce that
G = IY∞ = ΓY∞ .
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 57
We've shown the following picture of extensions.
L∞
Y∞
F∞
Ln
I∼
=Γ∼
= Zp
Yn G=IY∞
Fn
Z/pn Z
F
Let σ∈I map to the topological generator γ0 ∈ Γ under the natural isomorphism I∼
= Γ.
Lemma 8.6 . Let G be the closure of the commutator of G. Then
′
G′ = (γ0 − 1) · Y∞ = T Y∞ .
Proof. Recall that we have a decomposition G = ΓY∞ . Let a = αx, b = βy ∈ G, where
α, β ∈ Γ and x, y ∈ Y∞ . A straightforward calculation, using the de nition of the Λ(Γ)
structure of Y∞ , shows that
aba−1 b−1 = (xα )1−β (y β )α−1 .
Setting β = 1 and α = γ0 , we deduce that (γ0 − 1)Y∞ ⊆ G′ . To see the other inclusion,
c
P+∞ c n
P+∞ c n
write β = γ0 , where c ∈ Zp , so that 1 − β = − n=1 n (γ0 − 1) = − n=1 n T ∈ T Λ
and similarly for α − 1, which allows us to conclude.
Recall that the nth power of the Frobenius operator on Zp [[T ]] is given by φn (T ) =
pn
(1 + T ) − 1. Let φ0 (T ) = T .
Lemma 8.7 . We have
Yn = Y∞ /φn (T ).
Proof. We treat rst the casen = 0. Since L0 is the maximal unrami ed abelian p-
extension ofF and L∞ /F is a p-extension, L0 /F is the maximal unrami ed abelian subex-
tension of L∞ . In particular, Y0 = Gal(L0 /F ) is the quotient of G by the subgroup
′
generated by the commutator G and by the inertia group I of p. By the above lemma and
the decomposition G = IY∞ , we conclude that
Y0 = G/⟨G′ , I⟩
= Y∞ I/⟨(γ0 − 1)Y∞ , I⟩
= Y∞ /(γ0 − 1)Y∞ = Y∞ /T Y∞ .
For n ≥ 1, we apply the arguments of the last paragraph, replacing F by Fn and γ0 by
pn n
γ0 , so that σ0 becomes σ0p and (γ0 − 1)Y∞ becomes
n n
(γ0p − 1)Y∞ = ((1 + T )p − 1)Y∞ = φn (T )Y∞ ,
which gives the result.
We state next a variation of Nakayama's lemma for testing when is a Λ-module nitely
generated, whose standard proof is left as an exercise.
Lemma 8.8 (Nakayama's lemma). Let Y be a compact Λ-module. Then Y is nitely
generated over Λ if and only if Y /(p, T )Y is nite. Moreover, if the image of x1 , . . . , xm
generates Y /(p, T )Y over Z, then x1 , . . . , xn generate Y as a Λ-module. In particular, if
Y /(p, T )Y = 0, then Y = 0.
58 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Applying this in our particular situation we obtain the following result.
Proposition 8.9 .Y ∞ is a nitely generated Λ-module.
Proof. Since
p Ç å
p
X p
φ(T ) = (1 + T ) − 1 = T k ∈ (p, T ),
k
k=1
the module Y∞ /(p, T )Y∞ is a quotient of Y∞ /φ(T )Y∞ = Y1 = Cl(F1 ) ⊗ Zp , the p-Sylow
subgrop of Cl(F1 ), which is nite. Therefore, applying Lemma 8.8, we conclude that Y∞ is
a nitely generated Λ-module, as desired.
8.1.2. Second step . Once we know that the module Y∞ is a nitely generated Λ-module,
we can invoke the structure theorem for these modules (Theorem C.5) to get an exact
sequence
0 → Q → Y∞ → A → R → 0,
where Q and R are nite modules and where
s
ÅM t
ã ÅM ã
A =Λ ⊕ r
Λ/(p mi
) ⊕ kj
Λ/(fj (T ) ) .
i=1 j=1
for some integers s, r, t ≥ 0, mi , kj ≥ 0 and some distinguished polynomials fj (T ) ∈ Λ.
Recall that we want to calculate the size of Yn = Y∞ /φn (T ). The following lemma
reduces the problem to calculating the size of A /φn (T ).
Lemma 8.10 . There exists a constant c and an integer n 0 such that, for all n ≥ n0 ,
|Y∞ /φn (T )| = pc |A /φn (T )|.
Proof. Consider the diagram
0 φn (T )Y∞ Y∞ Y∞ /φn (T )Y∞ 0
0 φn (T )A A A /φn (T )A 0
By hypothesis, the kernel and cokernel of the middle vertical map are bounded. By
elementary calculations and diagram chasing, one ends up showing that the kernel and the
cokernel of the third vertical arrow stabilize for n large enough, which is what is needed to
conclude the proof. We leave the details of these calculations as an exercise.
