ALGEBRAIC NUMBER THEORY

RAGHAVAN NARASIMHAN
S. RAGHAVAN
S. S. RANGACHARI
SUNDER LAL
Tata Institute of Fundamental Research, Bombay

School of Mathematics
Tata Institute of Fundamental Research, Bombay

1966
 PREFACE

THIS pamphlet contains the notes of lectures given at a Summer

School on the Theory of Numbers at the Tata Institute of Fundamen-
tal Research in 1965. The audience consisted of teachers and students
from Indian Universities who desired to have a general knowledge of the
Bibliography
subject. The speakers were Raghavan Narasimhan, S. Raghavan, S. S.
Rangachari and Sunder Lal.
Chapter 1 sets out the necessary preliminaries from set theory and [1] N. Bourbaki: Algébre, Chapters I and II, Paris (1949).
algebra; it also contains some elementary number-theoretic material.
Chapter 2 deals with general properties of algebraic number fields; it [2] P.R. HALMOS: Finite-dimensional Vector Spaces, Van Nostrand,
includes proofs of the unique factorization theorem for ideals, the finite- Princeton (1958).
ness of class number, and Dirichlet’s theorem on units. Chapter 3 gives [3] N. JACOBSON: Lectures in Abstract Algebra, Vol. I and II, Van
a slightly more detailed analysis of quadratic fields, in particular from Nostrand, Princeton (1953).
the analytic aspect; the course ends with Dirichlet’s theorem on the
infinitude of primes in an arithmetic progression. [4] B.L. VAN DER WAERDEN: Modern Algebra, Ungar, New York
(1950).
[5] E. ARTIN: Theory of Algebraic Numbers, (Lecture notes)
Göttingen, 1959.
[6] H. HASSE: Vorlesungen über Zahlentheorie, Springer-Verlag,
(1950).
[7] E. HECKE: Vorlesungen über die Theorie der algebraischen Zahlen,
Chelsea, New York, (1948).
[8] D.HILBERT: Berichtüber die Theorie der algebraishchen
Zahlkörper, Jahresbericht der D.M.V., Bd.4 (1897), pp.177–
546.
[9] S. LANG: Algebraic Numbers, Addison-Wesley, 1964.
[10] H. WEYL: Algebraic Theory of Numbers, Ann. of Math. Studies,
Princeton, 1940.
[11] C.L. SIEGEL: Lectures on Analytic Number Theory, New York,
(1945).
[12] C.L. SIEGEL: Lectures on Advanced Analytic Number Theory, Tata
Institute, Bombay, (1961).

91
90 CHAPTER 3. QUADRATIC FIELDS

Theorem 3.5 (Dirichlet.) Let 1 ≥ 1, D > 1 be integers with (l, D) = 1.

Then there exist infinitely many primes p ≡ l( mod D).

Proof: For s > 1, we have, by (3.25)

X X 1
χ̄(l) log L(s, χ) = h + R1 (s)
χ∈Ĝ pm ≡1(D)
ps Chapter 1
where |R1 (s)| ≤ h p ps (p1s −1) which is bounded as s → 1 + 0.
P

Now, the term with χ = χ0 is log L(s, χ0 ) which tends to ∞ as

s → 1 + 0 by Lemma 3.6. Every other term χ̄(l) log L(s, χ) remains
PRELIMINARIES
bounded as s → 1 + 0 since L(1, χ) 6= 0. Hence
X
χ̄(l) log L(s, χ) → ∞ as s → 1 + 0;
χ
1.1 Sets and maps
A set is a collection of objects which are called the elements of the set.
since R1 (s) is bounded, we conclude that pm ≡l(D) p1s → ∞ as s → 1+0.
P
We shall suppose that if any object is given, we can decide whether it
Clearly this implies Theorem 3.5.
belongs to the set or not. The set of all rational integers (i.e. integers
Proposition 3.16 follows immediately from Proposition 3.9 and The-
positive, negative and zero) is denoted by Z, the set of all non-negative
orem 3.5, since there exist infinitely many primes p for which ( dp ) = 1
integers by Z+ , the set of all rational numbers by Q, the set of all real
and infinitely many primes q for which ( dq ) = −1. numbers by R, and the set of all complex numbers by C.
If x is an element of a set A, we write x ∈ A. If x is not an element
of A, we write x ∈ / A. If P is a property, the set of all objects with the
property P will be denoted by {x | x possesses the property P }. Thus
{x | x ∈ Z, x < 0} is the set of all negative integers. The set which does
not contain any element is called the empty set and is denoted by the
symbol ∅.
Let X and Y be two sets. If every element of X is an element of Y ,
we say that X is a subset of Y (or X is contained in Y ) and write X ⊂ Y
or Y ⊃ X. If X ⊂ Y, X 6= Y , we say that X is properly contained in Y .
It is clear that if X ⊂ Y and Y ⊂ X, then X = Y . If X and Y are two
sets, we define

(i) the union X ∪ Y of X and Y as the set

{z | z ∈ X or z ∈ Y };

(ii) the intersection X ∩ Y of X and Y as the set

{z | z ∈ X and z ∈ Y };

3
4 CHAPTER 1. PRELIMINARIES 3.6. PRIMES IN AN ARITHMETIC PROGRESSION 89

Q
(iii) the cartesian product X × Y of X and Y as Proof: Clearly, the series expansion for f (s) = χ L(s, χ) can be ob-
1 xk
in ex = ∞
X P
{(x, y) | x ∈ X and y ∈ Y }. tained by formal substitution of x = h ms k=0 k! .
m
mp
p ≡1 (D)
We say X and Y are disjoint if X ∩ Y = ∅. If X ⊂ Y , we define the Since all terms in the series for x are non-negative the corollary is proved.
complement, Y − X, as the set {z | z ∈ Y and z ∈
/ X}. Now we have
X χ(p) X X χ(pm )
log L(s, χ) = +
1.2 Maps p
ps p
mpms
m≥2

Let X and Y be two sets. A map f : X → Y is an assignment to each If σ = Re s we have

x ∈ X, of an element f (x) ∈ Y. If A is a subset of X, the image f (A) is
XX 1 XX 1 X 1
≤ =
the set {f (x) | x ∈ A}. The inverse image of a subset B of Y , denoted by p m≥2
|mpms | p
p mσ
p
pσ (pσ − 1)
m≥2
f −1 (B), is the set {x | x ∈ X, f (x) ∈ B}. The map f is said to be onto
or surjective, if f (X) = Y ; if f (x) = f (y) implies x = y, then f is said to This latter series converges for σ = 21 . Hence, for σ > 1,
be one-one or injective. If f : X → Y, g: Y → Z are two maps, we define X 1 X X 1
the composite (g ◦ f ): X → Z as follows: for x ∈ X, (g ◦ f )(x) = g(f (x)). log ζ(s) = s
+ R(s); log L(s, χ) = h + R1 (s) (3.25)
p
p χ m
ps
The map X → Xwhich associates to each x ∈ X, the element x itself p ≡1 (D)

called the identity map of X and denoted by IX (or by I, if there is where |R(s)| ≤
P 1
|R1 (s)| ≤ h p pσ (p1σ −1) . In particular,
P
p pσ (pσ −1) ,
no confusion). If f : X → Y is both one-one and onto, there is a map P 1
since ζ(s) → ∞ as s → 1+0 and R(s) is bounded as s → 1+0, p ps →
from Y → X, denoted by f −1 , such that f ◦ f −1 = IY , f −1 ◦ f = IX . P 1
∞ as s → 1 + 0, so that p p = ∞
The map f −1 is called the inverse of f . If A is a subset of X, the map
j = jA : A → X which associates to each a ∈ A the same element a in Lemma 3.8 If f (s) = χ∈Ĝ L(s, χ) = ∞
Q P −s
1 cm m , then cm ≥ 0, and
X is called the inclusion map of A in X. If f : X → Y is any map, the P∞ −s = ∞ for s = 1
c
m=1 m m φ(D) where φ(D) is the order of the group
map f ◦ jA : A → Y is called the restriction of f to A and is denoted by
G.
f | A.
Proof: We have already seen P that cm ≥ 0P(by the corollary to
Lemma 3.2). Now, for real s, pm ≡1 (D) mp1ms ≥ p>D φ(D)p1φ(D)s since,
1.3 Equivalence Relations
for p > D, we have pφ(D) ≡ 1 (D). Since p>D 1p = ∞ it follows that
P

Definition 1.1 Let X be a set. An equivalence relation in X is subset

P 1 1 x
pm ≡1( mod D) mpms = ∞ for s = φ(D) . Since e = 1+x+· · ·, and the se-
R of X × X such that ries for f (s) is obtained by formal substitution of x = h pm ≡1 (D) mp1ms
P

in the series for ex , ∞ −s diverges for s = 1 .

