Hecke's L-Functions
Hecke's L-Functions
Hecke's L-Functions
Kenkichi Iwasawa
Hecke’s
L-functions
Spring, 1964
SpringerBriefs in Mathematics
Series Editors
Nicola Bellomo, Torino, Italy
Michele Benzi, Pisa, Italy
Palle Jorgensen, Iowa City, USA
Tatsien Li, Shanghai, China
Roderick Melnik, Waterloo, Canada
Otmar Scherzer, Linz, Austria
Benjamin Steinberg, New York City, USA
Lothar Reichel, Kent, USA
Yuri Tschinkel, New York City, USA
George Yin, Detroit, USA
Ping Zhang, Kalamazoo, USA
SpringerBriefs present concise summaries of cutting-edge research and practical
applications across a wide spectrum of fields. Featuring compact volumes of 50 to
125 pages, the series covers a range of content from professional to academic.
Briefs are characterized by fast, global electronic dissemination, standard publish-
ing contracts, standardized manuscript preparation and formatting guidelines, and
expedited production schedules.
Hecke’s L-functions
Spring, 1964
123
Kenkichi Iwasawa
Princeton University
Princeton, NJ, USA
Foreword by
John Coates Masato Kurihara
Emmanuel College Department of Mathematics
Cambridge, UK Keio University
Yokohama, Japan
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
K. Iwasawa
Princeton, 1986
Foreword
ix
x Foreword
argument involving the compactness of the idèle class group of F also gives a
non-classical proof of the finiteness of the class number and the unit theorem.
In the final chapter, Iwasawa succinctly explains the link between the adelic
approach and the classical theory. He then goes on to give detailed proofs of key
classical results on the distribution of prime ideals, and on the class number for-
mulae for cyclotomic fields.
This volume provides an elegant and detailed account of questions which are of
seminal importance for modern number theory, and it covers material which is not
treated as fully or as elegantly in other basic texts on algebraic number theory. We
believe that it will provide an ideal text for future courses on this central part of
number theory. Finally, we warmly thank Takahiro Kitajima for his accurate
conversion of Iwasawa’s handwritten notes into LATEX, and Rei Otsuki for his help
with proofreading.
xi
Chapter 1
Algebraic Number Fields
Notation:
Z = the ring of integers, 0, ±1, ±2, . . .,
Q = the rational field,
R = the real field,
C = the complex field.
1.1 Ideals
For any a = 0 in F, let (a) = ao. Then (a) is an ideal in I; it is called the principal
ideal generated by a. (1) = o is of course the unit of I, and the set of all principal
ideals forms a subgroup H of I. I/H is called the ideal class group of F. It is a finite
group, and its order, [I : H], is called the class number of F.
Let p be a prime ideal of o. Then o/p is a finite field. The number of elements in
o/p, [o : p], is called the norm of p, and is denoted by N (p). Let a be any ideal in I,
and let a be written in the form (1.1). We define the norm of a by
N (a) = N (p)m p .
p
Then N (ab) = N (a)N (b) for any a, b in I, and if a is integral, then N (a) = [o : a].
If a = ao, then N (a) = |N (a)| with N (a) the norm of a.
An element a = 0 in F is called a unit of F if (a) = o; a = 0 is a unit if and
only if both a and a −1 are in o. The set of all units of F forms a subgroup E of
the multiplicative group F ∗ of F. Let W be the set of all roots of unity contained
in F. Then W is a finite subgroup of E, and E/W is a free abelian group of rank
r1 + r2 − 1 (see below).
Let n = [F : Q]. Then there exist exactly n isomorphisms σ1 , . . . , σn from F into
C. Let σ : F → C be one of them. Then σ : F → C, defined by σ (a) = σ (a) = the
complex conjugate of σ (a), is also an isomorphism. We call σ real or complex
according as σ = σ or σ = σ . We shall always fix the indices so that σ1 , . . . , σr1 are
real, σr1 +1 , . . . , σn are complex, and σr1 +i = σr1 +r2 +i , 1 ≤ i ≤ r2 , n = r1 + 2r2 .
For each a in F, let a (i) = σi (a) for i = 1, 2, . . . , n. Let α = (a1 , . . . , an ) be
( j)
any ordered basis of F over Q. Define an n × n matrix Mα by Mα = (ai ).
Then Mα Mα = (T (ai a j )), where T denotes the trace map from F to Q. Put
t
D(α) = det Mα tMα = (det Mα )2 = det(T (ai a j )). Then D(α) is a rational number,
and D(α) = 0. It follows that there exists a unique basis β = (b1 , . . . , bn ) such that
T (ai b j ) = δi j , i, j = 1, . . . , n,
and N (d) = || = d. By the definition, d−1 = o is the largest ideal of F such that
a = a−1 d−1 for any ideal a.
o) ⊂ Z. d is called the different of F. We have
T (
vi (a − ai ) < ε, i = 1, . . . , m.
S0 = {Pp }.
Thus there is a one-one correspondence P ↔ p between the set S0 and the set of all
prime ideals of o. The prime ideal corresponding to a prime spot P will be denoted
by p P . We also write v P , ν P , and Fp for vp , νp and FP respectively if P ↔ p.
Now, v P is defined for each P in S. For any a = 0 in F, v P (a) = 1 for almost
all P in S, and
v P (a) = 1.
P∈S
ν Pi (a − ai ) ≥ m i , i = 1, . . . , m,
ν P (a) ≥ 0, P ∈ S0 , P = P1 , . . . , Pm .
F ⊗Q Q Q ∼
= FP1 ⊕ · · · ⊕ FPg .
Let P be any one of P1 , . . . , Pg . Let n P = [FP : Q Q ], and let TP and N P denote the
trace and the norm map from FP to Q Q . Then it follows from the above that
g
g
g
n= n Pi , T (a) = TPi (a), N (a) = N Pi (a)
i=1 i=1 i=1
for any a in F.
Let p be a prime ideal of o. The extensions of the functions vp and νp or Fp (=
FP , P = Pp ) will be denoted again by the same letters. Let
op = {a | a ∈ Fp , νp (a) ≥ 0}
mp = {a | a ∈ Fp , νp (a) > 0}.
In general, let a be an ideal of F, a = {0}, and let a = psp , sp = νp (a). Let
s
ap be the closure of a in Fp (a ⊂ F ⊂ Fp ). Then ap = mpp = {a | a ∈ Fp , νp (a) ≥
sp = νp (a)}, and
a= (F ∩ ap ).
p
Since tp = νp (d) = 0 for almost all p, we have dp = op for almost all p namely,
ep (= e P ) = 1 for almost all p (almost all P in S).
The multiplicative group Fp∗ of the locally compact field Fp is a locally compact
abelian group in the natural topology. The groups 1 + msp , s ≥ 1, are open, compact
subgroups of Fp∗ , and form a base of neighborhoods of 1 in Fp∗ . Let
σi : FP → R or C.
Hence FP is again a locally compact field, and FP∗ , the multiplicative group of FP ,
is a locally compact abelian group. Let
U P = {a | a ∈ FP , v P (a) = 1}.
Then
{±1}, if σi is real,
UP ∼
=
C1 = {z | z ∈ C, |z| = 1} if σi is complex.
Remark 1.1 Let μ be a Haar measure of the additive group of FP . Then for any
measurable subset A of FP , we have
μ(a A) = v P (a)μ(A), a ∈ FP .
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2019 9
K. Iwasawa, Hecke’s L-functions, SpringerBriefs in Mathematics,
https://doi.org/10.1007/978-981-13-9495-9_2
10 2 Idèles
Proof Let O = o P × O1 × · · · × Or , r = r1 + r2 , where Oi = {x | x ∈ FP∞,i ,
v P∞,i (x) < 1}. Then O is a neighborhood of 0 in R , and hence in R. Let a be
n
in F ∩ O. Then a ∈ (F ∩ o P ) = o (cf Sect. 1.2). Hence N (a) = σi (a) is an
P i=1
algebraic integer in Q, namely, a rational integer. On the other hand, |N (a)| =
n r
|σi (a)| = v P∞,i (a) < 1. Hence a = 0. It follows that F ∩ O = {0} and that
i=1 i=1
F is discrete in R.
An invertible element of the ring R is called an idèle of F. The set of all idèles of
F forms a multiplicative group J , called the idèle group of F. It is easy to see that
α = (a P ) is an idèle of F if and only if a P = 0 for every P in S and v P (a P ) = 1 for
almost all P in S. Clearly the multiplicative group F ∗ of the field F is a subgroup
of J .
Let
J = UP × FP∗ .
P∈S0 P∈S∞
Problem 2.2 (i) For each neighborhood O of 1 in R, let O denote the set of all α
in J such that both α and α −1 are in O. Prove that {O } is a base of neighborhoods
of 1 in J in the idèle topology of J .
(ii) Prove that the two topologies of J are different.
(iii) Let P be fixed. Prove that the map R → FP (resp. J → FP∗ ) defined by
α = (. . . , a P , . . .) → a P is continuous (resp. in the idèle topology of J ).
2.1 Adèles and Idèles 11
Let
U= UP .
P∈S
J0 = {α | α = (a P ) ∈ J, a P = 1 for all P ∈ S∞ },
J∞ = {α | α = (a P ) ∈ J, a P = 1 for all P ∈ S0 }.
J = J0 × J∞ .
Put
U0 = U ∩ J0 , U∞ = U ∩ J∞ .
Then
U = U0 × U∞ , J = U0 × J∞ , J/J = J0 /U0 .
Since v P (a P ) = 1 for almost all P’s, V (α) is well defined. It is easy to see that
V : J → R+
J1 = ker(V )
= {α | α ∈ J, V (α) = 1}.
T ∼
= R+ .
J = J1 × T.
The product formula for F states that V (a) = v P (a) = 1 for any a in F ∗ .
Hence
F ∗ ⊂ J1 ,
J = J/F ∗ , J1 = J1 /F ∗ , T = F ∗ T /F ∗ .
Then
J = J1 × T , T ∼
=T ∼
= R+ ,
topologically.
For α = (a P ) in J , let
ι (α) = pνPP (a P ) .
P∈S0
Since ν P (a P ) = 0 for almost all P in S0 , the product is well defined, and ι (α) is an
ideal of F. The map
ι : J → I = the ideal group of F
J/J = J0 /U0 ∼
= I.
Let α = (a P ) be in J0 . Then
−1
V (α) = v P (a P ) = N (p P )−ν P (a P ) = N pνPP (a P ) = N (ι (α))−1 ,
P∈S0 S0
namely,
V (α) = N (ι (α))−1 , α ∈ J0 .
ι (a) = (a), a ∈ F ∗ .
Hence
ι (F ∗ ) = H = the group of principal ideals of F
and we have
J/F ∗ J ∼
= I/H = the ideal class group of F.
2.1 Adèles and Idèles 13
E = F ∗ ∩ J .
W = F ∗ ∩ U.
Remark 2.2 In general, we can define adèles for any algebraic variety V defined
over a number field F. We shall next explain such adèles in the case where V is
an affine algebraic group. A subgroup G of SL(m; F), m ≥ 1, is called algebraic if
there exists a set of polynomials f 1 , . . . , f s in F[X 11 , . . . , X i j , . . . , X mm ] such that
G consists of all m × m matrices A = (ai j ) over F, satisfying f k (· · · , ai j , . . .) = 0,
for k = 1, . . . , s. It is known in the theory of algebraic groups that an affine algebraic
group defined over F is always isomorphic to such a subgroup of SL(m; F) for some
m ≥ 1.