We now proceed to calculate the size of the module A.
Lemma 8.11 . Let s
ÅM t
ã ÅM ã
A = Λr ⊕ Λ/(pmi ) ⊕ Λ/(fj (T )kj ) ,
i=1 j=1
for some integers s, r, t ≥ 0 and mi , kj ≥ 1 and some distinguished polynomials fj (T ) ∈ Λ,
and write m = mi , ℓ = kj deg(fj ). Suppose A /φn (T )A is nite for all n ≥ 0. Then
P P
r = 0 and there exist constants n0 and c such that, for all n ≥ n0 ,
n
|A /φn (T )| = pmp +ℓn+c
.
Proof. A /φn (T ) is assumed to be nite and Λ/φn (T ) is in nite
First observe that, since
(use the division algorithm of Proposition C.2), we deduce that r = 0.
k n
We now deal with the second summand. Let V = Λ/p for some k ≥ 1. Since φ (T ) =
pn
Ppn −1 pn k
T + k=1 k T is distinguished, we have
n n
|V /φn (T )| = |Λ/(pk , T p )| = pkp ,
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 59
where the last equality follows again by the division algorithm of Proposition C.2. We
deduce from this that
s
M n
Λ/(pmi ) = pmp ,
i=1
P
where m= i mi .
Finally, we deal with the last summand. Let g(T ) ∈ OL [T ] be a distinguished poly-
nomial of degree d V = Λ/(g(T )). Hence
(that is not necessarily irreducible) and let
T d ≡ pQ(T ) mod for some Q ∈ OL [T ] so that T k ≡ p (poly) mod g for all k ≥ d, where
g
(poly) denotes some polynomial in OL [T ]. For pn ≥ d, we deduce that
n
φn (T ) = p (poly) + T p ≡ p (poly) mod g.
n+1 2
φ (T ) ≡ p (poly) mod g.
n+1 n+1
φn+2 (T ) = ((1 + T )(p−1)p + . . . + (1 + T )p + 1)φn+1 (T )
≡ p(1 + p (poly))φn+1 (T ) mod g.
n+2
φ (T )
Since ((1 + p (poly)) ∈ Λ× , we deduce that
φn+1 (T ) acts as p times a unit on V = Λ/(g(T ))
and hence
φn+2 (T )V = pφn+1 (T )V.
Therefore
|V /φn+2 (T )V | = |V /pV | |pV /pφn+1 (T )V |.
Since g(T ) is distinguished of degree d, we have
|V /pV | = |Λ/(p, g(T ))| = |Λ/(p, T d )| = pd .
Finally, we compute |pV /φn+1 (T )V |. Since (g(T ), p) = 1, multiplication by p is injective on
n+1
V and hence |pV /pφ (T )V | = |V /φn+1 (T )V |. Fix one n0 such that pn0 ≥ d. Then, using
the identity
φn+1 (T ) φn0 +2 (T )
φn+1 (T ) = n
◦ . . . ◦ n0 +1 ◦ φn0 +1 (T )
φ (T ) φ (T )
φk+1 (T )
and the fact that
φk (T )
act on V as p (unit) for any k > n0 , we deduce that φn+1 (T ) acts
(n−n0 −1) n0 +1
on V as p φ (T ) and hence
|V /φn+1 (T )V | = pd(n−n0 −1) |V /φn0 +1 (T )V |.
Putting everything together, we deduce that
|V /φn (T )V | = pnd+c ,
for some constant c and all n > n0 . Applying this to the third summand of A, we get
t
M
Λ/(fj (T )kj ) = pℓn+c ,
j=1
P
where ℓ= j kj deg(fj ) and some constant c. This nishes the proof of the proposition.
Along the way, we have proven the following fact.
Corollary 8.12 . Let Y be a nitely generated Λ-module. If Y /φn (T )Y is nite for
all n, then Y is torsion.
Proof. If A r = 0 in
is as in the statement of Proposition 8.11, then we showed that
the structure theorem for Y . This implies that A is torsion; each element is annihilated
by the characteristic ideal of A . If Y is any nitely generated Λ-module, then Y is quasi-
isomorphic to a module A as before, and as A is torsion, so is Y .
We can now complete the proof of Theorem 8.4.
60 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Proof of Theorem 8.4. Applying Lemma 8.10 and Lemma 8.11, we get
n
|Yn | = |Y∞ /φn (T )Y∞ | = pc |A /(φn (T ))| = pµp +λn+ν
.
This nishes the proof of the theorem.
8.2. Some consequences of Iwasawa's theorem. We have already seen one appli-
cation of Iwasawa's theorem (Proposition 7.25) during the statement of the main conjecture.