P
(i) for every x ∈ X, (x, x) ∈ R; m=1 cm m φ(D)

(ii) if (x, y) ∈ R, then (y, x) ∈ R; and Proposition 3.17 We have L(1, χ) 6= 0 for χ 6= χ0 .
Q Q
Proof: Consider f (s) = χ L(s, χ) = L(s, χ0 ) χ6=χ0 L(s, χ). Since
(iii) if (x, y) ∈ R and (y, z) ∈ R then (x, z) ∈ R. Q −s
L(s, χ0 ) = ζ(s) p|D (1 − p ), this is meromorphic in Re s > 0, with a
We say that x is equivalent to y with respect to R and write xRy if single simple pole at s = 1. If L(1, χ) = 0 for some χ 6= χ0 it follows
(x, y) ∈ R. Then the conditions above simply require that that f (s) is holomorphic for Re s > 0. Since, for Re s > 1, f (s) =
cm m−s , cm ≥ 0, it follows from Lemma 3.4 that the abscissa of
P
(i) every element x is equivalent to itself (reflexivity), convergence of this series is ≤ 0. This contradicts Lemma 3.8.
88 CHAPTER 3. QUADRATIC FIELDS 1.4. ABELIAN GROUPS AND HOMOMORPHISMS 5

where χ0 is the principal character defined by χ0 (m) = 1 if (m, D) = (ii) if x is equivalent to y, y is equivalent to x (symmetry), and
1, χ0 (m) = 0 otherwise. Hence, as in Proposition 3.11, we can show
that (iii) if x is equivalent to y and y to z then x is equivalent to z.
Xm  1 (transitivity).
χ(k) ≤ L for any m.
 
2

k=1 Let R be an equivalence relation in a set X. Then for any x ∈ X,
Hence, by Proposition 3.15, the series defining L(s, χ) converges for the set of all elements of X equivalent to x with respect to R is called
Re s > 0, uniformly in any bounded subset of Re s ≥ δ > 0. Hence we the equivalence class of R containing x and is denoted by x̄. Consider
have the family of distinct equivalence classes of X with respect to R. It is
easy to verify that they are pairwise disjoint and that their union is X.
Lemma 3.5 If χ 6= χ0 , L(s, χ) is holomorphic for Re s > 0. Moreover, The set of these equivalence classes x is called the quotient of X by R
and is denoted by X/R.
−1
Y Y
L(s, χ) = (1 − p−s ) = ζ(s) (1 − p−s ).
p∤D p|D
Example 1.1 The subset R ⊂ X × X consisting of elements (x, x), x ∈
X is an equivalence relation. This is called the identity relation.
From Lemma 3.5 and from (3.20), we deduce
Example 1.2 Let n ∈ Z, n > 0. Consider the set in Z × Z of pairs of
Q 1 integers (a, b) such a − b is divisible by n. This is an equivalence relation
Lemma 3.6 lims→1+0 (s − 1)L(s, χ0 ) = p|D (1 − p ); in particular,
L(s, χ0 ) → ∞ as s → 1 + 0. in Z and the quotient of Z by this relation is denoted by Z/(n) or Zn .

Now, for Re s > 1,

1.4 Abelian groups and homomorphisms
X X χ(pm ) ∞
X 1
log L(s, χ) = log = Definition 1.2 Let G be a nonempty set and ψ: G × G → G a mapping.
p
1 − χ(p)p−s p
mpms
m=1 Let, for x, y ∈ G, ψ((x, y)) be denoted by x·y or xy. Then the pair (G, ψ)
is said to be an abelian group if the following conditions are satisfied:
Let l ≥ 1 be any integer with (l, D) = 1. Using the orthogonality
relations, (Proposition 1.10, Chapter 1), we obtain (a) x(yz) = (xy)z for every x, y, z in G(associativity),

Lemma 3.7 For Re s > 1 we have (b) there exists an element e, called the identity element of G, which
X X 1 satisfies ex = xe = x for every x in G.
log L(s, χ) = h
χ
mpms (c) for every x ∈ G, there exists in G an element x−1 , called the inverse
pm ≡1 (D)
X X 1 of x, such that xx−1 = x−1 x = e, and
χ̄(l) log L(s, χ) = h
χ
mpms (d) for every x, y ∈ G, xy = yz (commutativity).
pm ≡l (D)

Here the summation is over all the characters of G, and h is the order Remark 1.1 We often abbreviate (G, ψ) to G when it is clear from the
of the character group Ĝ of G. context to which map ψ we are referring.

Remark 1.2 The map ψ is called the composition law in G. It is also

Q
Corollary 3.2 The function f (s) = χ L(s, χ) has, for Re s > 1, an
cm m−s where cm ≥ 0. called the multiplication in G.
P
expansion as a Dirichlet series
6 CHAPTER 1. PRELIMINARIES 3.6. PRIMES IN AN ARITHMETIC PROGRESSION 87

Remark 1.3 The identity element is unique. In fact, if there is an Since all terms here are non-negative, we may rearrange the series
element e′ in G such that condition (b) above is valid for every x in G as we like; hence the repeated series is equal to
with e replaced by e′ , we have, in particular, e = ee′ = e′ . ∞ ∞
X X (s1 − s)k
am m−s1 log m)k < ∞, s1 − δ < s < s1 .
Remark 1.4 The inverse of any element is unique. m=1
k!
k=0

Remark 1.5 In view of associativity, we define xyz = (xy)z = x(yz) But the inner sum is e(s1 −s) log m = ms1 −s . Hence
for every x, y, z in G. ∞
X ∞
X
am m−s1 · ms1 −s = am m−s < ∞ for s1 − δ < s < s1 .
More generally, the product x1 x2 · · · xn is well defined, where x1 , x2 , m=1 m=1
. . . , xn ∈ G. (Proof by induction). In particular, for any x ∈ G, we set Since s1 − δ < s0 , this contradicts the definition of s0 and f must be
singular at s = s0 .
xm = xx · · · x (m times) for m > 0 in Z, P Q
In what follows, sums and products of the type p , p will always
x0 = e, be taken over the primes p > 1. (If any additional conditions are to
xn = (x−1 )−n for n < 0 in Z. imposed, these will be indicated below the summation or product.)
Let D > 1 be a positive integer. If a ∈ Z is such that (a, D) = 1,
It is also customary to write the composition law in an abelian group then ab+ Dc = 1 for some b, c ∈ Z. Then the residue class ā containing a
additively, i.e. to write x + y for what has been denoted by x · y above. has an inverse b̄ in Z/(D). If a′ ∈ ā, then clearly (a′ , D) = 1. By a prime
In this case, one writes 0 for e, −x for x−1 , mx for xm , and refers to the residue class modulo D, we mean a residue class x̄ modulo D such that
composition law as addition. for any x′ ∈ x̄, we have (x′ , D) = 1. Clearly the prime residue classes
modulo D form a finite group G ;under the multiplication induced from
Definition 1.3 An abelian group G is said to be finite if it consists of
Z/(D). Let χ be a character of G. We define a function, also denoted
only finitely many elements; the number of elements of a finite group is
by χ on Z, by
referred to as its order. We say G is infinite if it is not finite.
χ(m) = χ(m̄) if (m, D) = 1, where m̄ is the residue class of m,
Example 1.3 The set Z (Q, R, C) of integers (rational numbers, real
χ(m) = 0 if (m, D) > 1.
numbers, complex numbers respectively) with the ‘usual’ addition as com-
position law is an abelian group. For s ∈ C, Re s > 1, we define
∞
Example 1.4 The set Z/(n) in Example 1.2 on page 5 can be seen to be
X
L(s, χ) = χ(m)m−s .
an abelian group with addition (+) defined by x̄ + ȳ = x + y for x, y ∈ Z. m=1
The order of Z/(n) is n as follows at once from the fact that for any
Using the results of §4, we see that for Re s > 1,
a ∈ Z, there is a unique b with 0 ≤ b < n for which a − b is divisible by
n. −1
Y
L(s, χ) = (1 − χ(p)p−s ) .
p
Example 1.5 Let Q∗ (R∗ , C∗ ) denote the set of non-zero rational (real,
complex) numbers. With the ‘usual’ multiplication for composition law, Further, because of the orthogonality relations (Proposition 1.10, Chap-
these from abelian groups. ter 1), we have
D
In what follows, we shall often drop the adjective abelian and speak
X X
χ(m) = χ(ā) = 0 if χ 6= χ0 .
simply of groups where we mean abelian groups. m=1 a∈G
86 CHAPTER 3. QUADRATIC FIELDS 1.4. ABELIAN GROUPS AND HOMOMORPHISMS 7