Let M(m; R) denote the ring of all n × n matrices over the adèle ring R of F;
it is a locally compact ring in the obvious manner. A matrix A = (αi j ) in M(m; R)
is called an adèle of G if f k (. . . , αi j , . . .) = 0 for k = 1, . . . , s. The set of all such
adèles of G forms a multiplicative locally compact group G relative to the topology
induced by that of M(m; R). We call G the adèle group of G. The group G itself is
then a discrete subgroup of G.
x p ≡ x mod Z.
p
T (a) = TP (a), a ∈ F,
P∈S∞
we obtain
ω P (a) = exp(−2πi T (a)), a ∈ F.
P∈S∞
Now, let P be finite, and let p(= Pp ) be the restriction of P on Q. Denoting again
by TP : FP → Q P the trace map, we put
T (a) = TP (a), a ∈ F.
P ext. of p
2.2 Characters of R and J 15
For α = (a P ) in R, let
ω(α) = ω P (a P ).
P∈S
ω(a) = 1, a ∈ F.
namely,
ω0 (a) = exp(2πi T (a)), a ∈ F.
16 2 Idèles
g
U0 (f) = (1 + mePii ) × U P × 1 × · · · × 1.
i=1 P= Pi
P∈S0
r
xk m k
χ∞ (α) = · |xk |isk , σ (α) = (x1 , . . . , xr ),
k=1
|xk |
so that −1
xk+r2 xk
|xk+r2 | = |xk |, = , r1 < k ≤ r.
|xk+r2 | |xk |
Let
m k = m k , sk = sk , 1 ≤ k ≤ r1 ,
mk = m k , sk = 21 sk , r1 < k ≤ r,
m k+r2 = 0, sk+r2 = 21 sk , m k ≥ 0,
m k = 0, sk = 21 sk , r1 < k ≤ r,
m k+r2 = −m k , sk+r2 = 21 sk ,
m k < 0.
Then
n
xk m k
χ∞ (α) = |xk |isk ,
k=1
|x k |
J0 (f) = {α | α = (a P ) ∈ J0 , a Pi ∈ 1 + mePii , 1 ≤ i ≤ g}
= {α | α = (a P ) ∈ J0 , a Pi ≡ 1 mod mePii , 1 ≤ i ≤ g}.
J0 (f)/U0 (f) ∼
= I(f).
Now, let χ be a character of J with conductor f. Since χ (U0 (f)) = 1, there exists
a homomorphism χ : I(f) → C1 such that
χ
J0 (f) C1
ι0 χ
I(f)
is commutative. We always consider the ideal groups I, I(f), etc., in discrete topology.
So χ : I(f) → C1 is a character of I(f). We call it the ideal character (of I(f)) induced
by the idèle character χ . If there is no risk of confusion, we shall write χ (a) for χ (a),
a ∈ I(f). By the definition,
Let R be a finite ring with identity 1; for simplicity, we shall assume that R is
commutative although this is not essential in the following theory. Let R∗ denote
the multiplicative group of all invertible elements in R. Let λ be a character of the
additive group of R, and ρ a character of the multiplicative group R∗ . Put
G(λ, ρ) = λ(x)ρ(x).
x∈R∗
Such a sum, defined for each pair of characters λ and ρ, is called a Gaussian sum on
R.
p−1
x 2πi x
G(λ, ρ) = e p .
x=1
p
2.3 Gaussian Sums 19
In particular,
G(λ, ρ) = 0.
Proof |G(λ, ρ)|2 = x∈R∗ λ(x)ρ(x) y∈R∗ λ(y)ρ(y) = x,y λ(x − y)ρ( xy ). Let
z = xy . Then the sum is equal to
⎛ ⎞
⎜
⎟
λ((z − 1)y)ρ(z) = ⎜ − ⎟.
⎝ ⎠
y,z∈R∗ z∈R∗ y∈R y∈R
/ ∗
y ∈R
However,
By the assumption (ii) and by Lemma 2.3 (1), we have G(λ y , ρ) = 0 for every y in
/ R∗ . Hence the above sum is 0, and the lemma is proved.
R, y ∈
p−1 2πi x
Example 2.3 For G(λ, ρ) = x=1 ( p )e
x p , the assumptions (i), (ii) are satisfied for
λ, ρ. Hence
|G(λ, ρ)|2 = #(R) = p.
20 2 Idèles
Let a be an ideal of F, a = {0}, a = pνp (a) . Put
ν (a)
Ra = mpp × FP
p P∈S∞
Ra = R /Ra , a ⊂ o.
Ra ∼
= o/a,
ι (α) = a, α ∈ (J0 × T ) ∩ J1 .
Now, let d be as before the different of F, and let d = N (d) = |Δ|, where Δ
is the discriminant of F. Let δ be an idèle associated with d, namely, δ = (d P ),
ν P (d P ) = ν P (d), P ∈ S0 , d P∞,i = σi−1 (d n ), P∞,i ∈ S∞ . Let d P be the different of
1
−1 1
d P = d P o P , oP = dP = oP , P ∈ S0 .
dP
R = Rf = R /Rf ∼
= o/f.
Then R is a finite ring with f elements. We shall fix χ , ϕ, and δ, and define a Gaussian
sum on the ring R.
I. The Additive Character λ : R → C1
ξ
Since δ, ϕ are in J , δϕ is in R. Hence ω0 is well defined, and is obviously a character
of R .
Let α = (a P ) be an element of R . We shall prove that ω0 (αξ ) = 1 for every
ξ = (x P ) in R ⇔ α is in Rf . Suppose that ω0 (αξ ) = 1 for any ξ = (x P ) in R . Then
αξ aP x P
ω0 (αξ ) = ω0 = ωP = 1.
δϕ P∈S0
dP f P
aP x P aP x P
ωP = exp 2πi T p = 1,
dP f P dP f P p
The above result then shows that λα (ξ ) = λ(αξ ) = ω0 (αξ ) = 1 for any ξ in R (for
any ξ in R ) ⇔ α = 0. Thus
ρ(R∗α ) = 1, α ∈
/ R∗ .
Thus the condition (ii) of Lemma 2.4 is satisfied for the character ρ.
We now put
C(χ ; δ, ϕ) = G(λ, ρ)
= λ(ξ )ρ(ξ ),
ξ ∈R∗
Proposition 2.4
|C(χ ; δ, ϕ)| = f , f = N (f),
C(χ ; δ, ϕ) = χf (−1)C(χ ; δ, ϕ).
Proof The first equality follows immediately from Lemma 2.4. As to the second,
24 2 Idèles
Here ρ(−1) = χf (−1), ρ(ξ ) = χ f (ξ ). Hence λ(ξ )ρ(ξ ) = G(λ, ρ) = C(χ ; δ, ϕ),
and the equality is proved.
Let a and b be ideals of F, a, b = {0}, b ⊂ a, and let ξ ∈ J . Then
1. Ra = a + Rb , b = a ∩ Rb so that a/b ∼ = Ra /Rb under x + b ↔ x + Rb , x ∈ a,
2. ξ Ra = Rι (ξ )a so that Ra ∼
= Rι (ξ )a under α ↔ ξ α, α ∈ Ra .
Problem 2.5 Prove the above 1, 2.
We now prove the following.
Proposition 2.5 Let α ∈ J , a = ι (α), a ∈ ad−1 f−1 (b ∈ F). Then
aξ α ξ
ω χ f (ξ ) = ω0 χ f (ξ ),
α δϕ
ξ ∈R ξ ∈R
We shall next state without proof some fundamental results on Haar measures of
locally compact groups. For simplicity, we shall always assume that our locally
compact groups are abelian and separable (i.e. have a countable basis for open sets).
−1
f (x)dμ(x) = f (x )dμ(x) = f (a −1 x)dμ(x),
G G G
namely,
dμ(x) = dμ(x −1 ) = dμ(ax), a ∈ G.
Let H be a closed subgroup of G. Then both H and G/H are again locally
compact, separable, abelian groups. Let μG , μ H , and μG/H be Haar measures of the
respective
groups. Let f be a continuous function with compact support on G. Then
g(x) = H f (xu)dμ H (u) is defined for any x in G, and g(x) depends only upon the
coset x = x H . Hence we put
g (x ) = f (xu)dμ H (x), x = x H ∈ G/H.
H
26 2 Idèles
The function g is then continuous, and has a compact support in G/H . Hence
G/H g (x )dμG/H (x ) is defined.
Now, we say that the measures μG , μ H , and μG/H are matched if
f (x)dμG (x) = g (x )dμG/H (x )
G G/H
= f (xu)dμ H (u) dμG/H (x ), (2.1)
G/H H
μG = μG/H ∗ μ H , μG/H = μG /μ H .
is defined for almost all x in G/H ; g is a function in L 1 (G/H, μG/H ), and (2.1)
again holds.
Let H be a closed subgroup of G; and U a closed subgroup of H . Let μG/H and
μ H/U be Haar measures of the respective groups. Let G = G/U , H = H/U . Then
G/H is naturally identified with G /H so that μG/H = μG /H . As μ H/U = μ H ,
μG/H ∗ μ H/U = μG /H ∗ μ H is defined, and it is a Haar measure on G/U .
8. Let V ⊂ U ⊂ H ⊂ G be a sequence of closed subgroups, and let μG/H , μ H/U ,
μU/V be given Haar measures on the respective groups. Then Associative law: μG/H ∗
(μ H/U ∗ μU/V ) = (μG/H ∗ μ H/U ) ∗ μU/V .
9. Let U ⊂ H ⊂ G, and let μG , μ H , and μU be given. Then
dμ(x) = d x1 · · · d xn .
2. G = R∗ or R+ , x ∈ R∗ or R+
dx
dμ(x) = .
|x|
3. G = C∗ , z ∈ C∗ , z = x + i y = r eiθ .
2d xd y 2dr dθ
dμ(z) = = .
|z| 2 r
We shall next fix Haar measures of various groups defined by J and its subgroups.
Let μU0 be the Haar measure of the compact group U0 such that μU0 (U0 ) = 1,
and let μ J0 /U0 be the point measure of the discrete group J0 /U0 . Let
μ J0 = μ J0 /U0 ∗ μU0 .
μ J∞ ≈ μ ,
under σ .
Let
μ J = μ J0 × μ J∞ .
μT ≈ μR+
28 2 Idèles
μ J = μ J /μ F ∗ , μ J1 = μ J1 /μ F ∗ .
Let
μT ≈ μT
μ J1 × μT ≈ μ J1 × μT
μ J = μ J /μ F ∗ ≈ μ J1 × μT .
Therefore
μ J = μ J1 × μT .
Remark 2.3 Let μ be any Haar measure of the additive group R. Let ξ be an element
−1
of J . Then σξ : R → R, α → ξ α, is a topological isomorphism. Let μ = μσξ
so that μ (A) = μ(σξ (A)) = μ(ξ A) for any Borel set A in R. Then μ is also a
Haar measure of R, and μ = cξ μ with some constant cξ > 0. It can be proved that
cξ = V (ξ ), namely, that
μ(ξ A) = V (ξ )μ(A), ξ ∈ J, A ⊂ R.
Then σa (A ) = A . As R/F is compact, 0 < μ(A ) < +∞. Hence it follows from
the above that ca = 1, namely, that V (a) = 1, a ∈ F ∗ . Thus the product formula for
F is essentially a consequence of the compactness of R/F.
Chapter 3
L-functions
3.1 Definition
N
lim (1 − u m (s)), s ∈ D,
N →∞
m=1
∞
(1 − u m (s)), s ∈ D.
m=1
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2019 31
K. Iwasawa, Hecke’s L-functions, SpringerBriefs in Mathematics,
https://doi.org/10.1007/978-981-13-9495-9_3
32 3 L-functions
we have
|χ (p)N (p)−s | < 1, χ (p)N (p)−s = 1, for Re (s) > 0.