Namely if one class number in a Zp -extension is coprime to p, then so are all the others.
We list here some further interesting applications.
Recall that if A is a nite abelian group, then
A[p] ..= {x ∈ A : px = 0}
denotes the subgroup of p-torsion elements and its p-rank rkp (A) is de ned to be
rkp (A) = dimFp (A/pA) = dimFp (A[p]).
Equivalently, we can decompose A uniquely as a direct sum of cyclic groups of prime power
order; then the rank at p is the number of direct summands of p-power order.
Corollary 8.13 .
Let F∞ /F be a Zp -extensions. Then µ = 0 if and only if rkp (Cl(Fn ))
is bounded independently of n.
Proof. Recall that
Cl(Fn ) ⊗ Zp = Yn ..= Y∞ /(φn (T )),
Ls Lt
that Y∞ = lim Yn is quasi-isomorphic to a Λ-module A =
mi
i=1 Λ/(p ) ⊕ j=1 Λ/(gj (T ))
←−
for some integers s, t ≥ 0, mi ≥ 1, and gi (T ) ∈ OL [T ] distinguished polynomials, and that
we have (cf. the proof of Lemma 8.11) an exact sequence
0 → Cn → Yn → An → Bn → 0,
where An ..= A /φn (T ), with |Bn | and |Cn | bounded independently of n. It su ces then to
show that µ = 0 if and only if dimFp (An /pAn ) is bounded independently of n.
We have
s
ÅM t
ã ÅM ã
A /pAn = A /(p, φn (T )) = Λ/(p, φn (T )) ⊕ Λ/(p, gj (T ), φn (T )) .
i=1 j=1
n
Take n big enough such that p ≥ deg(gj ) for all j and recall that gj and φn (T ) are
distinguished polynomials (in the sense that all but their leading coe cients are divisible by
p). The above formula then equals
s
ÅM t
ã ÅM ã
n n
Λ/(p, T p ) ⊕ Λ/(p, T deg(gj ) ) = (Z/pZ)sp +tg ,
i=1 j=1
P
where g = deg(gj ). This shows that rkp (Cl(Fn )) is bounded independently of n if and
only if s = 0, i.e. if and only if µ = 0. This nishes the proof.
Concerning Iwasawa's invariants, we have the following results:
Theorem 8.14 (Ferrero-Washington). If F is an abelian number eld and F ∞ /F is
the cyclotomic Zp -extension of F , then µ = 0.
Proof. The above theorem is proved by reducing the problem, using the duality coming
from Kummer theory, to calculating the µ-invariant (i.e. the p-adic valuation) of some p-adic
Was97
Dirichlet functions, which can be done explicitly from the constructions that we have given.
See [ , 7.5].
Finally, the following is an open conjecture of Greenberg (see [ Gre76 ]).
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 61
Conjecture 8.15 (Greenberg). For any totally real eld F , and any Z -extension
p
F∞ /F , we have µ = λ = 0. In other words, the values #Cl(Fn ) are bounded as n goes to
+∞.
62 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Appendix
Appendix A. The complex class number formula
K/Q be an abelian extension(17) of degree d = [K : Q]. By the Kronecker-Weber
Let
theorem, K ⊆ Q(µm ) for some minimal positive integer m. Recall the de nition of the
Dedekind zeta function ζK (s) of K ; this has an Euler product
YÅ ã−1
1
ζK (s) = 1− ,
p
N ps
where the product runs over all primes in the ring of integers of K. Since K is abelian, it
Was97 Was97
corresponds to a nite group X of Dirichlet characters of conductor m; these are precisely
the characters of Gal(Q/Q) that factor through Gal(K/Q) [ , 3]. From [ ,
Theorem 4.3], we have a decomposition
Y
ζK (s) = L(χ, s),
χ∈X
Looking at the residues at s=1 of the above equality, one obtains the following alternative
formulation of Theorem 1.2 from the introduction:
Theorem A.1 (Class Number Formula). Let r1 (resp. r2 ), hK , R, w, and D denote,
respectively, the number of real (resp. complex) embeddings, the class number, the regulator,
the number of roots of unity and the discriminant of the eld K . Then we have
2r1 (2π)r2 hK R Y
1/2
= L(χ, 1).
w|D| χ̸=1
In particular, if K is real, we get the formula
hK R Y 1
1/2
= L(χ, 1).
|D| χ̸=1
2
If K is a CM eld, then let h+ , R+ , and d+ = d/2 = r2 denote the class number, the
×
Was97
regulator and the degree respectively of its maximal totally real sub eld K +. Let Q = [OK :
×
µ(K)OK +] be Hasse's unit index (which is equal to either 1 or 2; see [ , Theorem
4.12]). Decomposing the formula of Proposition A.1 and using the formulas of Remark 4.20
one obtains the following:
Proposition A.2 . Let K be any CM abelian number eld. Then we have
Y 1
h = h+ Qw − B1,χ .