The existence of s0 and the uniform convergence of the series in the half- Definition 1.4 Let G, G′ be two groups. A homomorphisms f from G
plane Re s ≥ s0 + δ follows at once. As for P
the, derivative, if Re s > s0 to G′ is a map f : G → G′ such that f (xy) = f (x)f (y) for every x, y ∈ G.
and if s = σ + it, let s0 < σ1 < σ. We have am m−σ1 < ∞. Hence Let f : G → G′ be a homomorphism. Then f (e) = e. In fact, f (e) =
f (ee) = f (e)f (e) and multiplying both sides by f (e)−1 , we get f (e) = e
X X (log m)k If f : G → G′ and g: G′ → G′′ are homomorphisms, then g ◦ f : G → G′′
|am (log m)−k m−s | ≤ am m−σ1 <∞
mσ−σ1 is also a homomorphism. For any group G, the identity map IG : G → G
is a homomorphism.
(log m)k
since for each k, → 0 as m → ∞. Since the k-th derivative of
mσ−σ1 Definition 1.5 A homomorphism f : G → G′ is called an isomorphism
m−s is (−1)k (log m)k m−s the result follows. if there exists a homomorphism g: G′ → G such that f ◦ g = IG′ (the
identity map of G′ ) and g ◦ f = IG (the identity map of G).
Definition 3.8 The real number s0 defined by Lemma 3.3 is called the It is easy to see that a homomorphism f : G → G′ is an isomorphism
am m−s ; if the series converges
P
abscissa of convergence of the series if and only if it is both injective and surjective.
for all, we set s0 = −∞.
Example 1.6 The natural map η: Z → Z/(n) is a surjective homomor-
Lemma 3.4 Let {am } be a sequence of non-negative P numbers, and let phism. For n 6= 0 it is not one-one and hence not an isomorphism.
s0 be the
P abscissa of convergence of the series am m−s . Then, the
series am m−s = f (s) defines a holomorphic function Re s > s0 which Example 1.7 The map f : Z → Z given by f (a) = 2a for a ∈ Z is an
is singular at the point s = s0 . injective homomorphism. It is not onto and hence not an isomorphism.

Proof: That f (s)is holomorphic in Re s > s0 follows at once from Example 1.8 The map g: Q∗ → Q∗ given by g(x) = 1/x for x ∈ Q∗ is
Lemma 3.3. Suppose f is not singular at s = s0 . Then there exists a an isomorphism. Further g ◦ g = IQ∗ .
disc D: {|s − s1 | < δ} where s1 > s0 , such that |s0 − s1 | < δ and a
Definition 1.6 Let G be a group. A non-empty subset H of G is called
holomorphic function g in D such that g(s) = f (s) for Re s > s0 , s ∈ D.
a subgroup of G if for every x, y in H, xy −1 also belongs to H.
We have, by Taylor’s formula
∞ ∞ In particular e ∈ H and for any x ∈ H, x−1 ∈ H. It can be easily
X g(k) (s1 X f (k)(s1 )
g(s) = (s − s1 )k = (s1 − s)k checked that with the ‘composition’ induced by that of G, H is a group
k! k!
k=0 k=0 with e as the identity element and x−1 as the inverse of x in H.
Let H be a subgroup of G. Then the inclusion map j: H → G is an
(since g(s) = f (s) for s in a neighbourhood of s1 so that the series injective homomorphism.
P∞ (−1)k f (k) (s1 )
k=0 k! (s1 − s)k converges absolutely for any s in D and, in For any group G, the subsets {e} and G are subgroups of G. Let
particular, for real s with s1 − δ < s < s1 . Now, G1 , G2 be two groups and f : G1 → G2 , homomorphism of G1 into G2 .
∞
Then the set of x ∈ G1 , for which f (x) = e is easily verified to be a
subgroup of G1 . It is called the kernel of f and is denoted by ker f . The
X
(−1)k f (k) (s1 ) = am (log m)k m−s1
m=1 homomorphism f is an isomorphism if f is onto and ker f = {e}.
Let G be a group and H a subgroup of G. The relation “x ∼ y (x, y ∈
so that the repeated series G) if and only if xy −1 ∈ H” is an equivalence relation. If x̄ is the
∞ ∞
equivalence class containing x, then, on the set of equivalence classes we
X (s1 − s)k X can introduce the structure of a group by setting x̄ȳ = xy. These classes
am (log m)k m−s1 < ∞ for s1 − δ < s < s1 .
k! m=1 are called cosets of G modulo H. This group is denoted by G/H and is
k=0
8 CHAPTER 1. PRELIMINARIES 3.6. PRIMES IN AN ARITHMETIC PROGRESSION 85

called the quotient of G by H. If G/H is finite, then its order is called where F is a finite set ofQ(rational) primes, and p runs over all (rational)
−1
the index of H in G. There is natural mapping of G onto G/H taking primes. The product p(1 − N (p)−s ) would thus converge for all
x ∈ G to x̄ and this mapping is a surjective homomorphism with kernel s > 12 Since for s > 1, this product = ζK (s), we could conclude that
H. ζK (s) is bounded as s → 1 + 0. But we know that this is not the case,
Let f : G1 → G2 be a homomorphism of a group G1 into a group G2 since lims→1+0 (s − 1)ζK (s) 6= 0 by Proposition 3.13 .
with ker f = H. The image f (G1 ) of G1 under f is a subgroup of G2 and Suppose now that the degree of p is 1 for all but finitely many p. We
we define a homomorphism f¯: G1 /H → f (G1 ) by setting f (x̄) = f (x) for would have, for s > 1,
x ∈ G1 . Clearly f is an isomorphism of G1 /H onto f (G1 ) (fundamental
−1 −2 −1
Y Y Y Y
theorem of homomorphisms). ζK (s) = (1 − N (p)−s ) = (1 − p−s ) × (1 − N (p)−s ) ,
Let G be a group and a ∈ G such that every element of G is of the p p>p0 p≤p0 p⊃p
form an , where n ∈ Z. Then we say that G is a cyclic group generated
by a. where p0 is large enough. (If p0 ≥ |d|, then p splits or stays prime in
K for p > p0 , so that there would be exactly two prime divisors p of
Example 1.9 (Z, +) is an infinite cyclic group generated by the integer pOQexcept for finitely many p.) By (3.20) we have clearly lims→1+0 (s −
−1
1. The only subgroups of Z are of the form mZ = {mx | x ∈ Z} for 1) p>p0 (1 − p−s ) = c0 6= 0. This would give lims→1+0 (s−1)2 ζK (s) =
−1
m ≥ 0. c20 · p≤p0 p⊃p (1 − N (p)−1 ) 6= 0. This contradicts Proposition 3.13.
Q Q