Let
1 − u p (s) = (1 − χ (p)N (p)−s )−1 ,
namely,
−χ (p)N (p)−s
u p (s) = , Re (s) > 0.
1 − χ (p)N (p)−s
p −a
|u p (s)| ≤ = 2 p −a , s ∈ Da ,
1 − 21
For each prime number p, there exist at most n = [F : Q] prime ideals p such that
p| p (cf. I. §2). Hence, when p ranges over all prime ideals of o, we obtain
⎛ ⎞ ⎛ ⎞
|u p (s)| = ⎝ |u p (s)|⎠ ≤ ⎝ 2 p −a ⎠
p p p| p p p| p
≤ 2np −a = 2n p −a < +∞, s ∈ Da .
p p
This shows that p |u p (s)| converges uniformly to a bounded function on Da (and
also that {u p (s)} is a countable set). Applying the result stated at the beginning to
D = Da , we obtain a holomorphic function
(1 − u p (s)) = (1 − χ (p)N (p)−s )−1
p p
on Da . Since this holds for any a > 1, we see that the same product defines a function
of s for any s with Re (s) > 1. So, let
L(s; χ ) = (1 − χ (p)N (p)−s )−1 , for Re (s) > 1.
p
3.1 Definition 33
L(s; χ ) is called Hecke’s L-function for F with the character χ , as it follows from
the above, L(s, χ ) is holomorphic and L(s; χ ) = 0 for Re (s) > 1, and
m
(1 − χ (pk )N (pk )−s )−1
k=1
converges uniformly to L(s; χ ) in Da = {s | Re (s) > a}, a > 1, for any ordering
p1 , p2 , p3 , . . . , of the prime ideals of o.
Let χ ≡ 1. Then f = o, and χ (a) = 1 for any ideal a of F. In this case, we write
ζ F (s) for L(s; χ );
ζ F (s) = (1 − N (p−s )−1 , Re (s) > 1.
p
Proof For any real x > 0, there exist only a finite number of prime ideals p such that
N (p) ≤ x. We have
(1 − N (p)−a )−1 = (1 + N (p)−a + N (p)−2a + · · · )
N (p)≤x N (p)≤x
= N (a)−a ,
a
where the sum is taken over all integral ideals a such that the prime factors p satisfy
N (p) ≤ x. Clearly every integral ideal a with N (a) ≤ x is included in the above sum.
Hence
N (a)−a ≤
N (a)−a = (1 − N (p)−a )−1
N (a)≤x a N (p)≤x
a⊂o
≤ (1 − N (p)−a )−1 = ζ F (a).
p
34 3 L-functions
in Da . Since
Now, let s be |χ (a)N (a)−s | ≤ N (a)− Re (s) ≤ N (a)−a , we see from
−s
the above that χ (a)N (a) converges absolutely and uniformly in Da . A compu-
tation similar to the above also shows that
(1 − χ (p)N (p)−s )−1 =
χ (a)N (a)−s , s ∈ Da ,
N (p≤x) a
−a
−a
Let x → +∞. Since a N (a) converges, N (a)>x N (a) → 0. Hence we
obtain
L(s; χ ) = (1 − χ (p)N (p)−s )−1 = χ (a)N (a)−s , s ∈ Da .
p a
Since a(>1) is arbitrary, the above equality also holds for any s with Re (s) > 1.
For χ ≡ 1, we have
ζ F (s) = (1 − N (p)−s )−1 = N (a)−s , Re (s) > 1.
p a
∞
ζ (s) = (1 − p −s )−1 = m −s , Re (s) > 1.
p m=1
Hence
s ∞
dx
π−2 Γ m −s = e−πm x x 2
s 2 s
.
2 0 x
It follows that
s ∞
∞
− 2s dx
e−πm x x 2
2 s
π Γ ζ (s) =
2 m=1 0
x
∞∞
dx
e−πm x x 2
2 s
=
0 m=1
x
∞
s dx
= ω(x)x 2 ,
0 x
where
∞
e−πm x , x > 0.
2
ω(x) =
m=1
The interchange of the sum and the integral is allowed because e−πm x x 2 ≥ 0. Let
2 s
e−πm x = 1 + 2ω(x), x > 0.
2
θ (x) =
m∈Z
namely, that
1 1 1 1
ω(x) = √ ω − + √ , x > 0.
x x 2 2 x
Hence we have
∞ 1 ∞
dx s s dx s dx
ω(x)x 2 = ω(x)x 2 + ω(x)x 2 ,
0 x 0 x 1 x
1 1
s dx 1 1 1 1 s dx
ω(x)x 2 = √ ω − + √ x2 ,
0 x 0 x x 2 2 x x
s s−1
1
1
1 s 1 s−1 d x 1x2 1x 2
− x2 + x 2 = − s + s−1 , s>1
0 2 2 x 2 2 2 2
0
1 1
=− +
s s−1
1
= ,
s(s − 1)
36 3 L-functions
so that
1
1
s dx 1 s−1 d x 1
ω(x)x 2 = ω
x 2 +
0 x 0 x x s(s − 1)
∞
1−s d x 1
= ω(x)x 2 + ,
x s(s − 1)
s ∞
1
∞
s dx 1−s d x 1
π−2 Γ
s
ζ (s) = ω(x)x 2 + ω(x)x 2 + , s > 1.
2 1 x 1 x s(s − 1)
∞ ∞
s dx a dx
|ω(x)x |
2 ≤ ω(x)x 2 < +∞.
1 x 1 x
defines a holomorphic function of s for Re (s) < a. Since a (> 1) is arbitrary, η(s)
actually defines a holomorphic function on the entire s-plane.
Let
s
ξ(s) = π − 2 Γ
s
ζ (s), Re (s) > 1.
2
It follows from the above that
1
ξ(s) = η(s) + η(1 − s) + , for s > 1. (3.1)
s(s − 1)
Since both sides of the above are holomorphic function of s for Re (s) > 1, the equal-
ity holds for any s with Re (s) > 1. Moreover, the right-hand side is a meromorphic
function on the entire s-plane with the only simple poles at s = 0, 1. Hence, by the
above equality ξ(s) can be continuous analytically on the entire s-plane, and so is
s −1 −1
ζ (s) = π 2 Γ 2s ξ(s). Since Γ 2s has a zero at s = 0, ζ (s) is regular at s = 0.
1 1
Since Γ 2 = π 2 and since ξ(s) has a simple pole with residue 1 at s = 1, ζ (s)
also has a simple pole with residue 1 at s = 1. Furthermore, it follows from (3.1)
(which now holds for arbitrary s) that
3.1 Definition 37
n
A(x, y) = (x A, y) = (x, y ∗A) = a jk x j yk ,
j,k=1
A(x) := A(x, x), x, y ∈ Cn .
Since A is symmetric, A(x, y) = A(y, x); and since A is real and positive definite,
there exists an a > 0 such that
For x, y ∈ Cn , let
θ (x, y; A) = exp(−π A(x + m) + 2πi(y, m)),
m
Proof Let
q(m; x, y) = −π A(x + m) + 2πi(y, m).
where u = −2π x A + 2πi y. For the given bounded domain D in Cn , there exists a
constant C > 0 such that
u ≤ C, |π A(x)| ≤ C, x, y ∈ D.
Let m = (m 1 , . . . , m n ) ∈ Zn ,
m
< aπ
2c
. Then |m k | < aπ
2c
, m k ∈ Z. Hence there exist
n
only a finite number of such m’s in Z . Therefore it is sufficient to show that
exp(q(m; x, y))
m
≥ 2C
aπ
l∈Z
2
∞
πa n
= C 1+2 exp − l 2
l=1
2
∞
πa
≤ C 1 + 2 exp − l
l=1
2
πa n
e− 2
= C 1 + 2 πa .
1 − e− 2
Proof We first note that if A is real, positive definite and symmetric, then so is A−1 ,
and |A| > 0. Hence θ (x, y; A−1 ) is defined, and is a holomorphic function of x and
y. Since both sides of the above are holomorphic in x ∈ Cn , it is sufficient to prove
the equality for real x ∈ Rn . Let
f (x) = θ (x, 0; A) = exp(−π A(x + m)), x ∈ Rn .
m
where the right-hand side converges absolutely and uniformly on each bounded
domain of Rn , and where a(m) is given by
1 1
a(m) = ··· f (x) exp(−2πi(x, m))d x1 · · · d xn .
0 0
m 1 +1 m n +1
= ··· exp (−π A(x) − 2πi(x, m)) d x
m m 1 m n
∞ ∞
= ··· exp (−π A(x) − 2πi(x, m)) d x1 · · · d xn .
−∞ −∞
Since A is positive definite, there exists an n × n, real, positive definite and symmetric
matrix B such that A = B 2 . Let
y = x B, t = m B −1 .
Therefore
1
a(m) = √ exp −π A−1 (m) .
|A|
It follows that
1
θ (x, 0; A) = f (x) = √ exp −π A−1 (m) + 2πi(x, m)
m
|A|
1
= √ θ (0, x; A−1 ).
|A|
1
θ (x, y; A) = √ exp (−2πi(x, y)) θ (−y, x; A−1 ).
|A|
Proof In Lemma 3.2, we make the substitution x → x − i y A−1 . Then the exponent
of the m-term on the left is
Here
Hence
−π A−1 (m) + 2πi(x − i y A−1 , m) = −π A−1 (m) + 2πi(x, m) + 2π A−1 (y, m).
3.2 Theta-Formulae (Analytic Form) 41
Since
−π A−1 (m) + 2π A−1 (y, m) − π A−1 (y) = −π A−1 (−y + m),
Remark 3.1 Let y = 0 in the above. Then we have the equality in Lemma 3.2. Let
also x = 0. Then
1
θ (0, 0; A) = √ θ (0, 0; A−1 ),
|A|
namely,
1
exp (−π A(m)) = √ exp(−π A−1 (m)).
m
|A| m
u(x) = x Mα
= (u 1 (x), . . . , u n (x)),
n
u k (x) = a (k)
j x j , k = 1, . . . , n.
j=1
u k (x) = u k (x), 1 ≤ k ≤ r1 ,
u k (x) = u k+r2 (x),
u k+r2 (x) = u k (x), r1 + 1 ≤ k ≤ r1 + r2 .
n
A(x) = ck |u k (x)|2 , x ∈ Rn . (3.2)
k=1
n
a= ak m k .
k=1
θ (x, 0; A)
r
1 1 +r2
r
(k) 2 (k) (k+r2 )
= exp −π ck (u k + a ) + 2 ck (u k + a )(u k+r2 + a ) .
a∈a k=1 k=r1 +1
On the other hand, we know that (det Mα )2 = ΔN (a)2 (cf. Sect. 1.1). Hence det A =
(det C) |det Mα |2 = c1 · · · cn d N (a)2 , where d = |Δ| = N (d). Let β = (b1 , . . . , bn )
be a basis of F/Q, complementary to α. Then β is a basis of the ideal a = a−1 d−1
3.3 Theta-Formulae (Arithmetic Form) 43
−1 −1
A−1 = A = tMα−1 C −1 M α = Mβ C −1t M β .
Let
v(x) = x Mβ , x = (x1 , . . . , xn ) ∈ Cn .