2
χ odd
De ne
Y 1
h− ..= h/h+ = Qw − B1,χ .
2
χ odd
Note that the above gives an easily computable product formula for h− that has dispensed
with the transcendental terms of the regular class number formula.
(17) We will only be using the following results when K = Q(µm ) or K = Q(µm )+ .
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 63
Appendix B. Class eld theory
K be a number eld
We recall some necessary basic statements of class eld theory. Let
O ×
= (K ⊗ R)× = v|∞ Kv× the group
Q
and denote by K∞
its ring of integers. Denote by
×
of archimedean units of K and, for every nite place l of K , denote by Ol the units of the
localisation Kl of K at l. If v | ∞, we just let Ov = Kv .
×
De nition B.1
Y′
. The idèles of K are de ned as the restricted product
A× Kv× ×
, xl ∈ Ol× for all but nitely
K := = x∞ , (xl )l : x∞ ∈ K∞ many l ,
v
where the product runs over all places of K and l over its nite places.
We equip A×K with aQtopology, where a basis of open neighbourhoods of the identity is
l nite Ul such that Uv ⊆ Kv is open and Ul = Ol for
×
Q Q
given by U = U
v v = v|∞ Uv ×
× ×
almost all l, which makes AK a locally compact topological group. The global units K of
×
K are diagonally embedded into AK and have discrete image.
De nition B.2 . The quotient CK := K × \A×
K is called the idèle class group of K.
If E/K is a nite extension and P is a prime of E above a prime p of K, then the norm
maps NEP /Kp : EP → Kp de ne a map
NE/K : A× ×
E → AK
sending E× to K× and hence inducing a map between idèle class groups. The main state-
ments of global class eld theory can be stated in the following way.
Proposition B.3 (Global Class Field Theory). Let K be a number eld. Then
nite abelian extensions are in bijective correspondence with open subgroups of CK of nite
index. Precisely, if E/K is any nite abelian extension, then
Gal(E/K) ∼
= CK /NE/K CE ;
and, conversely, for every such nite index open subgroup H of CK there exists a unique
nite abelian extension E of K with NE/K CE = H . Moreover, a place v of K is unrami ed
in E if any only if Ov× ⊆ NE/K CE .
Remark B.4 . Let K ab be the maximal abelian extension of K . Passing to the limit in
the above theorem, one gets an isomorphism between Gal(K ab /K) and the pro nite com-
pletion ofCK . In particular, continuous characters of CK biject with continuous characters
ab
of Gal(K /K).
We will give two examples. Let K be a number eld and let HK be its Hilbert class eld,
i.e. its maximal abelian unrami ed extension. By the above theorem, the extension HK /K
corresponds to the subgroup K ×U
fK of CK , where U
fK = Q O × ,
v v and we therefore have
Gal(HK /K) = A×
K /K UK .
× f
A× lvl (xl ) ,
Q
As usual, there is a natural map K → {ideals of K }, sending (xv )v to l nite
which is surjective and whose kernel is exactly U
fK , and hence induces an isomorphism
CK /UfK ∼
= Cl(K) between the quotient of the idèle class group and the ideal class group of
K . We conclude that
Gal(HK /K) ∼
= Cl(K).
Let now
MK = maximal abelian p-extension of K unrami ed outside every prime p | p;
LK = maximal unrami ed abelian p-extension of K.
64 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Note that LK /K is a subextension of the nite extension HK /K , and by de nition, we
have
Gal(LK /K) = Gal(HK /K) ⊗ Zp
∼
= Cl(K) ⊗ Zp = p-Sylow subgroup of Cl(K).
UK = (O ⊗ Zp )× = p|p Op× be the local units of K at p and EK
Q
Let let be the p-adic
closure of the image VK of O inside UK (diagonally embedded).
×
Proposition B.5 . We have
Gal(MK /LK ) = UK /EK .
Proof. De ne
(p) fK = UK × U (p) ,
Y
UK = Ov× , U K
v∤p
(where Ov× = Kv× if v is an archimedean place). By class eld theory, we have
Gal(MK /K) = A×
K /H,
(p)
where H = K × UK , and the subgroup of Gal(MK /K) corresponding to LK is
fK /H ∼
J ′′ = K × U = UK H/H
∼
= UK /(UK ∩ H).
Observe that we are considering all the modules inside the idèle class group and that the
inclusion of EK inside A×
K is not the inclusion induced by K × ⊆ A×
K: the rst inclusion has
trivial components at places away from p, while the last inclusion is the diagonal one. For
the sake of clarity, we will note
ι : VK → A ×
K
the inclusion induced by VK ⊆ EK ⊆ UK ⊆ A×
K and we are going to see any global unit
inside the idèles by the diagonal embedding.