Proposition 1.1 Any cyclic group G is isomorphic to Z or Z/(m) for Remark 3.9 Proposition 3.14 asserts simply the existence of infinitely
some (m > 0). many primes p of which a given discriminant d is (is not) a quadratic
residue. This would of course follow from Dirichlet’s theorem on the
Proof: Let G be generated by a. Consider the map f : Z → G taking existence of infinitely many primes in an arithmetic progression (to be
n ∈ Z to an . This is a surjective homomorphism and ker f is a subgroup proved in the following section).
mZ of Z generated by m ≥ 0 in Z. If m = 0, G is isomorphic to Z. If
m ≥ 0, G is isomorphic to Z/(m).
3.6 Primes in an arithmetic progression
Definition 1.7 Let G be a group and a ∈ G. We say that a is of order
n if the cyclic group generated by a in G is of order n. Lemma 3.3 Let {am } be a sequence of non-negative numbers. Suppose
that there is a real s such that ∞ −s is convergent. Then there
P
P∞am m −s
m=1
Example 1.10 The element −1 in Q∗ is of order 2. exists s0 ∈ R such that f (s) = m=1 am m converges for s > s0 ,
diverges for s < s0 (unless the series converges for all values of s).
Remark 1.6 If G is a group of order h and an element a ∈ G is of Further, the series converges for any complex s with Re s > s0 uniformly
order n, then n divides h and therefore ah = e. in any half plane Re sP≥ s0 + δ, with δ > 0. Also, for any integer k, we
have f (k) (s) = (−1)k ∞ k −s for Re s > s , where f (k) (s)
m=1 am (log m) m 0
th
denotes the k derivative of f (s).
1.5 Rings, modules and vector spaces.
′
am m−s < ∞ for some s′ ∈ R. We claim that
P
Proof: Suppose
Definition 1.8 Let R be a nonempty set and let φ, ψ be mappings of
|am m | < ∞ for Re s > s′ , and the convergence is uniform in this
−s
P
R × R into R. Writing φ((x, y)) = x + y, ψ((x, y)) = xy, the triple
region. In fact, since am ≥ 0
(R, φ, ψ) is said to be a ring if the following conditions are satisfied:
′
X
(i) (R, φ) is an abelian group, |am m−s | = am m−Re s ≤ am m−s
84 CHAPTER 3. QUADRATIC FIELDS 1.5. RINGS, MODULES AND VECTOR SPACES. 9

Proof: We have (ii) x(yz) = (xy)z for x, y, z in R (associativity).

b b
(iii) x(y + z) = xy + xz, (y + z)x = yx + zx in R (distributivity), and
Am (m−s − (m + 1)−s )
X X
am m−s = Ab b−s − Aa a−s +
m=a m=a+1 (iv) there exists in R an element 1, called the unit element of R, such
−s R m+1 s −δ−1 that x1 = 1x = x for every x in R.
For σ ≥ δ,we have |m−s− (m + 1) | = |s m u−s−1 du| ≤ σm .
Hence, if |Am ≤ M for all m > 0 in Z, then Remark 1.7 We call φ and ψ respectively the addition and the multipli-
b b cation in R; we write + for φ and · or × for ψ.
X  n |s| X o
am m−s  ≤ M b−δ + a−δ + m−δ−1 → 0
 
σ

a m=a+1 Remark 1.8 The identity element of (R, φ) is called the zero element
of R and is denoted by 0.
as a, b → ∞ uniformly in any bounded subset of σ ≥ δ. This proves the
proposition. Remark 1.9 The unit element in R is unique.
Using the fact that (s − 1)ζ(s) → 1 as s → 1 + 0 we obtain, by
Propositions 3.13, 3.14 and Theorem 3.3, the following. Remark 1.10 Associativity is valid for any finite number of elements
in R (in a sense which is obvious).
Theorem 3.4 (Dirichlet) Let K be a quadratic field of discriminant d.
Let h be the class number of K. Then we have
Remark 1.11 We denote by R∗ the set of non-zero elements of R.
 √
d
Ld (1) if d > 0

Definition 1.9 A ring R is commutative if xy = yz for every x, y in

2 log η

h= √ (3.24) R.
 w d Ld (1) if d > 0.


2π
Hereafter, by a ring we shall always mean a commutative ring.
Note that (3.24) implies that Ld (1) > 0. It is possible to express
the series Ld (1) as a finite (elementary) sum. Moreover, the fact that Definition 1.10 A subring S of a ring R is a subgroup of (R, +) such
Ld (1) > 0 is one of the key propositions in the proof of Dirichlet’s that 1 ∈ S and for x, y ∈ S, xy ∈ S.
theorem that for any integer l with (l, d) = 1, there exist infinitely many
primes p ≡ l( mod d); this theorem will be proved in the next section. We observe that S is a ring under the operations induced from R.
If R is a ring, S a subring and E a subset of R, the ring S[E]
Definition 3.7 The degree of prime ideal p is the integer f for which generated by
PnE is the set consisting of 1 and all the elements of the
N (p) = pf where p is a rational prime. form α = i=1 si , ei , si ∈ S where each ei is a product of finitely
many elements of E. It is trivial that S[E] is the smallest subring of R
Proposition 3.16 Let K be a quadratic field of discriminant d. Then containing S and E.
there exist in K infinitely many prime ideals of degree 1, and infinitely
many prime ideals of degree 2. Definition 1.11 Let R be a ring and a, b ∈ R. Then we say that a
divides b (or that a is a divisor of b, a|b in symbols), if there exists c in
p′
Proof: Clearly, for any prime ideal p, we have N (p) = with f = 1
R such that b = ac. If a does not divide b, we write a 6 | b.
or 2. Suppose that f > 1 for all but finitely many p. Then for s > 1,
Y −1
Y Y −1 Definition 1.12 If a, b, c ∈ R, then a is congruent to b modulo c ( in
(1 − N (p)−s ) ≤ (1 − q −s )−2 (1 − p−2s )
p p
symbols a ≡ b (mod c)) if c|(a − b).
q∈F
10 CHAPTER 1. PRELIMINARIES 3.5. THE DIRICHLET CLASS-NUMBER FORMULA 83

Definition 1.13 An element a ∈ R is called a zero-divisor if there exists  −1

d
x 6= 0 in R such that ax = 0. Trivially, 0 is a zero-divisor. Y
(1 − (N (p))−s )−1 = (1 − p−s )
−1
1− p−s
p⊃p
p
Definition 1.14 A commutative ring R in which 1 6= 0, is an integral
domain if R contains no zero-divisors other than 0. In fact, by Propositions 3.2 and 3.3 ( pd )
= 0 for p | d and there is a
unique p ⊃ pO with p2 = pO and N (p) = p. If ( pd ) = −1, there is
Definition 1.15 An element u ∈ R is called a unit in R if there exists a unique p ⊃ pO; indeed p = pO, so that N (p) = p2 and then 1 −
v ∈ R such that uv = 1. (R is commutative). (N (p))−s = (1 − p−2s = (1 − p−s )(1 + p−s ). If ( dp ) = 1, there are two
Q prime ideals p with p ⊃ pO and N (p) = p for each of them, so
distinct
The units u in R clearly form a multiplicative group which we denote that p⊃p (1 − N (p)−s )−1 = (1 − p−s )2 . We see therefore that
by U .
d −s −1
   
−1
Y
ζK (s) = (1 − p−s ) 1− p
Definition 1.16 We call two elements a, b in R associates with respect p
p
to U if a = ub for some u ∈ U. We say, briefly, that a and b are
associates and otherwise we say that a and b are non-associate. The Euler product for ζK (s) applied to the field K = Q gives us
∞
Definition 1.17 A commutative ring R is called a field, if the set R∗ −1
X Y
ζ(s) = m−s = (1 − p−s ) .
of elements a 6= 0 in R forms a group under multiplication (i.e. if every m=1 p
a 6= 0 in R is a unit ). Clearly a field is an integral domain.
The remark after the proof of the Euler product for ζK (s) applied to
d
Definition 1.18 A subring R of a field K is called a subfield of K, if am = ( m )m−s gives us
R is a field (with respect to the operations which make R a ring viz. the ∞   −1
X d  −s Y d
operations induced from K). Ld (s) = m = 1− p−s .
m=1
m p
p
Example 1.11 Z, Q, R, and C are commutative rings with the usual
Hence we have
addition and multiplication of complex numbers, and, in fact, integral
domains. C, R, Q are fields and Q, R are subfields of C; however, Z is Proposition 3.14 For s > 1, we have
not a field.
ζK (s) = ζ(s)Ld (s).
Example 1.12 The additive group Z/(m) consisting of residue classes
(of Z) modulo (an integer) m(6= 0) is a ring, the multiplication being We assert that the series defining Ld (s) converges for s > 0. In fact,
defined by r̄s̄ = rs for r, s ∈ Z. after Proposition 3.11, this follows from