Then
n
A−1 (x) = ck−1 |vk (x)|2 , x ∈ Rn ,
k=1
= nk=1 bk m k ranges
just as in (3.2). When m = (m 1 , . . . , m n ) ranges over Zn , b
over a = a−1 d−1 . Since v(m) = (b(1) , . . . , b(n) ), A−1 (m) n
n = k=1 ck−1 |b(k) |2 . Fur-
t −1 (k)
thermore, (x, m) = (x Mα , m Mα ) = (u(x), v(m)) = k=1 u k b . Therefore
n
n
−1
θ (0, x; A ) = exp −π ck−1 |b(k) |2 + 2πi (k)
b uk ,
b∈a−1 d−1 k=1 k=1
and we have
r
1 1 +r2
r
(k) 2 (k) (k+r2 )
exp −π ck (u k + a ) + 2 ck (u k + a )(u k+r2 + a )
a∈a k=1 k=r1 +1
1
n
n
= exp −π ck−1 |b(k) |2 + 2πi (k)
b u k . (3.3)
c1 · · · cn d N (a)2 b∈a−1 d−1 k=1 k=1
The equality holds for any u = u(x) = x Mα , x ∈ Cn , and hence also for arbitrary
u = (u 1 , . . . , u n ) in Cn .
Let e1 , . . . , en be rational integers such that ek ≥ 0, 1 ≤ k ≤ n, and
ek = 0 or 1, 1 ≤ k ≤ r1 ,
ek ek+r2 = 0, r1 + 1 ≤ k ≤ r1 + r2 .
Let
ek = ek , 1 ≤ k ≤ r1 ,
ek = ek+r2 , ek+r
2
= ek , r1 + 1 ≤ k ≤ r1 + r2 .
n
(−2π ck (u k + a (k) ))ek exp(−π(· · · ))
a∈a k=1
1
n
= (2πib(k) )ek exp(· · · ),
c1 · · · cn d N ((a))2 b∈a−1 d−1 k=1
namely
n
e
n
ckk (u k + a (k) )ek exp(−π(· · · ))
k=1 a∈a k=1
(−i)e
n
= (b(k) )ek exp(· · · ),
c1 · · · cn d N (a)2 b∈a−1 d−1 k=1
where e = nk=1 ek = nk=1 ek .
Let r be any element of F. Put u k = r (k) in the above. Then we obtain the following
result.
Proposition 3.3 Let a be a non-zero ideal of F, and let r be an element of F. Then
n
e
n
n
ek
ckk (r (k) + a (k) ) exp −π ck |r (k) + a (k) |2
k=1 a∈a k=1 k=1
(−i) e
n
n
= (k) ek
(b ) exp −π ck−1 |b(k) |2 + 2πi T (r b).
c1 · · · cn d N (a)2 b∈a−1 d−1 k=1 k=1
Here c1 , . . . , cn > 0, c j = c j+r2 for r1 + 1 ≤j ≤ r1 + r
2 , e1 , . . . en and e1 , . . . , en ,
n n
are integers as stated in the above, and e = k=1 ek = k=1 ek . Furthermore, the
series on the both sides of the equality are absolutely convergent.
Now, let χ be a Hecke character of F, f the conductor of χ , f = N (f), and
{m 1 , . . . , m n ; s1 , . . . sn } the signature of χ . Then the conjugate χ of χ has the same
conductor f and the signature {m 1 , . . . , m n ; −s1 , . . . , −sn } where
m k = m k , 1 ≤ k ≤ r1 ,
m k = m k+r2 , m k+r2 = m k , r1 + 1 ≤ k ≤ r1 + r2 .
When a runs over a−1 f, and r over a set of representatives of a−1 modulo a−1 f, then
r + a obviously runs over all elements of a−1 (once over each element). Hence
θ (α; χ ) = g(α(r + a); χ ).
r a∈a−1 f
and consequently,
m k
θ (α; χ ) = tk χ f (αr ) (r (k) + a (k) )m k exp(· · · ).
r a∈a−1 f
|tk |2
ck = √
n
, ek = m k , ek = m k , 1 ≤ k ≤ n,
df
in Proposition 3.3, we see that the series a∈a−1 f is absolutely convergent. Since r
ranges over a finite set, it follows that the right-hand side of (3.4) is also absolutely
convergent, and hence θ (α; χ ) is well-defined. Furthermore, the same proposition
shows that
−1
(−i) M
m k m
θ (α; χ ) = tk χ f (αr )
ck k
r c1 · · · cn d N (a−1 f)2
n
(k) m k −1 (k) 2
× (b ) exp −π σk |b | + 2πi T (r b)
b∈ad−1 f−1 k=1
n √ 2
√ n
m
d f b(k) k π n d f b(k)
=h exp − √
−1 −1
tk n
d f k=1 tk
b∈ad f
× χ f (αr )ω0 (r b),
r
46 3 L-functions
where
(−i) M m
m −1
tk m k
n
h= tk k ck k √ , M= mk ,
c1 · · · cn d N (a−1 f)2
n
df k=1
ω0 (r b) = exp(2πi T (r b)).
n
1
c1 · · · cn d N (a−1 f)2 = |tk |2 · · d · N (a)−2 f −2
k=1
df
= V (α∞ )2 · V (α0 )2 · f
= V (α)2 f.
m
On the other hand, since tkm k is the complex conjugate of tk k , we have
tk m k
|tk |2 m k
m −1
m m −1
tk k ck k √n
= √n
ck k = 1.
df df
Hence
(−i) M
h= √ ,
V (α) f
where
(−i) M n
W (χ ; δ, ϕ) = √ C(χ ; δ, ϕ), M = mk ,
f k=1
Hence
|W (χ ; δ, ϕ)| = 1,
and
iM
W (χ ; δ, ϕ) = √ χf (−1)C(χ ; δ, ϕ) = (−1) M χf (−1)W (χ ; δ, ϕ).
f
Therefore
W (χ ; δ, ϕ) = W (χ; δ, ϕ).
Here V (α) is the volume of α, and g(α; χ ) is the function defined in Sect. 3.3.
Namely, let α = (a P ), tk = σk (a P∞,k ), 1 ≤ k ≤ r , tk+r2 = t k , r1 + 1 ≤ k ≤ r . Then
48 3 L-functions
n
m π 2
n
g(α; χ ) = χ f (α) tk k exp − √
n
|tk | .
k=1
d f k=1
where
χ (α0 )χ f (α0 )V (α0 )s , if a = ι (α0 ) ⊂ o,
f 0 (α, s; χ ) =
0, otherwise,
m
π
f ∞ (α∞ , s; χ ) = χ (α∞ ) tk exp − √
k
n
|tk | V (α∞ )s .
2
df
However, f 0 (α0 ξ ) = f 0 (α0 ) for any ξ in U0 , and μU0 (U0 ) = 1. Hence the inner
integral on the right is | f 0 (α0 )|. As μ J0 /U0 is the point measure on the discrete group
J0 /U0 , we see that
3.4 The Function f (α, s; x) 49
| f 0 (α0 , s; χ )|dμ J0 (α0 ) = | f 0 (α0 )|dμ J0 /U0 (α 0 )
J0 J0 /U0
= | f 0 (α 0 )|
α 0 ∈J0 /U0
= |χ (a)N (a)−s |.
a⊂o
It follows from Sect. 3.1 that J0 | f 0 (α0 , s; χ )|dμ J0 (α0 ) < +∞ if Re (s) > 1. We
also see from the above that in such a case,
f 0 (α0 , s; χ )dμ J0 (α0 ) = χ (a)N (a)−s
J0 a⊂o
= L(s; χ ).
n
tk m k m π
f ∞ (α∞ ) = |tk |isk tk k exp − √ |t k | 2
|tk |s
k=1
|t k | n
d f
n
s+m k +isk π
= |tk | exp − √ n
|tk | 2
k=1
df
r
π
= |tk |s+m k +isk exp − √ n
|t k | 2
k=1
df
r
2s+m k +m k +2isk 2π
× |tk | exp − √ n
|tk | .
2
k=r +1
df
1
where
π
Ik = |t|s+m k +isk exp − √ n
|t| 2
dμ(t), 1 ≤ k ≤ r1 ,
R∗ df
2π
= |t|2s+m k +m k +2isk exp − √ n
|t| 2
dμ(t), r1 + 1 ≤ k ≤ r,
C∗ df
where
m +m
r
√
n
ek s+ k 2 k +isk
df 2
A(s; χ ) = (2π )r2 ,
k=1
ek π
r
ek m k + m k
γ (s; χ ) = Γ s+ + isk .
k=1
2 2
s r1
γ (s; 1) = Γ Γ (s)r2 .
2
Now, since f (α, s; χ ) = f 0 (α0 , s; χ ) f ∞ (α∞ , s; χ ), J = J0 × J∞ , μ J = μ J0 ×
μ J∞ , we immediately obtain from the above the following result.
Proposition 3.5 The function f (α, s; χ ) is continuous on J × C. For Re (s) > 1,
it is integrable on J :
| f (α, s; χ )|dμ J (α) < +∞, Re (s) > 1.
J
Let
ξ(s; χ ) = f (α, s; χ )dμ J (α), Re (s) > 1.
J
Then
ξ(s; χ ) = A(s; χ )γ (s; χ )L(s; χ ), Re (s) > 1,
where A(s; χ ) and γ (s; χ ) are the functions defined in the above.
Let
O = {α | α ∈ J, V (α) > 1}.
O is an open set in J .
Proposition 3.6 For any complex s, f (α, s; χ ) is integrable on O, and
η(s; χ ) = f (α, s; χ )dμ(α)
O
we see that
| f (α, s; χ )| ≤ | f (α, c; χ )|
by Proposition 3.5. This proves the first half. At the same time, it was shown that
|η(s; χ )| ≤ | f (α, s; χ )|dμ J (α) ≤ K c < +∞
O
whenever Re (s) ≤ c.
Let C be an arbitrary circle on the s-plane, and let c > 1 be chosen so that Re (s) ≤
c for s ∈ C. For fixed α ∈ O, f (α, s; χ ) is of course holomorphic in s. Hence, for
any s inside of the circle C, we have
1 f (α, z; χ )
f (α, s; χ ) = dz.
2πi C z−s
Therefore
1 f (α, z; χ )
η(s; χ ) = dz dμ J (α).
2πi O C z−s
However,
f (α, z; χ ) Kc
dμ J (α)dz ≤ dz < +∞.
z−s |z − s|
O×C C
Note that the functions χ (α), V (α), and θ (α; χ ) depend only upon the coset α = α F ∗
of α mod F ∗ so that they may be also denoted by χ (α), V (α), and θ (α; χ ). Note
also that since
n
m π
g(α; χ ) = χ f (α) tk k exp − √n
|t k | 2
,
k=1
df
we have
1, if f = o and m k = 0, 1 ≤ k ≤ n,
εχ = g(0; χ ) =
0, otherwise.
54 3 L-functions
for any s with Re (s) > 1. The integral over O is equal to η(s; χ ). For the integral
over O −1 , we obtain
f (α, s; χ )dμ J (α) = −1 χ (α)(θ (α; χ ) − εχ )V (α)s dμ J (α),
O −1 O
It follows from the last remark in Sect. 3.4 that εχ = εχ = 1 or 0 and that if εχ = 1,
then f = o, m k = 0, 1 ≤ k ≤ n, so that
(−i) mk
W (χ ; δ, ϕ) = − √ C(χ ; δ, ϕ) = 1.
f
Hence we may write the integrand of the last integral in the form
−1
As O −1 → O (O → O) under α → α −1 δϕ (α → α −1 δϕ), we see that
−1
W χ (α)(θ (α −1 δϕ; χ ) − εχ )V (α)s−1 dμ J (α)
O
=W χ(α)(θ (α; χ ) − εχ )V (α)1−s dμ J (α)
O
= W η(1 − s; χ ),
with
W = W (χ ) = χ (δϕ)W (χ ; δϕ).