We now claim that UK ∩ H = EK . One inclusion is clear, since clearly ι(VK ) ⊆ UK and,
(p) (p)
if x ∈ VK , we can write ι(x) = x (ι(x)/x) ∈ K UK , which shows that ι(VK ) ⊆ K UK
× ×
and we conclude by taking the closure on both sides of the inclusion (recall that EK = ι(VK )
by de nition). To prove that UK ∩ H ⊆ EK , de ne, for every n ≥ 1, the subgroup UK,n =
(p)
p|p 1 + p Op . Observe that the sets K UK UK,n (resp. ι(VK )UK,n ), for n ≥ 1, de ne a
n ×
Q
(p)
co nal subset of closed neighbourhoods of K × UK (resp. ι(VK )) and that
(p) (p)
\ \
K × UK = K × UK UK,n , EK = ι(VK )UK,n ,
n≥1 n≥1
(p) (p)
so it su ces to prove K × UK UK,n ⊆ ι(VK )UK,n for every n. Let x ∈ K × , u′ ∈ UK , u ∈
UK,n be such that xu u ∈ UK . So in particular xu′ ∈ UK . Since u′ has component 1 at all
′
primes p | p, then x must be a unit at those primes. Since any element in UK has component
1 at all primes v ∤ p and xu′ ∈ UK , then x must be a unit at all those primes. Hence x is
a global unit. Now observe that, at primes above p, we have xu = x ∈ VK (since it has
′
′
component 1 at any place above p), and at primes outside p, xu = 1, so we conclude that
′ ′
xu ∈ ι(VK ), hence xu u ∈ ι(VK )UK,n , which concludes the proof of the proposition.
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 65
Appendix C. Power series and Iwasawa algebras
In this section we state and give references for some basic yet fundamental results on
Iwasawa algebras and the structure theory of modules over Λ(Zp ).
Fix a nite extension L/Qp and denote OL , p|p, π = πL and k = kL its ring of integers,
its maximal ideal, a uniformizer and its fraction eld respectively.
De nition C.1 . A polynomial P (T ) ∈ OL [T ] is called distinguished if P (T ) = a0 +
a1 T + . . . + an−1 T n−1 + T n with ai ∈ p for every 0 ≤ i ≤ n − 1.
The following two results are some of the main tools in dealing with power series.
Proposition C.2 . Let f (T ) = P i≥0 ai T i ∈ OL [[T ]] and assume that ai ∈ p for 0 ≤ i ≤
n − 1 and that an ∈ OL× . Then:
(Division algorithm) For every g(T ) ∈ OL [[T ]] there exist unique q(T ) ∈ OL [[T ]] and
r(T ) ∈ OL [T ], where r has degree at most n − 1, such that
g(T ) = q(T )f (T ) + r(T ).
(Weierstrass Preparation Theorem) The power series f may be uniquely written in the
form
f (T ) = P (T )U (T ),
where U (T ) ∈ OL [[T ]] is a unit and P (T ) ∈ OL [T ] is a distinguished polynomial of degree n.
Proof. See [ Was97 , Proposition 7.2; Theorem 7.3].
Let
Λ = lim OL [Zp /pn Zp ] ∼
= OL [[T ]]
←−
be the Iwasawa algebra (identifying Λ(Zp ) with OL [[T ]] using the Mahler transform and
dropping Zp from the notation, as is standard). As we have seen, this ring is hugely impor-
tant. Most of the ideas surrounding Iwasawa theory rely on the fact that one has a structure
theorem for nitely generated modules over Λ.
Proposition C.3 .
The prime ideals of Λ are exactly 0, (π, T ), (π) and ideals (P (T ))
where P (T ) is an irreducible distinguished polynomial. Moreover, Λ is a local Noetherian
ring with maximal ideal (π, T ).
Proof. See [ Was97 , Proposition 13.9; Lemma 13.11]
M, M ′ be two Λ-modules. We say that M is pseudo-isomorphic
Let to M ′, and we write
M ∼ M ′ , if there exists a homomorphism M → M ′ with nite kernel and co-kernel, i.e, if
there is an exact sequence
0 → A → M → M ′ → B → 0,
with A and B nite Λ-modules (just in case: A and B have nite cardinality!).
Remark C.4 . not an equivalence relation. In particular, M ∼ M ′ does
Was97
Note that ∼ is
′
not imply M ∼ M . For example, (π, T ) ∼ Λ, but Λ ̸∼ (π, T ) (see [ , 13.2]). This
′
problem goes away if we restrict to the case where M and M are nitely generated torsion
Λ-modules.
The following is the main result concerning the structure theory of nitely generated
Λ-modules, and says that Λ almost behaves as if it was a principal ideal domain.