Proposition 3.15 If {am } is a sequence of complex numbers such that

Example 1.13 A finite integral domain A is a field. In fact, let A =
Am = m
P
{x1 , x2 , . . . , xh }, and let a ∈ A, a 6= 0. Then the elements ax1 , . . . , axh k=1 ak is bounded as m → ∞ , then the series

are distinct, since A has no non-zero zero-divisors. Hence they must be ∞

X am
all the xi in some order, so that axi = 1 for some i. f (s) =
ms
1

Definition 1.19 A positive integer p in Z is called a prime if p > 1 converges for σ = Re s > 0, and uniformly on any bounded subset of the
and its only divisors in Z are ±1, ±p. half plane σ ≥ δ, for fixed δ > 0.
82 CHAPTER 3. QUADRATIC FIELDS 1.6. THE LEGENDRE SYMBOL 11

(The last equation holds by Lemma 3.2.) If d > 0, we set u1 = ξ1 β1 + Proposition 1.2 For p > 0 in Z, Z/(p) is a field if and only if p is a
√
ξ2 β2 , u2 = ξ2 β1′ + ξ2 β2′ then, since |β1 β2′ − β2 β1′ | = N (b) d, we have prime.
4
ZZ ZZ
dξ1 dξ2 = √ du1 du2 , Proof: In fact, if p = rs, r, s 6= ±1, then rs̄ = 0̄. But neither r̄ nor
Ω N (b) d U∗ s̄ is equal to 0̄. Hence Z/(p) is not even an integral domain, if p is not
where a prime. Conversely, if p is a prime, then for any x ∈ Z with p 6 | x,
there exists y ∈ Z such that xy ≡ 1(mod p). For, consider the set of
n u1 o
U ∗ = (u1 , u2 ) | 0 < u1 u2 < 1; 1 < < η 2 , u1 , u2 > 0 . r ∈ Z such that rx ≡ 0(mod p). This is an additive subgroup G of Z
u2 and hence, by the example 1.9 on page 8, of the form mZ, m ≥ 0. Since
Making the change of variables v1 = u1 u2 , v2 = u1 /u2 we see that p ∈ G, m = 1 or p. But m 6= 1, since p ∤ x. Thus m = p and as a
RR 4 log η consequence, the elements,
√
Ω dξ1 dξ2 = N (b) d so that, with (3.23) this gives us Theorem 3.3
when d > 0. If d < 0, we set u1 = Re (ξ1 β1 +ξ2 β2 ), u2 = Im (ξ1 β1 +ξ2 β2 ) 0, x, 2x, . . . , (p − 1)x.
and find that
are all distinct modulo p. Since the order of Z/(p) is p (see Example 1.4
2 2π
ZZ ZZ
dξ1 dξ2 = √ du1 du2 = √ page 6) there exists y ∈ Z (1 ≤ y ≤ p − 1) such that yx = 1̄, i.e.
Ω N (b) d u21 +u22 <1 n(b) d yx ≡ 1( mod p). [The last part of the argument is that of Example 1.13
above].
and Theorem 3.3 is completely proved.
Let K be, as above, a quadratic field of discriminant d, and, for
X > 0, N (X, K) the number of integral ideals a of norm N (a) < X. 1.6 The Legendre symbol
Since the number κ in Theorem 3.3 is independent of the class C, we
conclude that Definition 1.20 Let p ∈ Z, p > 2 be a prime. An integer a with p 6 |a
limX→∞ N (X;K)
X = h · κ, h being the class number is said to be a quadratic residue modulo p, if there exists x ∈ Z such
Further, if am denotes the number of integral that x2 ≡ a( mod p), and a quadratic non-residue modulo p if no such x
P ideals of norm =
m, then N (X; K) = m<X am , while ζK (S) = ∞ am exists in Z.
P
m=1 m2 . Hence, by
Lemma 3.1, we obtain
From the definition, it is clear that a is a quadratic residue modulo
Proposition 3.13 (Dedekind) We have p if and only if a + mp (for arbitrary m ∈ Z) is so. We can thus talk of a
(non-zero) residue class modulo p being quadratic residue or non-residue
lim (s − 1)ζK (S) = h · κ, modulo p.
s→1+0
Consider now, for p ∤ a, the quadratic congruence x2 ≡ a( mod p) for
where h is the class number of the quadratic field K, and κ is the number x ∈ Z. If a is a quadratic non-residue modulo p, this congruence has no
defined in Theorem 3.3. solution x. If a is a quadratic residue modulo p, say a ≡ b2 ( mod p), then
x2 ≡ b2 ( mod p), i.e. (x− b)(x+ b) ≡ 0( mod p). Since by Proposition 1.2
We shall evaluate the above limit in another way. We have already Z/(p) is an integral domain, it follows that x ≡ ±b( mod p) are the
proved that only two solutions of the congruence x2 ≡ a( mod p). Further, since
−1
Y
ζK (s) = (1 − (N (p))−s ) . p > 2, b 6≡ −b( mod p). Now consider the mapping x̄ → x̄2 of (Z/(p))∗
p
into itself. The image of x̄ (6= 0) under this mapping is always a quadratic
Let p be any (rational) prime number; then there are at most two residue modulo p and by what we have seen, each quadratic residue
prime ideals p, p′ dividing p. We claim that the product modulo p in (Z/(p))∗ is the image of exactly two elements of (Z/(p))∗
12 CHAPTER 1. PRELIMINARIES 3.5. THE DIRICHLET CLASS-NUMBER FORMULA 81

under this mapping. Thus there are (p − 1)/2 quadratic residues modulo Conversely, if ω1 , ω2 are associate elements of b satisfying (3.21),
p. It follows that there are (p − 1)/2 quadratic non-residues modulo p. then ω1 = ǫω2 , where ǫ is a unit with 1 ≤ |ǫ| < η, so that ǫ = ±1. Hence
The product of two quadratic residues modulo p is again a residue. 
 the number of ω ∈ b with 
For, if a ≡ x2 ( mod p) and b ≡ y 2 ( mod p) then ab ≡ (xy)2 ( mod p).

2N (X, C) = 0 < |NK (ω)| < Y, 0 ≤ log 
 ω  < log η (3.22)

The product of a quadratic residue a modulo p and a non-residue b   1
|NK (ω)| 2

modulo p is a quadratic non-residue modulo p. For, there exists x ∈ Z
for which x2 ≡ a( mod p) and if ab ≡ z 2 ( mod p) for z ∈ Z choose by Case (ii) d < 0. Clearly we have now, wN (X, C) = the number of
Proposition 1.2, y ∈ Z such that xy ≡ 1( mod p). Clearly we have then integers ω ∈ b with 0 < |NK (ω)| < Y.
the congruence b ≡ (yz)2 ( mod p), and b would be a residue modulo p. In either case, let (β1 , β2 ) be an integral base of b and let β1′ , β2′ be
The product of two quadratic non-residues a, b modulo p is a quadratic the conjugates of β1 , β2 respectively. Let Ω denote the following open
residue modulo p. For let ā1 , . . . , āq (q = 21 (p − 1)), be the quadratic set in the plane: if d > 0,
residues modulo p. Then since Z/(p) is an integral domain, it follows

from what we have seen above that aa1 , . . . , aaq are precisely all the Ω = ζ = (ξ1 , ξ2 ) ∈ R2 | 0 < |ξ1 β1 + ξ2 β2 ||ξ1 β1′ + ξ2 β2′ | < 1,
residue classes modulo p which are non-residues. Since ab is distinct 
|ξ1 β1 + ξ2 β2 | ′ ′ 12
from these, it must be a residue. 0 < log 1 |ξ 1 β1 + ξ 2 β2 | < log η ,
|ξ1 β1 + ξ2 β2 | 2
Definition 1.21 Let p 6= 2 be a prime and a ∈ Z. We define the and if d < 0,
Legendre symbol ( ap ) by
Ω = {ζ = (ξ1 , ξ2 ) ∈ R2 | 0 < |ξ1 β1 + ξ2 β2 |2 < 1}.