Problem 3.1 Prove that W (χ ) depends only upon χ , and is independent of the
choice of δ and ϕ such that δ, ϕ ∈ J1 ∩ (J0 × T ), ι (δ) = d, ι (ϕ) = f.
It follows that
ξ(s; χ ) = +
O −1
O
= η(s; χ ) + W η(1 − s; χ ) + εχ χ (α) V (α)s−1 − V (α)s dμ J (α),
O −1
for any s with Re (s) > 1. The computation also shows that the last integral converges
for Re (s) > 1.
Now, let
α = α 1 · τ x , α 1 ∈ J 1 , τ x ∈ T , x ∈ R+ ,
Since dμT (τ x ) → dx
x
under T ∼
= R+ , we see that
−1
εχ χ (α) V (α)s−1 − V (α)s dμ J (α)
O
1
dx
= εχ χ (α 1 )dμ J 1 (α 1 ) χ (τ x )(x s−1 − x s ) , (χ (τ x ) = 1).
J1 0 x
μ J 1 (J 1 ) < +∞.
and that εχ χ (α 1 )dμ J 1 (α 1 ) = εχ χ (α 1 )dμ J 1 (α 1 ) = 0.
J1 J1
Thus we see that
εχ v
ξ(s; χ ) = η(s; χ ) + W (χ )η(1 − s; χ ) + , Re (s) > 1. (3.5)
s(s − 1)
|W (χ )| = |χ (δϕ)||W (χ ; δ, ϕ)| = 1,
W (χ ) = χ (δϕ)W (χ; δ, ϕ) = W (χ ),
so that
W (χ )W (χ ) = 1.
On the other hand, if χ ≡ 1, then W (1) = 1. Hence we obtain from (3.5) that
εχ v
ξ(s; χ ) = W (χ ) W (χ )η(s; χ ) + η(1 − s; χ ) +
s(s − 1)
= W (χ )ξ(1 − s; χ ),
namely,
ξ(s; χ ) = W (χ )ξ(1 − s; χ ).
for Re (s) > 1. Here A(s; χ )−1 is a function of the form eas+b , where a, b = constant,
and γ (s; χ )−1 is a product of functions of the form Γ (cs + d)−1 . Since Γ (s)−1
is holomorphic on the entire s-plane, A(s; χ )−1 γ (c; χ )−1 is also holomorphic for
arbitrary s. It then follows from the above that the L-function L(s; χ ), which was
originally defined for s with Re (s) > 1, is a meromorphic function of s on the entire
s-plane, satisfying (3.6) for arbitrary s. If χ ≡ 1, then ξ(s; χ ) is an entire function
of s. Hence L(s; χ ) is also an entire function of s. Let χ ≡ 1. ζ F (s) = L(s; 1) is
holomorphic at s = 0, 1, and since
s −r1
γ (s; 1)−1 = Γ Γ (s)−r2 , r1 + r2 > 0,
2
with A(s; χ ) and γ (s; χ ) in Sect. 3.3. Then ξ(s; χ ) satisfies the functional equation
ξ(s; χ ) = W (χ )ξ(1 − s; χ ), s ∈ C,
|W (χ )| = 1, W (χ ) = W (χ ).
r
ek m k + m k
γ (s; χ ) = Γ s+ + isk
k=1
2 2
does not vanish for Re (s) > 1. Hence ξ(s; χ ) = A(s; χ )γ (s; χ )L(s; χ ) also has no
zero for Re (s) > 1. By the functional equation, we then see that ξ(s; χ ) has no zero
for Re (s) < 0. Hence the zeros of L(s; χ ) in the domain Re (s) < 0 are obtained
from those of γ (s; χ )−1 , and they can be easily determined because Γ (s)−1 has only
simple zeros at s = 0, −1, −2, . . .. Thus all “non-trivial" zeros of L(s; χ ) are in the
critical strip {s | 0 ≤ Re (s) ≤ 1}. The generalized Riemann hypothesis states that
all such zeros are on the straight line Re (s) = 21 .
In general, a locally compact group is compact if and only if it has a finite total
Haar measure. In the above, we have shown that μ J 1 (J 1 ) < +∞. Hence
(A) J (B) J
F∗ J (E × T )U
F∗ × T J E×T U
E×T W
1 1
3.5 Fundamental Theorems 59
F ∗ J /F ∗ × T ∼
= J /E × T, (E × T )U/E × T ∼
= U/W.
be defined by
λ : J /U → Rr .
Let H = λ(J1 ∩ J ) and L = λ(T ). Then H is the hyperplane consisting of all vectors
(x1 , . . . , xr ), xk ∈ R, satisfying rk=1 ek xk = 0, where ek = 1 or 2 according as
1 ≤ k ≤ r1 or r1 + 1 ≤ k ≤ r , and L is the straight line generated by ω = ( n1 , . . . , n1 ):
L = Rω. Therefore
Rr = H ⊕ L .
J /(E × T )U ∼
= (H ⊕ L)/(λ(E) ⊕ L) = H/λ(E).
E/W ∼
= EU/U ∼
= λ(E).
J 1 = J1 /F ∗ ∼
= J/F ∗ × T.
Hence
Thus
compactness of J 1 ⇐⇒ compactness of J/F ∗ J and H/λ(E).
UV
U V
U ∩V
Lemma 3.3 To each pair of groups X , Y , Y ⊂ X , in the diagram on the above, there
exists a unique Haar measure μ X/Y such that
(i) For X = G, U , V , U ∩ V , μ X/1 coincides with the given μ X ,
(ii) For any Z ⊂ Y ⊂ X , μ X/Y · μY/Z = μ X/Z ,
(iii) μU V /U ≈ μV /U ∩V , μU V /V ≈ μU/U ∩V in the obvious manner.
Proof Let
μU/U ∩V = μU /μU ∩V , μV /U ∩V = μV /μU ∩V , μU V /U ∩V = μU/U ∩V × μV /U ∩V ,
μ X/Y = μ X /μY ,
for any X , Y , Y ⊂ X . Then (i) is trivially satisfied. Also μ X/Y · μY/Z = (μ X /μY ) ·
(μY /μ Z ) = μ X /μ Z = μ X/Z (cf. Sect. 2.4). It is known that μU V /U ∩V /μU/U
∩V ≈
μV /U ∩V . However, μU/U ∩V = μU/U ∩V , μV /U ∩V = μU/U ∩V , and μU V /U ∩V =
μU V /μU ∩V = μU V /U ∩V . Hence μV /U ∩V ≈ μU V /U ∩V /μU/U ∩V = (μU V /μU ∩V )/
(μU /μU ∩V ) = μU V /μU = μU V /U (Sect. 2.4). Similarly, μU/U ∩V ≈ μU V /V . The
uniqueness is obvious, because there is no other choice for μU V and μ1 .
μ J = μU0 × μ J∞
μ F ∗ ×T = μ F ∗ × μT ,
μ E×T = μ E × μT ,
62 3 L-functions
where μ E is the point measure of the discrete group E. We apply the lemma to the
diagram (A) of Sect. 3.5, and fix the measures of the groups in that diagram by the
above lemma. Since J = J1 × T , we have
μ J/(F ∗ ×T ) = μ J /μ F ∗ ×T = (μ J1 × μT )/(μ F ∗ × μT )
≈ (μ J1 /μT ) × (μT /μT )
= μ J1 /μT = μ J 1 .
Hence
v = μ J 1 (J 1 ) = μ J/(F ∗ ×μT ) (J/(F ∗ × T )).
Also
Since μ F ∗ and μ E are both point measures, so is μ F ∗ /μ E . Hence μ(F ∗ ×T )/(E×T ) , and
consequently μ F ∗ J /J are also point measures. As
μ J/J = μ J /μ J = (μ J0 × μ J∞ )/(μU0 × μ J∞ )
≈ μ J0 /μU0
= μ J0 /U0 ,
μ J/J is a point measure. Since μ J/J = μ J/F ∗ J · μ F ∗ J /J , we see that μ J/F ∗ J is also
a point measure. Hence
v = μ J/(F ∗ ×T ) (J/(F ∗ × T ))
= μ J/F ∗ J (J/F ∗ J ) · μ F ∗ J /(F ∗ ×T ) (F ∗ J /(F ∗ × T ))
= [J : F ∗ J ]μ J /(E×T ) (J /(E × T ))
= hμ J /(E×T ) (J /(E × T )).
Let J∞ be the inverse image of (R+ )r under the isomorphism
Then
J∞ = U∞ × J∞ , J = U0 × J∞ = U × J∞
.
We denote by μU∞ and μ J∞ the translate of the standard measures on ({±1}r1 × Cr12 )
and (R+ )r respectively under the above isomorphism. Then μU∞ (U∞ ) = 2r1 (4π )r2 ,
and
3.6 The Residue of ζ F (s) at s = 1 63
μ J∞ = μU∞ × μ J∞ ,
because μ J∞ is, by definition, the translate of the standard measure on (R∗ )r1 × (C∗ )r2
under the same isomorphism. Put
μU = μU0 × μU∞ .
Then
μ J = μU0 × μ J∞ = μU0 × μU∞ × μ J∞ = μU × μ J∞ .
where μλ(E) is the point measure of λ(E), and μ L is the translate of μT under T → L,
namely, the measure defined by dμ L (xw) = d x, x ∈ R.
Now, let ε1 , . . . , εr −1 be units in E such that ω1 = λ(ε1 ), . . . , ωr −1 = λ(εr −1 )
form a basis of the lattice λ(E) over Z. Then ω1 , . . . , ωr −1 , ωr form a basis of Rr
over R. Let
dμRr (ξ ) = d x1 · · · d xr
64 3 L-functions
Hence
μRr /(λ(E)⊕L) (Rr /(λ(E) ⊕ L)) = (μRr /μλ(E)⊕L )(Rr /(λ(E) ⊕ L))
= c · (μRr /μλ(E)⊕L )(Rr /(λ(E) ⊕ L))
= c,
and consequently,
r −1
( j) 1
yj = xk log |εk | + xr · , j = 1, . . . , r.
k=1
n
where ⎛ ⎞
e1 log |ε1(1) | · · · er log |ε1(r ) |
⎜ .. .. ⎟
⎜ . . ⎟
R = det ⎜ ⎟ .
⎝ e log |ε(1) | · · · e log |ε(r ) | ⎠
1 r −1 r r −1
e1
··· er
n n
2r1 +r2 π r2 R
√ h.
w d
namely,
1 1
< ζ (s) < 1 + .
s−1 s−1
Hence
1 < (s − 1)ζ (s) < s,
and
lim (s − 1)ζ (s) = 1.
s→1
Now, let a denote the residue of ζ F (s) at s = 1. Then Theorem 3.3 shows that
√
w da
h = r +r r .
2 1 2π 2 R
Hence, if we can complete the value of a by a different method, we obtain a formula
which gives us the value of the class number h of F. Most of the classical class
number formulae for algebraic number fields are obtained by this principle.
Chapter 4
Some Applications
Let I(m) be as before the subgroup of I consisting of all ideals in I which are prime
to m. Then the map ι : J → I induces a surjective homomorphism J0 (m) → I(m)
with kernel U0 (m) so that we obtain an isomorphism
Now, let χ be a Hecke character of F with conductor f; f is the largest integral ideal
of F such that χ (U0 (f)) = 1. It follows from the above that χ defines a character
ψ = ψχ of I(f). Namely, given any ideal a of I(f), we define
ψ(a) = χ (α),
ψ(a) = 0,
whenever (a, f) = 1. Thus ψ(a) is defined for a which is either prime to f or integral.