66 JOAQUÍN RODRIGUES JACINTO & CHRIS WILLIAMS
Theorem C.5 . Let M be a nitely generated Λ-module. Then
s
ÅM t
ã ÅM ã
r ni mj
M ∼Λ ⊕ Λ/(p ) ⊕ Λ/(fj (T ) ) ,
i=1 j=1
for some integers r, s, t ≥ 0, ni , mj ≥ 1 and irreducible distinguished polynomials fj (T ) ∈
O[T ].
Proof. Was97
See [ , Theorem 13.12].
Remark C.6. We do not have a similar result for the nite level group algebras
OL [Zp /p Zp ], only for the projective limit. A fundamental concept in Iwasawa theory is the
n
idea that it is easier to study a tower of p-extensions all in one go, and use results about the
tower to deduce results at nite level, and a major reason for this is this structure theorem.
De nition C.7 . Suppose M is a nitely generated torsion Λ-module. Then r=0 in
the structure theorem. We de ne the characteristic ideal of M to be the ideal
s
Y t
Y Ä m ä
ChΛ (M ) = (pni ) fj j ⊂ Λ.
i=i j=1
The above de nition generalizes slightly to other groups other than those isomorphic to
Zp . Let G = H × Γ ∼= Zp , were H is a nite commutative group of order prime to p and
Γ∼
= p Z . Then we have a decomposition
Λ(G) ∼
= OL [G] ⊗ Λ.
Let M be a nitely generated torsion Λ(G)-module. Let H∧ denote the group of characters
of H and de ne, for any ω ∈ H ∧ ,
1 X −1
eω ..= ω (a)[a] ∈ OL [H].
|H|
a∈H
Lemma C.8 . The group H acts on M (ω) .
. = eω M via multiplication by ω and we have
a decomposition of Λ(G)-modules
M = ⊕ω∈H ∧ M (ω) .
Moreover, each M (ω) is a nitely generated torsion Λ-module.
Proof. See [ CS06 , A.1] or the exercises.
In view of the above lemma, we we have the following de nition.
De nition C.9 . Let G be as above and let M be a nitely generated torsion Λ(G)-
module. We de ne the characteristic ideal of M to be the ideal
ChΛ(G) (M ) ..= ⊕ω∈H ∧ ChΛ (M (ω) ) ⊆ Λ(G).
We will be using the following a basic property of characteristic ideals:
Lemma C.10 . The characteristic ideal is multiplicative in exact sequences.
Proof. See [ CS06 , A.1 Proposition 1]
AN INTRODUCTION TO p-ADIC L-FUNCTIONS 67
References
[Bel09] Joël Bellaïche. An introduction to the conjecture of Bloch and Kato. Unpublished, 2009.
[BSW19] Daniel Barrera Salazar and Chris Williams. Exceptional zeros and L-invariants of Bianchi
modular forms. Trans. Amer. Math. Soc., 372(1):134, 2019.
[Cas86] J. W. S. Cassels. Local elds, volume 3 of London Mathematical Society Student Texts.
Cambridge University Press, Cambridge, 1986.
[Col] Pierre Colmez. La fonction zeta p-adique, notes du cours de M2. Unpublished notes.
[Col00] P. Colmez. Fonctions L p-adiques. Astérisque, (266):Exp. No. 851, 3, 2158, 2000. Sémi-
naire Bourbaki, Vol. 1998/99.
[Col10a] Pierre Colmez. Fonctions d'une variable p-adique. Asterisque, 330:1359, 2010.
[Col10b] Pierre Colmez. Représentations de GL2 (Qp ) et (φ, Γ)-modules. Asterisque, 330:281509,
2010.
[CS06] J. Coates and R. Sujatha. Cyclotomic elds and zeta values. Springer Monographs in
Mathematics. Springer, 2006.
[Dar01] Henri Darmon. Integration on Hp × H and arithmetic applications. Ann. of Math.,
154:589639, 2001.
[DS05] Fred Diamond and Jerry Shurman. A rst course in modular forms, volume 228 of Grad-
uate Texts in Mathematics. Springer-Verlag, New York, 2005.
[Gre76] Ralph Greenberg. On the Iwasawa invariants of totally realy number elds. Amer. J.
Math., 98(1):263284, 1976.
[GZ86] Benedict Gross and Don Zagier. Heegner points and derivatives of L-series. Invent. Math.,
84:225320, 1986.
[Hid93] Haruzo Hida. Elementary theory of L-functions and Eisenstein series, volume 26 of LMS
Student Texts. Cambridge University Press, 1993.
[HK03] Annette Huber and Guido Kings. Bloch-Kato conjecture and main conjecture of Iwasawa
theory for Dirichlet characters. Duke Mat. J., 119, no. 3:393464, 2003.