   +1 if p ∤ a and ais a quadratic residue modulo p We verify that Ω is bounded as follows. For d > 0, since
a
= −1 if p ∤ a and a is a non-residue modulo p,
p |ξ1 β1 + ξ2 β2 ||ξ1 β1′ + ξ2 β2′ | < 1
0 if p | a.


and
It is clear from the earlier considerations that 1 ≤ |ξ1 β1 + ξ2 β2 |/|ξ1 β1′ + ξ2 β2′ | < η 2 ,
we see that both ξ1 β1 + ξ2 β2 , ξ1 β1′ + ξ2 β2′ are bounded in Ω. Thus ξ1 , ξ2
    
ab a b
= for a, b ∈ Z (1.1) √
p p p are again bounded in Ω, since β1 β2′ − β2 β1′ 6= 0 (in fact = ±N (b) d by
′ ′
(3.2)). If d < 0, then |ξ1 β1 + ξ2 β2 | = |ξ1 β1 + ξ2 β2 | < 1, and again, since
1.7 The quotient field of an integral domain β1 β2′ − β2 β1′ 6= 0, ξ1 , ξ2 are bounded in Ω. According to what we have
proved above, we have
Definition 1.22 Let R, R′ be two rings. A map f : R → R′ is called a 
 number of lattice points in Ω√Y ifd < 0
homomorphism if  number of lattice points in Ω√ + the number AY

Y
wN (X, C) =
of lattice points (ξ1 , ξ2 ) with |ξ1 β1 + ξ2 β2 |2 ≤ Y
(i) f (x + y) = f (x) + f (y),


and |ξ1 β1 + ξ2 β2 | = |ξ1 β1′ + ξ2 β2′ | =

6 0 if d < 0.
(ii) f (xy) = f (x)f (y) for every x, y ∈ R, and √ √
Since, as is easily verified. AY = O( Y ) = O( X), we conclude that
√
(iii) f (1) = 1. wN (X, C) NΩ ( Y )
lim = N (b) lim
X→∞ X Y →∞ Y
For any ring R, the identity map IR is a homomorphism. The composite
Z Z
of two homomorphisms is again a homomorphism. = N (b) dξ1 dξ2 (3.23)
Ω
80 CHAPTER 3. QUADRATIC FIELDS 1.7. THE QUOTIENT FIELD OF AN INTEGRAL DOMAIN 13

Lemma 3.2 Let Ω be a bounded open set in the plane R2 . For X > 0, For any ring R, the identity map IR is a homomorphism. The com-
let ΩX = {ζ = (ξ1 , ξ2 ) ∈ R2 | ( ξX1 , ξX2 ) ∈ Ω}. Let NΩ (X) denote the posite of two homomorphisms is again a homomorphism.
number of lattice points in ΩX . Then
ZZ Definition 1.23 A homomorphism f : R → R′ is said to be an isomor-
lim X −2 NΩ (X) = dξ1 dξ2 = area of Ω phism if there exists a homomorphism g: R′ → R such that g ◦ f = IR
X→∞ Ω and f ◦ g = IR′ . The rings R and R′ are then said to be isomorphic and
provided that this integral exists in the sense of Riemann. we write R ≃ R′ .

Proof: Divide the plane into closed squares S of sides X1 parallel to An isomorphism f : R → R is called an automorphism of R. The
the coordinate axes. For any S, let P (S) denote the point whose coor- image f (R) of a ring R under a homomorphism f : R → R′ is a subring
dinates have smallest values (“the lower left vertex”). Clearly NΩ (X) = of R′ . A homomorphism is an isomorphism if and only if it is injective
{number of S with P (S) ∈ Ω}. and surjective.
Now, if N1 , N2 denote, respectively, the number of S with S ⊂ Ω,
S ∩ Ω 6= ∅,R Rthen, by the definition of the Riemann integral, X −2 N1 , Remark 1.12 The natural map q: Z → Z/(m) is a homomorphism.
X −2 N2 → Ω dξ1 dξ2 ; since N1 ≤ NΩ (X) ≤ N2 , the lemma follows.
Proposition 1.3 Every integral domain R can be embedded isomorphi-
Theorem 3.3 (Dedekind.) Let K be a quadratic field of discriminant d cally in a field.
and w the number of roots of unity in K. Let C be an ideal class of K and
N (X, C) the number of non-zero integral ideals a ∈ C with N (a) < X. Proof: Let R be an integral domain and R∗ the set of non-zero ele-
Then ments of R. On R × R∗ , we define the relation: (a, b) ∼ (c, d) if ad = bc.
N (X, C) Since R contains no zero-divisors, it can be verified that this is an equiv-
lim =κ
X→∞ X alence relation. We make the quotient K = R × R∗ / ∼ a ring by defin-
exists and we have ing the ring operations as follows. If x/y denotes the equivalence class
 containing (x, y) ∈ R × R∗ then define a/b + c/d = (ad + bc)/bd and
 2 log η (a/b)(c/d) = ac/bd. These operations are well defined and K is a ring.
√ if d > 0, η > 1 being the fundamental unit;
κ= d √ In fact, K is a field, since b/a is an inverse for a/b, a 6= 0. The map
2π/(w |d| if d < 0.

i: R → K given only by i(a) = a/1 for a ∈ R is a one-one homomorphism
of R into K.
Proof: Let b be an integral ideal in C −1 then, for any integral ideal
We shall identify R with the subring i(R) of K.
a ∈ C, ab = αO where α ∈ b. Conversely, if α ∈ b, a = b−1 αO is an
integral ideal in C; moreover, |NK (α)| = N (a)N (b) so that N (a) < X
Remark 1.13 K is called the quotient field of R. If f : R → L is a one-
if and only if |NK (α)| < XN (b) = Y (say). Consequently N (X, C) is
one homomorphism of R into a field L, then f can be extended in a
the number of non-zero principal ideals αO, α ∈ b, |NK (α)| < Y ; in
unique way to a one-one homomorphism f¯ of K into L, by prescribing
other words, N (X, C) = the number of α ∈ b, α 6= 0, which are pairwise
f¯(a/b) = f (a)f (b)−1 , for b 6= 0. Further, this property characterises the
non-associates and for which |NK (α)| < Y.
quotient field upto isomorphism. Thus L contains the isomorphic image
Case (i) d > 0. Let η > 1 be the fundamental unit. Clearly for any
f¯(K) of K. We see then that if L contains an isomorphic image of K
α ∈ b, α 6= 0, there is an integer m such that if ω = η m α, we have
and in this sense, K is the “smallest” field containing R.
 
 ω 
0 ≤ log  1  < log η.
 (3.21) Example 1.14 Q is (isomorphic to) the quotient field of Z.
|NK (ω| 2
14 CHAPTER 1. PRELIMINARIES 3.5. THE DIRICHLET CLASS-NUMBER FORMULA 79

1.8 Modules Since (s − 1)ζ(s) → 1 as s → 1 + 0, we obtain

Let R be a (commutative) ring (containing 1) lim sup (s − 1)f (s) − cζ(s)| ≤ ǫ.

s→+0
Definition 1.24 An R-module M is a triple (M, +, ψ) where
Since this is true for any ǫ > 0, we obtain
(i) (M, +) is an (abelian) group
lim (s − 1)f (s) = c lim (s − 1)ζ(s) = c.
(ii) ψ: R × M → M is a map such that if we set s→1+0 s→1+0

ψ((λ, a)) = λa, then for λ, µ ∈ R, and x, y ∈ M, we have As an application of these remarks, we prove

(1) λ(x + y) = λx + λy, Proposition 3.12 ζ(s) is meromorphic in Re s > 0, its only singularity
in the half-plane Re s > 0 is at s = 1 where it has a simple pole with
(2) (λ + µ)x = λx + µx residue 1.
(3) (λµ)x = λ(µx),
Proof: We have, for s > 1,
(4) 1 · x = x.
∞
X ∞
X Z m+1
The elements of R are called scalars and ψ is called scalar multiplication. ζ(s) = m{m−s − (m + 1)s } = s m u−s−1 du
m=1 m=1 m
Definition 1.25 If R is a field, M is called a vector space over R (or ∞ Z m+1 ∞
[u]
X Z
an R-vector space) and the elements of R called vectors. = [u]u−s−1 du = s du
m=1 m 1 us+1
Example 1.15 Every (abelian) group (G, +) can be regarded as a Z-
module (G, +, ψ) by defining ψ((n, x)) = nx for x ∈ G and n ∈ Z. where [u] is the largest integer ≤ u. Now [u] = u−(u), where 0 ≤ (u) < 1.
Conversely, every Z-module is of this type. Hence, for s > 1,
∞ ∞ ∞
(u) s (u)
Z Z Z
Example 1.16 Every (commutative) ring R with 1 is an R-module. ζ(s) = s u−s du − s du = −s du.
1 1 us+1 s−1 1 us+1
Example 1.17 R and C are vector spaces over Q.
Now, for Re ≥ δ > 0, |(u)u−s−1 | < u−δ−1 , so that the latter integral
Definition 1.26 Let M be an R-module (resp. vector space over R). converges uniformly for Re s ≥ δ > o for any δ > 0, and so defines a
Then a subgroup N of M is called an R-submodule (resp. a subspace s
holomorphic function g(s) for Re s > 0. Since, for s > 1, ζ(s) − s−1 =
over R) if ψ maps R × N into N. −sg(s) the proposition follows.