In Chap. 3, we denoted the above (generalized) character of ideals by χ (a) or χ (a).
However, in the following, we shall denote it by ψ or ψχ to distinguish clearly from
the Hecke character χ . By the definition,
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2019 67
K. Iwasawa, Hecke’s L-functions, SpringerBriefs in Mathematics,
https://doi.org/10.1007/978-981-13-9495-9_4
68 4 Some Applications
L(s; χ ) = (1 − ψχ (p)N (p)−s )−1
p
= ψχ (a)N (a)−s , (Re (s) > 1)
a⊂o
Let X denote the group of all Hecke characters of F. (Since X is the group of all
characters of J/F ∗ × T , it is a locally compact abelian group in Pontryagin topology.
However, in the following we shall not be interested in such a topology on X .) Let
m be any non-zero integral ideal of F, and let X (m) denote the set of all Hecke
characters χ in X such that χ (U0 (m)) = 1, namely, such that m is contained in the
conductor f of χ ; m ⊂ f, f|m. X (m) is obviously a subgroup of X . If m1 ⊂ m2 then
U0 (m1 ) ⊂ U0 (m2 ), and X (m2 ) ⊂ X (m1 ). Clearly, X = m X (m).
Let χ ∈ X (m) so that m ⊂ f = conductor of χ . Let ψχ be the character of I(f)
corresponding to χ . Since I(m) ⊂ I(f), ψχ induces a character ψχm of I(m). Let
Ch(I(m)) denote the group of all characters of I(m). Then χ → ψχm obviously
defines a homomorphism
X (m) −→ Ch(I(m)).
Proof Let P1 , . . . , Ph consist of all infinite prime spots of F and of all finite spots
of F such that ν P (m) > 0. Let α = (a P ) be any idèle in J , and let 0 < ε < 1. By
the approximation theorem, there exists an element a in F such that
v P (a P − a) < εv P (a P ),
Now, suppose that ψχm = 1 for a character χ in X (m). Then χ (J0 (m)) =
ψχ (I(m)) = 1. By the definition of a Hecke character, χ (F ∗ ) = 1. Hence
m
χ (J0 (m)F ∗ ) = 1. Since χ is continuous, it follows from the above lemma that
χ (J ) = 1, namely, χ ≡ 1.
0
J∞ is the set of all α = (a P ) in J∞ such that σk (a P∞,k ) > 0 for k = 1, . . . , r1 . It is
an open subgroup of index 2r1 in J∞ , and it contains the subgroup T ; T ⊂ J∞ 0
.
0
Problem 4.1 Prove that J∞ is the connected component of 1 in J (and in J∞ ).
(U0 (m) × J∞ 0
) = F(m)(U0 (m) × J∞ 0
). Hence we have the above isomorphism.
∗
Since U0 (m) × J∞ is open in J , J/F (U0 (m) × J∞
0 0
) is discrete. However, F ∗ × T ⊂
F (U0 (m) × J∞ ), and J/F × T is compact (Theorem 3.2). Hence J/F ∗ (U0 (m) ×
∗ 0 ∗
0
J∞ ) is also compact. It then follows that J/F ∗ (U0 (m) × J∞ 0
) is a finite group.
Now, let
r(m) = ι (F(m));
r(m) is the group of all principal ideals (a) with a in F(m). Since ι (J0 (m) × J∞ 0
)=
ι (J0 (m)) = I(m), and since the kernel of J0 (m) × J∞ → I(m) is U0 (m) × J∞ , we
0 0
see that r(m) is a subgroup of I(m), and that the isomorphism J0 (m) × J∞ 0
/U0 (m) ×
0 ∼
J∞ = I(m) induces isomorphisms
X (m) ∼
= Ch(I(m)/r(m)).
that when χ ranges over X (m), ψχm gives us all characters of I(m) that are trivial
on r(m).
Proposition 4.3 X is the subgroup of all elements of finite order in X .
Proof Since I(m)/r(m)is a finite group, so are Ch(I(m)/r(m)) and X (m). Hence
every element in X = X (m) has a finite order. Suppose conversely that χ has a
finite order in X . The restriction χ of χ on J∞
0
then also has a finite order. However, as
0 ∼ + + ∗ ∗
J∞ = R × · · · × R × C × · · · × C , no non-trivial character of J∞ 0
has a finite
order. Hence χ = 1, χ (J∞ ) = 1, and χ belongs to X .
0
X (m2 ) X (m1 )
Since the map in the upper level is injective, so is the map in the lower level.
Let χ be a character in X (m) with conductor f. Then m ⊂ f, and f is the largest
integral ideal = {0} with the property χ ∈ X (f). Hence f = m if and only if χ ∈ /
X (m ) for any m = m, m ⊂ m ⊂ o.
A character in Ch(I(m)/r(m)) (namely, a character of I(m) which is triv-
ial on r(m)) is called primitive if it is not in the image of Ch(I(m )/r(m )) →
Ch(I(m)/r(m)) for any m = m, m ⊂ m ⊂ o. It follows from the above that a char-
acter χ in X (m) has conductor m if and only if the corresponding ψχm is primi-
tive in Ch(I(m)/r(m)). In such a case, ψχ (a) = ψχm (a) for a prime to f = m, and
ψχ (a) = 0 for integral a with (a, f) = 1. Therefore in order to obtain all L-functions
L(s; χ ) = (1 − ψχ (p)N (p)−s )−1 for the characters χ in X with conductor m, it
is sufficient to find all primitive characters of the finite group I(m)/r(m).
4.1 Hecke Characters and Ideal Characters 71
G be the Galois group of A/F; G is a compact abelian group in Krull topology. One
of the fundamental theorems in class field theory states that J/F ∗ J∞ 0 is topologically
isomorphic to G in a natural manner, or equivalently, that X is naturally isomorphic
to the group of all characters of the compact abelian group G. It was C. Chevalley who
first introduced the groups J , X , X etc., and simplified (clarified) the classical class
field theory which had been based upon the groups I(m)/r(m) and Ch(I(m)/r(m)),
m ⊂ o.
we obtain
−s
log L(s; χ ) = − log 1 − ψχ (p)N (p )
p
∞
1
= ψχ (p)m N (p)−ms , Re (s) > 1. (4.1)
p m=1
m
Here log z denotes the branch of the log function such that log 1 = 0. The double
sum in the above is absolutely convergent, and we see that
∞
1 −ms
L(s; χ ) = exp ψχ (p) N (p)
m
, Re (s) > 1.
p m=1
m
L(1 + ε; 1)3 L(1 + ε + i y; χ )4 L(1 + ε + 2i y; χ 2 )
≥ 1.
Proof Let A denote the product of the L-functions in the above. Then it follows from
(4.1) that
∞
A = exp a(p, m) ,
p m=1
72 4 Some Applications
with
1
a(p, m) = 3 + 4ψχ (p)m N (p)−imy + ψχ (p)2m N (p)−2imy N (p)−m(1+ε) .
m
If ψχ (p) = 0, then
Re (a(p, m)) = m3 N (p)−m(1+ε) > 0. If ψχ (p) = 0, then
ψχ (p) N (p)
m −imy
= 1. Hence, let ψχ (p)m N (p)−imy = eiθ , θ ∈ R. Then
1
Re (a(p, m)) = (3 + 4 cos θ + cos 2θ )N (p)−m(1+ε)
m
1
= (3 + 4 cos θ + 2 cos2 θ − 1)N (p)−m(1+ε)
m
2
= (1 + cos θ )2 N (p)−m(1+ε)
m
≥ 0.
It follows that
|A| = exp Re (a(p, m)) ≥ 1.
p m
L(1 + i y; χ ) = 0
Proof (1) Suppose that χ 2 ≡ 1. Then ψχ (p) = 0 or ±1 for any prime ideal p, and
−1 −1
1 − N (p)−s 1 − ψχ (p)N (p)−s
⎧
⎪ −s −1 −s −2s
⎨1 − N (p) = 1 + N (p) + N (p) + ··· ,
−s −2
= 1 − N (p) = 1 + 2N (p) + 3N (p)−2s + · · · ,
−s
⎪
⎩ −1
1 − N (p)−2s = 1 + N (p)−2s + · · · .
Let
Then we see from the above that am ∈ Z, am ≥ 0, and that am ≥ 1 if m = N (p)2 for
some prime ideal p. Hence, for any prime number p, there exists m such that m is a
power of p and am ≥ 1. It follows that
4.2 The Existence of Prime Ideals 73
∞
am = +∞.
m=1
Now, suppose that L(1; χ ) = 0. Since L(s; 1) has a unique simple pole at s = 1,
(Theorem 3.1), Z (s) is then holomorphic at any complex s. Hence
∞
Z (l) (2)
Z (s) = (s − 2)l , s ∈ C
l=0
l!
∞ ∞
1
Z (2) = (−2)l am (− log m)l m −2
l=0
l! m=1
∞
∞
2 l
l −2
= am (log m) m
l=0 m=1
l!
∞
∞
1
= (2 log m)l am m −2
m=1 l=0
l!
∞
= e2 log m am m −2
m=1
∞
= am
m=1
= +∞.
It follows that
It follows from the proposition that ξ(s; χ ) = 0 on the line Re (s) = 2. By the
functional equation, we then have ξ(s; χ ) = 0 also on the line Re (s) = 0. Therefore
the zeros of L(s; χ ) at s = 0 on Re (s) = 0 are given by the zeros of
r −1
−1 ek m k + m k
γ (s; χ ) = Γ s+ + isk
k=1
2 2
on the same line; at s = 0, we have to take into account the fact that ξ(s; χ ) has a
m +m
simple pole at s = 0. If Re (s) = 0, then Re ( e2k (s + k 2 k + isk )) = e4k (m k + m k ).
m +m
Hence Γ ( e2k (s + k 2 k + isk ))−1 has zero on Re (s) = 0 only when m k = m k = 0,
and in such a case, it has a unique simple zero at s = −isk . Thus the zeros of L(s; χ )
on Re (s) = 0 are explicitly determined. For example, if χ ≡ 1, then m k = m k =
sk = 0, 1 ≤ k ≤ r , and γ (s; 1)−1 has a zero of order r at s = 0. Hence ζ F (s) has a
zero of order r − 1 (r = r1 + r2 ) at s = 0, but ζ F (s) = 0 for s = 0 with Re (s) = 0.
Let m be any non-zero integral ideal of F. Let χ be a character in X (m) with
conductor f, ψ = ψχ the corresponding character of I(f) with ψ(a) = 0 for integral
a, not prime to f, and ψχm the restriction of ψχ on I(m).