[Iwa52] Kenkichi Iwasawa. Some properties of L-groups. Proceedings of the International Congress
of Mathematicians, Cambridge Mass., 1950, 1952.
[Kat04] Kazuya Kato. p-adic Hodge theory and values of zeta functions of modular forms.
Astérisque, 295:117290, 2004.
[Kol88] Victor Kolyvagin. Finiteness of E(Q) and X(E/Q) for a subclass of Weil curves. Izv.
Akad. Nauk. USSR ser. Matem., 52, 1988. In Russian.
[Lan90] Serge Lang. Cyclotomic elds I and II, volume 121 of Graduate Texts in Mathematics.
Springer-Verlag, 1990.
[Loe14] David Loe er. P -adic integration on ray class groups and non-ordinary p-adic L-
functions. In T. Bouganis and O Venjakob, editors, Iwasawa 2012: State of the art and
recent advances, volume 7 of Contributions in Mathematical and Computational Sciences,
pages 357 378. Springer, 2014.
[MM91] M. Ram Murty and V. Kumar Murty. Mean values of derivatives of modular L-series.
Ann. of Math., 133:447475, 1991.
[MTT86] Barry Mazur, John Tate, and Jeremy Teitelbaum. On p-adic analogues of the Birch and
Swinnerton-Dyer conjecture. Invent. Math., 84:1 48, 1986.
[MW84] Barry Mazur and Andrew Wiles. Class elds of abelian extensions of Q. Invent. Math.,
76(2):179330, 1984.
[NSW99] Jürgen Neukirch, Alexander Schmidt, and Kay Wingberg. Cohomology of Number Fields.
Springer, 1999.
[PR95] B. Perrin-Riou. Fonctions L p-adiques des représentations p-adiques. Astérisque,
(229):198, 1995.
[Rub00] Karl Rubin. Euler Systems. Annals of Mathematics Studies. Princeton University Press,
2000.
[SU14] Christopher Skinner and Eric Urban. The Iwasawa Main Conjectures for GL(2). Invent.
Math., 195 (1):1277, 2014.
[Tat50] John Tate. Fourier analysis in number elds and Hecke's zeta functions. PhD thesis,
Princeton, 1950.
[Was97] Lawrence C. Washington. Introduction to cyclotomic elds, volume 83 of Graduate Texts
in Mathematics. Springer-Verlag, New York, second edition, 1997.
Joaquín Rodrigues Jacinto, É.N.S. de Lyon • E-mail : joaquin.rodrigues-jacinto@ens-lyon.fr
Chris Williams, University of Warwick • E-mail : christopher.d.williams@warwick.ac.uk
You might also like
- Iwasawa Theory of Elliptic Curves With Complex Multiplication P-Adic L Functions (Ehud de Shalit)100% (2)Iwasawa Theory of Elliptic Curves With Complex Multiplication P-Adic L Functions (Ehud de Shalit)165 pages
- An Introduction To The Theory of L-Functions100% (1)An Introduction To The Theory of L-Functions205 pages
- Theory of L-Functions: An Introduction To TheNo ratings yetTheory of L-Functions: An Introduction To The205 pages
- Riemann'S Zeta Function and Beyond: AbstractNo ratings yetRiemann'S Zeta Function and Beyond: Abstract54 pages
- A P-Adic Measure Attached To The Zeta FunctionsNo ratings yetA P-Adic Measure Attached To The Zeta Functions37 pages
- Iwasawa's Construction of P-Adic L-FunctionNo ratings yetIwasawa's Construction of P-Adic L-Function22 pages
- Problems of The Millennium: The Riemann Hypothesis (2004)No ratings yetProblems of The Millennium: The Riemann Hypothesis (2004)9 pages
- A P-Adic Measure Attached To The Zeta Functions 2No ratings yetA P-Adic Measure Attached To The Zeta Functions 284 pages
- The Riemann Hypothesis - Arithmetic and Geometry - Jeffrey C. LagariasNo ratings yetThe Riemann Hypothesis - Arithmetic and Geometry - Jeffrey C. Lagarias16 pages
- Larry Joel Goldstein - Analytic Number Theory100% (1)Larry Joel Goldstein - Analytic Number Theory300 pages
- Hida Elliptic Curves and Arithmetic Invariants0% (1)Hida Elliptic Curves and Arithmetic Invariants463 pages
- Jornadas Sobre Los Problemas Del Milenio Barcelona, Del 1 Al 3 de Junio de 2011No ratings yetJornadas Sobre Los Problemas Del Milenio Barcelona, Del 1 Al 3 de Junio de 201147 pages
- Algebraic K Theory of Number Fields Alexey Beshenov 2025 Download NowNo ratings yetAlgebraic K Theory of Number Fields Alexey Beshenov 2025 Download Now132 pages
- A - Closer - Look - On - Irrationality - of ZetaNo ratings yetA - Closer - Look - On - Irrationality - of Zeta27 pages