Definition 1.27 Let M be an R-module. A subset S of M is said to

Remark 3.8 If we set A′ (X) =
P
m≤X am , then (3.18) holds if and
generate M over R if every x ∈ M can be written in the form x = −1 ′
Pn + only if X A (X) → c as X → ∞. In fact, either of these conditions
i=1 ai xi with xi ∈ s, ai ∈ R and n = n(x) ∈ Z . implies that m−1 am → 0 as m → ∞, and they are therefore equivalent.
The elements of S are called generators of M over R. In particular,
if there exists a finite subset S of M over R, then M is finitely generated Definition 3.6 By a lattice point in the plane R2 we mean a point
over R. If S = ∅, by definition, the module generated by S is {0}. ζ = (ξ1 , ξ2 ) with ξ1 , ξ2 ∈ Z.
78 CHAPTER 3. QUADRATIC FIELDS 1.8. MODULES 15

Proof: Since A(1) = 0, we have, for M > 0 in Z. Proposition 1.4 Let α1 , α2 , . . . , αn be a system of generators of a Z-
module M , and let N be a sub-module of M . Then there exist P β1 , . . . , βm
M M
X X in N (m ≤ n) that generate N over Z and have the form βi = j≤i kij αj
am m−s = {A(m + 1) − A(m)}m−s
with kij ∈ Z, kii ≥ 0 and 1 ≤ i ≤ m.
m=1 m=1
M
X −1
= A(M + 1)M −s + A(m + 1){m−s − (m + 1)−s }. Proof: Let us suppose that the proposition has been proved for all
m=1 Z-modules with n − 1 generators at most, where n ≥ 1. (Note that the
(3.19) proposition is trivial when M = {0}.) Let M be a module generated
Now
P∞ −s = ζ(s) converges for s > 1, and as M → ∞, over Z by n elements α1 , α2 , . . . , αn and N a sub-module of M . Define
m=1 m
A(M + 1)M −s → 0, for s > 1. We see, on applying this to the sequence M ′ to be the module generated by α2 , α3 , . . . , αn over Z and N ′ to be
{am = 1}, that N ∩ M ′ . If n = 1, M ′ = {0}.) If N = N ′ the proposition is true by
the induction hypothesis. If N 6= N ′ , then let A be the subgroup of
∞
X Z consisting of integers k for which there exist k2 , . . . , kn in Z with
m{m−s − (m + 1)−s } = ζ(s). kα1 + k2 α2 + · · · + kn αn ∈ N. Then A is of the formP k11 Z, k11 ≥ 0;
m=1 n
let β1 = k11 α1 + k12 α2 + · · · + k1n αn ∈ N. If α = i=1 hi αi , with
Now M −s − (m + 1)−s = s
R m+1
x−s−1 dx, and 0 ≤ x − m ≤ 1 in the h1 , h2 , . . . , hn ∈ Z, belongs to N , then h2 ∈ A so that h1 = mk11 for
m
interval (m, m + 1). Hence some m ∈ P Z. Thus α − mβ1 ∈ N ′ . By the induction hypothesis, there
exist βi = j≤i kij αj (i = 2, 3, . . . , m, kij ∈ Z, kii ≥ 0), in N ′ , which
∞ m+1 ∞ generate N ′ . It is clear that β1 , β2 , . . . , βm are generators of N having
 X Z  Z
ζ(s) − s x−s dx < s x−s−1 dx = 1,
 
m 1
the required form.
m=1

i.e. ∞ Definition 1.28 A subset S of an R-module M is linearly independent

s 
 Z  
ζ(s) − s x−s dx = ζ(s) −  < 1; over R, if, for any finite set of elements x1 , . . . , xn in S, a relation
  
1 s−1 P n
i=1 ai xi = 0, ai ∈ R implies necessarily that a1 = · · · = an = 0. We
in particular say that S is linearly dependent, if it is not linearly independent.
lim (s − 1)ζ(s) = 1. (3.20)
s→1+0
Definition 1.29 A subset S of M is called a base of M over R (or an
Since, by (3.18), |A(m + 1)| ≤ K1 m for some Ki > 0, and since
∞ −s R-base) if S is linearly independent and generates M over R.
1)−s } converges (absolutely) for s > 1,(3.19) im-
P
m=1 m{mP − (m +−s
plies that ∞m=1 a m m = f (s) converges for s > 1; moreover, we have
Example 1.18 The Z-module of even integers has {2} for a base.
∞
X
−s −s
|f (s) − cζ(s)| = |A(m + 1) − cm|{m − (m + 1) }
Example 1.19 Q is a Z-module but is not finitely generated over Z.
m=1
Trivially, Q if finitely generated over Q.
Given ǫ > 0, there exists, by (3.18), an m0 = m0 (ǫ) with
|A(m + 1) − cm| < ǫm for m > m0 . Then, for s > 1, Example 1.20 R2 = R × R has {(1, 0), (0, 1)} as a base over R.
m0
X X
|f (s) − cζ(s)| ≤ |A(m + 1) − cm| + ǫ m{m−s − (m + 1)−s } Proposition 1.5 Let V be a vector space (over a field K) which has a
m=1 m>m0 set of generators consisting of m elements. If S is any linearly indepen-
≤ C(m0 ) + ǫζ(s). dent subset of V , then S contains at most m elements.
16 CHAPTER 1. PRELIMINARIES 3.5. THE DIRICHLET CLASS-NUMBER FORMULA 77

Proof: Let x1 , . . . , xm be a set of generators for V over K. If possible, and then

let S P contain more than m elements, say y1 , . . . , ym+1 . We can write g g
y1 = m
Y Y
i=1 λi xi , λi ∈ K and λ 6= 0 for at least one i. WeP may assume, pn = N (pO) = N (pi )ei = pfi ei ≥ pg (since fi ≥ 1, ei ≥ 1),
m
without loss of generality, that λ1 6= 0. Then x1 = λ−1 1 y1 −
−1
i=2 λ1 λi xi . i=1 i=1
Hence y1 , x2 , . . . , xm generate V . Assume for some i, 1 ≤ i ≤ m, that
so that g ≤ n. Hence (3.15) gives
(y1 , . . . , yi , xi+1 , . . . , xm ) generate V . (This is true for i = 1.) Then
−n
X Y
i m
(N (a))−s ≤ (1 − p−s ) ;
p≤x
X X
N (a)≤x
yi+1 = αj yj + αk xk , α1 , . . . , αm ∈ K.
j=1 k=i+1 Q −s −1
P product (1 − p ) is absolutely convergent for s > 1, the
since the
Since y1 , . . . , yi+1 are linearly independent, αk 6= 0 for some k ≥ i + 1. series (N (a))−s converges for s > 1. If we now let x → ∞ in (3.17) we
Without loss of generality, we may suppose that αi+1 6= 0. Then obtain the Euler product for ζK (S), viz.
−1
Y
i+1
X m
X ζK (s) = (1 − (N (p))−s ) .
xi+1 = βj yj + βk xk p
j=1 k=i+2
Note that this equation also holds when s is complex and Re s > 1.
for β1 , . . . , βi+1 , βi+2 , . . . , βm ∈ K. Thus y1 , . . . , yi+1 , xi+2 , . . . , xm gen-
Remark 3.7 The same reasoning shows that if {am } is a sequence of
erate V . By iteration of this P process, it follows that y1 , . . . , ym generate
V . In particular ym+1 = m P numbers with a1 = 1, amk = am ak for all integers m, k ≥ 1,
complex
i=1 γi yi , γi ∈ K, contradicting the linear and if ∞m=1 |am | < ∞,then
independence of S.
∞
X Y
Corollary 1.1 If x1 , . . . , xm and y1 , . . . , yn are bases of V , then m = n. am = (1 − ap )−1 .
m=1 p