Let Re (s) ≥ a > 21 . Then
4.2 The Existence of Prime Ideals 75
∞
1
ψ(p)m N (p)−ms
≤ N (p)−am
m
p m=2 p m=2
N (p)−2a
=
p
1 − N (p)−a
1
≤ −a
N (p)−2a < +∞,
1−2 p
∞
1
because N (a)−2a < +∞ for 2a > 1. Hence ψ(p)m N (p)−ms is a holo-
a⊂o p m=2
m
morphic function of s for Re (s) > 21 , and we obtain from (4.1) that
log L(s; χ ) ∼ ψχm (p)N (p)−s .
p
Here and in the following, we shall use the notation ∼ to indicate the fact that the
difference of the both sides of ∼ is holomorphic at s = 1. If p is in I(m), namely,
if p | m, then ψχ (p) = ψχm (p). However, there exist only a finite number of p’s such
that p|m. Hence it follows from the above that
log L(s; χ ) ∼ ψχm (p)N (p)−s .
p∈I(m)
Since χ is a character of X (m), we know from Sect. 4.1 that ψχm (r(m)) = 1 so that
ψχm may be considered as a character of I(m)/r(m). Hence, for any c in I(m)/r(m),
ψχm (c) is defined. In particular, ψχm (cp ) = ψχm (p) if cp denote the coset of p in
I(m)/r(m), and we have
log L(s; χ ) ∼ ψχm (cp )N (p)−s . (4.3)
p∈I(m)
Now, when χ ranges over all characters in X (m), ψχm ranges over all characters
of I(m)/r(m) (Proposition 4.2). Therefore it follows from (4.3) that
m m
ψ χ (c) log L(s; χ ) ∼ ψ χ (c)ψχm (cp )N p−s
χ∈X (m) χ p∈I(m)
= ψχ ψχm (c−1 cp )N (p)−s ,
p∈I(m) χ
= ψ(c−1 cp )N (p)−s ,
p∈I(m) ψ∈Ch(I(m)/r(m))
Hence m
ψ χ (c) log L(s; χ ) ∼ N N (p)−s .
χ∈X (m) p∈c
1
log L(s; 1) ∼ log .
s−1
Since ψχm (c) = 1 for χ ≡ 1, we obtain from the above the following theorem.
Theorem 4.1 Let m be any integral ideal of F. Let N be the order of I(m)/r(m),
and let c be any coset of I(m)/r(m). Then
1 1
N (p)−s ∼ log .
p∈c
N s−1
In particular, each coset of I(m)/r(m) (i.e., each ideal class of the ideal class group
I(m)/r(m)) contains infinitely many prime ideals in it.
which a constant α > 0. Let π(x) denote the number of prime ideals p of F such
that N (p) ≤ x (x ≥ 2). Then we obtain from the above that
x √ log x
du
π(x) = + O(xe−α n ).
0 log u
π(x)
lim x = 1.
x→∞
log x
G(m) is the multiplicative group of residue classes ū (= u mod m) such that (u, m) =
1. It is an abelian group of order ϕ(m) (ϕ = Euler’s function). Let u ∈ Z, u ≥ 1,
(u, m) = 1. Then the principal ideal (u) is in I(m), and the coset of (u) mod r(m)
will be denoted by c(u). Let v ∈ Z, v ≥ 1, (v, m) = 1. Then
c(uv) = c(u)c(v).
Λ(m 2 ) −→ Λ(m 1 ).
where the first isomorphism is defined in Sect. 4.1. Let χ → ψχm → λ under the
above maps. Then one sees easily that χ has the conductor (m) if and only if λ is
primitive in Λ(m). Suppose that this is so. Let u ∈ Z, u ≥ 1. If (u, m) = 1, then (u)
is in I(m), and ψχ ((u)) = ψχm ((u)) = λ(ū) = λ(u) because ū ↔ c(u) = (u) mod
r(m) under G(m) → I(m)/r(m). On the other hand, if (u, m) = 1, then ψχ ((u)) =
0, λ(u) = 0, both from the definition of ψχ ((u)) and λ(u). Hence
ψχ ((u)) = λ(u)
4.3 Dirichlet’s L-functions 79
Therefore L(s; χ ) is also denoted by L(s; λ). We see that when λ ranges over all
primitive characters in Λ(m), L(s; λ) gives us all L-functions L(s; χ ) of Q for the
characters χ in X (m) with conductor (m). The L-functions of Q were first introduced
by Dirichlet in this manner by means of the primitive characters of G(m). Hence
they are called Dirichlet’s L-functions.
Let χ → λ as above, with λ primitive in Λ(m). Let {m 1 ; s1 } be the signature of
χ . Since χ ∈ X = X , we have s1 = 0 (cf. Sect. 4.1). We shall next show how to
determine e = m 1 by means of λ. Let α ∈ J∞ , α = (1, . . . , 1, x), x ∈ R∗ . Then
m 1 e
x x
χ (α) = |x|is1 = .
|x| |x|
Let m ∈ Z, m ≥ 1. Consider
−1
L(s; χ ) = 1 − ψχ (( p)) p −s
χ∈X (m) χ p
−1
−s
= 1 − ψχ (( p)) p , Re (s) > 1.
p χ
Let f be the order of p̄ in G(m). Then f is the least positive integer such that p f ≡ 1
2πi
mod m, and is a factor of ϕ(m) = #G(m). We put ϕ(m) = f g. Let η = e f . Then
it is well-known that when λ ranges over Λ(m), λ( p̄) ranges over {ηk | 0 ≤ k < f },
g-times for each. Therefore the above product over Λ(m) equals
f −1 g
1 − ηk p −s = (1 − p − f s )g .
k=0
2πi
Let K be the cyclotomic field of mth roots of unity, i.e., K = Q(e m ). Then the
ideal ( p) is decomposed in K into the product of prime ideals as follows:
( p) = p1 · · · pg , N (pk ) = p f , k = 1, . . . , g;
here f and g are the integers defined in the above. It follows that
g
−s
1 − N (p) = 1 − N (pk )−s
p| p k=1
= (1 − p − f s )g
so that
1 − ψχ (( p)) p −s = 1 − N (p)−s .
χ∈X (m) p| p
This is proved for any prime number p which is prime to m. However, we can obtain
the same formula also for p which divides m. (The proof is similar, but is slightly
more complicated). Therefore
4.3 Dirichlet’s L-functions 81
−1
L(s; χ ) = 1 − N (p)−s
χ∈X (m) p p| p
−1
= 1 − N (p)−s
p
= ζ K (s), Re (s) > 1,
The product is taken our all Dirichlet’s L-functions with characters χ in X (m).
The above theorem has many interesting consequences. We know that both ζ K (s)
and L(s; 1) = ζQ (s) have a simple pole at s = 1 and that L(s; χ ), χ ≡ 1, is holo-
morphic at s = 1. Hence it follows from the above that L(1; χ ) = 0 for χ ≡ 1 (cf.
Sect. 4.2); for, otherwise, the zero of L(s; χ ) at s = 1 would cancel the pole of L(s; 1)
at s = 1, and ζ K (s) would be holomorphic at s = 1. Furthermore, since the residue
of ζQ (s) at s = 1 is 1, it also follows from the above that the residue of ζ K (s) at
s = 1 is equal to
L(1; χ ).
χ∈X (m)
χ≡1
χ≡1
for the class number h of K . In the next section, we shall compute the value of
L(1; χ ) (χ ≡ 1), and obtain the classical class number formula for the cyclotomic
field K (with m a prime number).
Actually, the equality in Theorem 4.2 is a special case of the following important
general result in class field theory. Let F be any number field and let Y be any finite
subgroup of the group X of Hecke characters of F. Since X is the torsion subgroup
of X , Y is contained in X . As stated in Sect. 4.1, X may be identified with the
character group of G, the Galois group of the maximal abelian extension A over F.
Let H be the closed subgroup of all σ in G such that χ (σ ) = 1 for χ in Y , and let
K be the field, F ⊂ K ⊂ A, corresponding to the subgroup H of G. Then K /F is a
finite abelian extension with Galois group G/H . For such an extension K of F, we
then have
ζ K (s) = L(s; χ ), s ∈ C.
χ∈Y
82 4 Some Applications
∼ G/H (not
Note that Y may be considered as the character group of G/H so that Y =
canonically) and [K : F] = [G : H ] = order of Y . In Theorem 4.2, we have F = Q
and Y = X (m).
e1 log |ε1(1) · · · er log |ε1(r ) |
⎜ .. .. .. ⎟
⎜ . . . ⎟
R(ε1 , . . . , εr −1 ) =
det ⎜ ⎟
,
⎝e log |ε(1) | (r ) ⎠
· · · er log |εr −1 |
1 r −1
e1
··· er
n n
Now, let l be a fixed prime number, l > 2, and let K denote the cyclotomic field
of l-th roots of unity;
2πi
K = Q(ζ ), ζ = e l .
Let
K 0 = Q(ζ + ζ );
n = [K : Q] = ϕ(l) = l − 1,
l −1
n 0 = [K 0 : Q] = ,
2
h = class number of K ,
h 0 = class number of K 0 ,
R = regulator of K ,
R0 = regulator of K 0 ,
d = |discriminant of K | = l n−1 ,
4.4 The Class Number Formula for Cyclotomic Fields 83
d0 = |discriminant of K 0 | = l n 0 −1 ,
E = group of units in K ,
E 0 = group of units in K 0 ,
W = group of roots of unity in K ,
W0 = group of roots of unity in K 0 .
E = {ζ a } × E 0 .
R = 2n 0 −1 R0
so that
√ n0
g/g0 = π n 0 2n 0 −1 /l l .
σa (ζ ) = ζ a .
Clearly σa depends only upon the residue class a of a mod l. The map a → σa then
defines an isomorphism
G(l) ∼
= G = Galois group of K /Q.
G(l)/{±1} ∼
= G 0 = Galois group of K 0 /Q.
We fix an integer t (a primitive root mod l) such that t is a generator of G(l), and
we put σ = σt so that G = {σ k }, 0 ≤ k < n. For any α in K , we then denote the
conjugates of α by
l = (1 − ζ ).
(1 − ζ t )/(1 − ζ ) = (ζ 2 − ζ − 2 )/(ζ 2 − ζ − 2 ).
1−t t t 1 1
η=ζ 2
1
(Note that ζ 2 is in W because ζ has order l, (l, 2) = 1). It follows from the above
that η is a unit, in fact, a real unit: η ∈ E 0 . It is called the circular unit of K (for fixed
t).
Since G(l) = (Z/lZ)∗ is a cyclic group of order n = l − 1, so is the character
group Λ(l) (and also X (l) = X (l)). Hence Λ(l) has a unique character ρ of order
2: ρ 2 ≡ 1, ρ ≡ 1. Since a → al , a ∈ Z, is such a character, we have
a
ρ(a) = , a ∈ Z.
l
Clearly ρ takes only real values (namely, 0, ±1). However, if conversely λ is a
character in Λ(l) taking only real values, then λ(a) = ±1 for (a, l) = 1. Hence
λ2 ≡ 1, and we see that either λ ≡ 1 or λ = ρ in the above. By this reason, we call
ρ the real character in Λ(l).
Let Λ0 (l) denote the set of all λ in Λ(l) such that λ(−1) = 1. Then Λ0 (l) is a
subgroup of order n 0 = l−1 in Λ(l), and it may be considered as the group of all
2
characters of G(l)/{±1̄}. Since ρ(−1) = −1
l−1
l
= (−1) 2 , we see that ρ ∈ Λ0 (l) or
ρ∈/ Λ0 (l) according as l ≡ 1 mod 4 or l ≡ 3 mod 4.
Let
2πia
ω(a) = ζ a = e l , a ∈ Z.
Then ω(a) depends only upon the coset ā of a mod l, and it defines a character of
the additive group of the finite ring (field) R = Z/lZ. For any λ in Λ(l), we put
l−1
v(λ) = ω(a)λ(a) = λ(a)ζ a .
a a=0
4.4 The Class Number Formula for Cyclotomic Fields 85
v(λ) is nothing but the Gaussian sum G(ω, λ) on R defined in Sect. 2.3, and we see
immediately from the lemmas proved there that
λ(a)ζ ua = λ(u)v(λ), (u, l) = 1 or λ ≡ 1, (4.5)
a
v(λ) = 0, λ = 1,
√
|v(λ)| = l, λ ≡ 1.