- Giuseppe Romanazzi (University of Insubria) Gromanazzi@uninsubria - ItNo ratings yetGiuseppe Romanazzi (University of Insubria) Gromanazzi@uninsubria - It4 pages
- Algebraic Cycles, Chow Motives, and L-Functions - CommelinNo ratings yetAlgebraic Cycles, Chow Motives, and L-Functions - Commelin39 pages
- The Analytic Class Number Formula and - Functions: Akshay Venkatesh Notes by Tony FengNo ratings yetThe Analytic Class Number Formula and - Functions: Akshay Venkatesh Notes by Tony Feng89 pages
- Andrei Jorza - Math 160c Spring 2013 Caltech - Applications of Global Class Field Theory (2013)No ratings yetAndrei Jorza - Math 160c Spring 2013 Caltech - Applications of Global Class Field Theory (2013)58 pages
- P-Adic Hodge Theory (Brinon and Conrad)No ratings yetP-Adic Hodge Theory (Brinon and Conrad)290 pages
- Algebraic K Theory of Number Fields Alexey Beshenov Updated 2025100% (1)Algebraic K Theory of Number Fields Alexey Beshenov Updated 2025106 pages
- (University Lecture Series) H. Iwaniec - Lectures On The Riemann Zeta Function (2014, American Mathematical Society) PDF100% (1)(University Lecture Series) H. Iwaniec - Lectures On The Riemann Zeta Function (2014, American Mathematical Society) PDF130 pages
- Cyclotomic Fields I and II - Serge LangNo ratings yetCyclotomic Fields I and II - Serge Lang449 pages
- Ambrus Theta, Eisenstein Series Over Function FieldNo ratings yetAmbrus Theta, Eisenstein Series Over Function Field21 pages
- 18.784: Seminar in Number Theory: Lecturer: Professor Ju-Lee KimNo ratings yet18.784: Seminar in Number Theory: Lecturer: Professor Ju-Lee Kim135 pages
- Algebraic K Theory of Number Fields Alexey Beshenov Available Full Chapters100% (2)Algebraic K Theory of Number Fields Alexey Beshenov Available Full Chapters102 pages
- On The Sum of Squared Logarithms InequalNo ratings yetOn The Sum of Squared Logarithms Inequal13 pages
- Automorphic Representations and L-Functions For The General Linear Group - Volume 2cNo ratings yetAutomorphic Representations and L-Functions For The General Linear Group - Volume 2c210 pages
- Cohomology of Arithmetic Groups, L-Functions and Automorphic - T. Venkatamarana PDFNo ratings yetCohomology of Arithmetic Groups, L-Functions and Automorphic - T. Venkatamarana PDF132 pages
- Analytic Number Theory 1st Edition Shigeki Akiyama Hideaki Ishikawa Auth DownloadNo ratings yetAnalytic Number Theory 1st Edition Shigeki Akiyama Hideaki Ishikawa Auth Download89 pages
- Recent Perspectives in Random Matrix Theory and Number Theory Mezzadri F Instant DownloadNo ratings yetRecent Perspectives in Random Matrix Theory and Number Theory Mezzadri F Instant Download76 pages
- Automorphic Forms Beyond Gl2 Lectures From The 2022 Arizona Winter School Hang Xue PDF Download100% (1)Automorphic Forms Beyond Gl2 Lectures From The 2022 Arizona Winter School Hang Xue PDF Download59 pages
- Universality of L-Functions J. SteudingNo ratings yetUniversality of L-Functions J. Steuding12 pages
- Elliptic Curves and Big Galois Representations 1st Edition Daniel Delbourgo - PDF Download (2025)No ratings yetElliptic Curves and Big Galois Representations 1st Edition Daniel Delbourgo - PDF Download (2025)54 pages
- Elliptic Tales Curves Counting and Number Theory Avner Ash - Download The Entire Ebook Instantly and Explore Every Detail100% (1)Elliptic Tales Curves Counting and Number Theory Avner Ash - Download The Entire Ebook Instantly and Explore Every Detail61 pages
- Links Between String Theory and The Riemann's Zeta FunctionNo ratings yetLinks Between String Theory and The Riemann's Zeta Function23 pages
- Learning Algebraic Number Theory: Sam Ruth March 31, 2010100% (1)Learning Algebraic Number Theory: Sam Ruth March 31, 201015 pages
- (Dorian Goldfeld) Automorphic Forms and L-FunctionNo ratings yet(Dorian Goldfeld) Automorphic Forms and L-Function509 pages
- Elliptic Tales Curves Counting and Number Theory Avner Ash Instant Download100% (1)Elliptic Tales Curves Counting and Number Theory Avner Ash Instant Download47 pages