It follows that
v(λ) = λ(a)ζ −a = λ(−1)v(λ). (4.6)
a
√ = ρ(−1)v(ρ).
In particular, v(ρ) √ Hence v(ρ)2 = ρ(−1)|v(ρ)|2 = ρ(−1) · l = ±l,
and v(ρ) = ± l or ±i l according√as l ≡ 1√mod 4 or l ≡ 3 mod 4. It is not quite
easy to determine the sign ± before l and i l. One can show however that
√
l, l ≡ 1 mod 4,
v(ρ) = √ √
i l, l ≡ 3 mod 4, ( l > 0).
Applying a similar argument to K 0 (or, from the general theorem stated at the end
of Sect. 4.4), we also see that
g0 h 0 = L(1; λ).
λ∈Λ0 (l)
λ≡1
For any ε > 0, the above series converges uniformly for s ≥ ε so that f (s) defines
a continuous function of s for s > 0. In particular
∞
lim f (s) = f (1) = ζ m m −1 = − log(1 − ζ ),
s→1
m=1
86 4 Some Applications
where
π π
log(1 − ζ ) = log |1 − ζ | + i arg(1 − ζ ), − < arg(1 − ζ ) < .
2 2
Proof Let S(a) = am=1 ζ m for a ≥ 1, and S(0) = 0. If ζ is an N -th root of unity,
a+N
ζ = 1, then ζ m = 0. Hence S(a) = S(b) for a ≡ b mod N , and one obtains
m=a
immediately that
|S(a)| ≤ N , a ∈ Z.
Let s ≥ ε, b ≥ a. Then
b
b
ζ m m −s = (S(m) − S(m − 1)) m −s
m=a m=a
b−1
= −S(a − 1)a −s + S(b)b−s + S(m) m −s − (m + 1)−s .
m=a
b−1
m −s
−s −s
ζ m
≤ Na + Nb + N S(m) m −s − (m + 1)−s
m=a
m=a
= 2N a −s
≤ 2N a −s −→ 0, for a −→ +∞
∞
This proves the uniform convergence. The fact that ζ m m −1 = − log(1 − ζ ) is
m=1
well-known in the function theory.
2πi
Now, let λ ≡ 1, s > 1, and ζ = e l . Using
1 (a−m)u
l−1
1, m ≡ a mod l,
ζ =
l u=0 0, m ≡ a mod l,
l−1 ∞
l−1
1
= λ(a) ζ (a−m)u m −s
a=0 m=1
l u=0
4.4 The Class Number Formula for Cyclotomic Fields 87
∞
1
l−1 l−1
= λ(a)ζ au ζ −mu m −s
l u=0 a=0 m=1
∞
v(λ)
l−1
= λ(u) ζ −mu m −s (by (4.5))
l u=0 m=1
∞
v(λ)
l−1
= λ(u) ζ −mu m −s (∵ λ(0) = 0).
l u=1 m=1
v(λ)
l−1
L(1; λ) = lim L(s; λ) = − λ(u) log(1 − ζ −u )
s→1 l u=1
v(λ)
l−1
=− λ(u) log(1 − ζ −u ) + λ(−1) log(1 − ζ u ) .
2l u=1
α = (1 − ζ )(1 − ζ −1 ).
Then
v(λ)
l−1
L(1; λ) = − λ(u) log(1 − ζ −u ) + log(1 − ζ u )
2l u=1
v(λ)
l−1
=− λ(u) log σu (α)
2l u=1
v(λ)
l−1
=− λ(t)k log α (k) ,
2l k=1
n 0 −1
v(λ)
L(1; λ) = − λ(t)k log α (k) .
l k=0
It follows that
h 0 = g0−1 L(1; λ) =
g0−1
|L(1; λ)|
λ∈Λ0 (l) λ
λ≡1
88 4 Some Applications
−1
n −1
2n 0 −1 R0 1 n 0 −1
0
= √ n 0 −1 √
λ(t)k log α (k)
l l
k=0
n −1
−1
0
= 2n 0 −1 R0
λ(t)k log α (k)
.
k=0
2πi
m
When λ ranges over the character group Λ0 (l), λ(t) takes the values e n0 , m =
0, 1, . . . , n 0 − 1. Hence by a well-known theorem on determinants, we have
⎛ ⎞
⎜
n 0 −1
⎜ log α (1) log α (2) · · · log α (0) ⎟ ⎟
λ(t) α (k)
det ⎜
.
k
log .. .. .. .
.. ⎟
λ∈Λ0 (l) k=0
⎝ . . . ⎠
log α (n 0 −1) log α (0) · · · log α (n 0 −2)
(1−t)t k k+1
1−ζ t 1−ζ t
, η(k) = ζ
1−t
Since η = ζ 2
1−ζ
2 ,
1−ζ t k
(k)
2 t k+1 −t k+1
η
= (1 − ζ k )(1 − ζ k )
(1 − ζ t )(1 − ζ −t )
(k+1)
α
= (k)
α
Therefore the above determinant equals
⎛ ⎞
2 log |η(0) |
2 log |η(1) | · · · 2 log |η(n 0 −1) |
⎜ 2 log |η(1) |
2 log |η(2) | · · · 2 log |η(0) | ⎟
⎜ ⎟
⎜ .. .. .. .. ⎟
det ⎜ . . . . ⎟.
⎜ ⎟
⎝2 log |η(n 0 −2) | 2 log |η(n 0 −1) | · · · 2 log |η (n 0 −3) ⎠
|
log α (n 0 −1) log α (0) ··· log α (n 0 −2)
0 −1
n
Using log |η(k) | = log N K 0 /Q (η) = log 1 = 0 (because η is a unit), we then see
k=0
that the determinant is again equal to the following:
⎛ ⎞
0 log |η(1) | · · · log |η(n 0 −1) |
n −1 ⎜0 log |η(2) | · · · log |η(0) | ⎟
0 ⎜ ⎟
⎜ ⎟
±2n 0 −1 log α (k) det ⎜ ... ..
.
..
.
..
. ⎟
⎜ ⎟
k=0 ⎝0 log |η(n 0 −1) | · · · log |η(n 0 −3) |⎠
1 log α (0) · · · log α (n 0 −2)
4.4 The Class Number Formula for Cyclotomic Fields 89
⎛ ⎞
0 log |η(1) | · · · log |η(n 0 −1) |
n −1 ⎜0 log |η(2) | · · · log |η(0) | ⎟
0 ⎜ ⎟
⎜ ⎟
= ±2n 0 −1 log α (k) det ⎜ ... ..
.
..
.
..
. ⎟
⎜ ⎟
k=0 ⎝0 log |η(n 0 −1) | · · · log |η(n 0 −3) |⎠
1 0 ··· 0
⎛ ⎞
log |η(0) | log |η(1) | · · · log |η(n 0 −1) |
n −1 ⎜ log |η(1) | log |η(2) | · · · log |η(0) | ⎟
0 ⎜ ⎟
n 0 −1 (k) ⎜ .. .. .. .. ⎟
= ±2 log α det ⎜ . . . . ⎟
⎜ ⎟
k=0 ⎝log |η(n 0 −2) | log |η(n 0 −1) | · · · log |η(n 0 −3) |⎠
1
n0
1
n0
··· 1
n0
n −1
0
n 0 −1 (k)
= ±2 log α R(η(0) , . . . , η(n 0 −2) ).
k=0
If λ ≡ 1, then
0 −1
n 0 −1
n
λ(t)k log α (k) = log α (k) = log N K 0 /Q (α)
k=0 k=0
= log N K /Q (1 − ζ ) = log p
= 0.
n 0
(k)
λ(t) log α
= 2n 0 −1 R(η(0) , . . . , η(n 0 −2) ),
k
Since η(0) · · · η(n 0 −1) = N K 0 /Q (η) = 1, E(η(0) , . . . , η(n 0 −2) ) is the subgroup of E 0
generated by all conjugates η(k) of η, and E(η(0) , . . . , η(n 0 −2) )W0 is the subgroup of
E 0 generated by the conjugates ±η(k) of ±η. By the remark stated at the beginning
of this section, we have the following formula for the class number h 0 of K 0 :
h 0 = [E 0 : H ];
here E 0 is the group of all units in K 0 , and H is the subgroup of E 0 generated by the
conjugates ±η(k) of ±η.
We shall next compute h 1 = hh0 from
90 4 Some Applications
h g0
h1 = = L(1; λ)
h0 g λ∈Λ(l)
λ∈Λ
/ 0 (l)
√ n0
l l
= n n −1 L(1; λ).
π 02 0 λ
v(λ)
l−1
L(1; λ) = − λ(u) log(1 − ζ −u ) − log(1 − ζ u )
2l u=1
v(λ)
l−1
=− λ(u) · 2i arg(1 − ζ −u ).
2l u=1
However,
1 u
arg(1 − ζ −u ) = 2π − 2π
2 l
u
= π −π .
l
O• 1
−u
l−1
Since λ(u) = 0 for λ = 1 (or λ = 1), we obtain
u=1
πiv(λ)
l−1
L(1; λ) = uλ(u).
l 2 u=1
Therefore √ n0
π n0 i n0 l−1
l l
h 1 = n n −1 · 2n · v(λ) · uλ(u).
π 02 0 l 0 λ λ u=1
Let l ≡ 3 mod 4, n 0 = l−1 ≡ 1 mod 2. Then ρ ∈ / Λ0 (l), and the characters in the
2 √
above product consist of the real character ρ and n 02−1 pairs {λ, λ}. Since v(ρ) = i l,
we have √ n 0 −1 √ n0
v(λ) = i l(−l) 2 = i n 0 l .
λ
and we obtain the following formula for h 1 = hh0 (the first factor of the class number
h of K ):
h 1
l−1
h1 = = 2l − uλ(u) .
h0 λ∈Λ(l)
2l u=1
λ(−1)=−1
By the same method, we can obtain similar class number formulas for more
fields as well as their subfields. For example, let l ≡ 3
general types of cyclotomic√
mod 4, l > 3. Then F = Q( −l) is a subfield of K in the above. Let h be the class
number of F, and let
π
g = √ .
l
Then
g h = L(1; ρ),
namely, √
l
h = L(1; ρ).
π
However, we have seen in the above that
√ l−1
πiv(ρ) πi(i l) u
l−1
L(1; ρ) = uρ(u) = u.
l 2 u=1 l2 u=1
l
Hence
1 u
l−1
h = − u.
l u=1 l
92 4 Some Applications
For l = 3, we have g = 3π√l so that we have to multiply the right-hand side of the
above by 3. (We then obtain√ h = 1).
Let l ≡ 1 mod 4. Then Q( l) is contained
√ in the above cyclotomic field K , and we
can obtain a class number formula for Q( l) similar to the formula h 0 = [E 0 : H ]
for K 0 .
Bibliography
1. E. Artin, The theory of algebraic numbers, Notes by Gerhald Würges from lectures held at the
Mathematisches Institut, Göttingen, Germany in the Winter-Semester, 1956/57, Translated and
distributed by George Sticker. Göttingen 172 pages, 1959.
2. E. Hecke, Vorlesungen über die Theorie der algebraischen Zahlen. Chelsea, 1948.
3. J. Tate, Fourier analysis in number fields and Hecke’s zeta-functions, in “Algebraic Number
Theory” Edited by J. Cassels and A. Frohlich, pp. 305–347, Thompson Washington, D.C.,
1967.
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2019 93
K. Iwasawa, Hecke’s L-functions, SpringerBriefs in Mathematics,
https://doi.org/10.1007/978-981-13-9495-9