Remark 1.14 We shall be concerned only with vector spaces, which are In particular, for Re s > 1, we have
finitely generated over a field.
∞
−1
X Y
ζ(s) = m−s = (1 − p−s ) .
Remark 1.15 Let V be a vector space, finitely generated over a field K.
m=1 p
From the finitely many generators, we can clearly pick out a maximal
set of linearly independent elements which suffice to generate V and Lemma 3.1 Let {amP} be a sequence of real numbers, X a real positive
constitute a base of V over K. number. Let A(X) = m<X am . Suppose that

Remark 1.16 Let V be a vector space generated by x1 , . . . , xm over K A(X)

lim = c. (3.18)
and let x 6= 0 in V . Then x, x1 , . . . , xm again generate V over K and, X→∞ X
from this set, we can always pick out a maximal linearly independent Then the series
∞
subset containing x. In other words, any x 6= 0 in V can be completed X am
f (s) =
to a base of V . ms
m=1

Definition 1.30 A vector space V is said to be of dimension n over a converges for s > 1 and we have
field K (in symbols dimk V = n) if there exists a base of V over K lim (s − 1)f (s) = c.
containing n elements. s→1+0
76 CHAPTER 3. QUADRATIC FIELDS 1.8. MODULES 17

Let C0 = 1, C1 , . . . , Ch−1 denote the different ideal classes. For each Observe that, by Corollary 1.1, the dimension is defined indepen-
class C, we define the zeta-function of C, denoted ζK (s, C), by dently of the base used.
X′ We shall write merely dim V for dimk V when it is clear from the
ζK (s, C) = (N (a))−1 ; context to which field K we are referring.
a∈C
Corollary 1.2 Let W be a subspace of V with dimk V = n. Then W
the summation is over all non-zero integral ideals a in C and s is a real has a base consisting of at most n elements, i.e. dimk W ≤ dimk V. If
number > 1. The zeta-function ζK (s) of the field K is defined by W is a proper subspace of V , then dim W < dim V.
X X′
ζK (s) = ζK (s, C) = (N (a))−s For, we know first, by Proposition 1.5, that any linearly independent
C∈h a set in W contains at most n elements. Choose a maximal set of linearly
independent elements in W . This is a base for W and therefore dim W ≤
the summation now being over all nonzero integral ideals of K.
n = dim V. Since any n linearly independent elements of V generate V ,
We assert now that all these series converge (absolutely) for s > 1.
it follows that dim W < dim V if W is properly contained in V .
It is, of course, sufficient to verify this for the series defining ζK (s). Let
x > 0 be any real number. We have Example 1.21 Let K be a field and k a subfield of K. Let L be a
1 field of which K is a subfield. We may clearly consider L as a K-
−1
X Y
≤ (1 − (N (p))−s ) (3.15) (or k-) vector space, K as a k-vector space. Suppose the vector spaces
N (a))s
N (a)≤x N (p)≤x L/K, K/k are of finite dimension, and that u1 , . . . , un ∈ K form a
k-base of K, v1 , . . . , vm ∈ L form a K-base of L. Then L is a finite
the product being over all prime ideal p with N (p) ≤ x. To prove this,
dimensional k-vector space, and the products ui vj , 1 ≤ i ≤ n, 1 ≤ j ≤
we remark that
m(in L) form a k-base of L.
−1
(1 − (N (p))−s ) = 1 + (N (p))−s + (N (p))−2s + · · · (3.16) Definition 1.31 Let V, V ′ be two vector space over a field K. By a
homomorphism (or K-linear map) of V into V ′ we mean a homomor-
and that any integral ideal a can be written uniquely as a product of
phism f of (V, +) into (V ′ , +) which satisfies in addition the condition
prime ideals; further if N (a) ≤ x, then every prime divisor p of a satisfies
N (p) ≤ x. Inequality (3.15) follows on multiplying out the finitely many f (λx) = λf (x) for x ∈ V, λ ∈ K.
absolutely convergent series (3.16) with N (p) ≤ x; moreover, we have If the homomorphism f is one-one and onto, then f is an isomor-
phism and V, V ′ are isomorphic.
−1
Y X X
(1 − (N (p))−s ) − (N (a))−s = (N (a))−s (3.17)
By an endomorphism of a vector space V , we mean a homomorphism
N (p)≤x N (a)≤x N (a)>x
of V into itself.
where the latter summation is over the integral ideals a of norm > x all Let f : V → V ′ be K-linear. Let N be the kernel of f . Then N is a
of whose prime divisors are of norm ≤ x. subspace of V and, similarly, the image f (V ) is a subspace of V ′ . The
Any prime ideal p contains a unique prime number p ∈ Z, by Re- quotient group V /N can be made into a vector space over K. Further
mark 2.18. We have N (p) = pf for a certain integer f ≥ 1, so that V /N is isomorphic of f (V ) (Noether homomorphism theorem). If dimk V
p ≤ x if N (p) ≤ x. Further, there are at most n distinct prime ide- is finite, then dimk V = dimk f (V ) + dimK N. Let V, V ′ be vector spaces
als p1 , . . . , pg , g ≤ n containing a given p; in fact, they are uniquely over K, and let (e1 , . . . , en ) form a base of V over K. Then given
determined by the equation arbitrary elements a1 , . . . , an ∈ V ′ here is a unique
P K-linear map f : V →
. In fact, for x ∈ V, if x = ni=1 λi ei , then we have
V ′ with f (ei ) = aiP
e
pO = pe11 · · · pgg only to set f (x) = ni=1 λi ai . The uniqueness is obvious.
18 CHAPTER 1. PRELIMINARIES 3.5. THE DIRICHLET CLASS-NUMBER FORMULA 75

Definition 1.32 Let V be a vector space over a field K and dimk V = Proof: CASE 1. d is odd. If d is odd, then, for any odd n > 0 we
n. By a K-linear form (or briefly, a linear form) on V , we mean a have, by Proposition 3.9, (case (ii) of the proof)
homomorphism of V into K (regarded as a vector space over itself ).      
d n n
= = .
n d |d|
The set V ∗ of linear forms on V forms a vector space over K and
is called the dual of V . If α1 , . . . , αn is a base of V over K, define Let |d| = pa, where p is an odd prime. We have p ∤ a a since d is square-
the linear forms α∗1 , . . . , α∗n by α∗i (αj ) = δij (the Kronecker delta)= 1 free. Let u be a quadratic non-residue modulo p. By our remark above,
if i = j and 0 for i 6= j. We at once that α∗1 , . . . , α∗n are linearly there is an (odd) n > 0 such that n ≡ u( mod p), n ≡ 1( mod 2a). Then
Psee
n ∗
independent over K. For, if i=1 i αi = 0 with a1 , . . . , a)n = K, then
a        
d n n u 1
aj = ni=1 ai α∗i (αj ) = 0 forPj = 1, . . . , n. Any linear form α∗ can be
P = = = −1.
n
n p a p a
written in the form α∗ = ∗ ∗
i=1 b1 αi where α (αi ) = bi ∈ K. Thus
dimK V ∗ = n. The base α1 ∗, . . . , α∗n ofV ∗ is called the dual base of the CASE 2. d is even. Let d = d1 d2 , where d1 is an even discriminant,
base α1 , . . . , αn of V. and d2 is an odd discriminant. Then, for n > 0 in Z, we have ( nd ) =
( dn1 )( dn2 ) (by definition). Again, by definition, it is easy to check that
there exists a with ( da1 ) = −1. Choose n > 0 such that n ≡ a( mod d1 )
Definition 1.33 Let V be a vector space over a field K. A bilinear
and n ≡ 1( mod d2 ). We then have ( nd ) = −1.
form B on V is a mapping B: V × V → K such that for any fixed
y ∈ V the mappings By′ , By′′ of V into K, defined by By′ (x) = B(x, y) Proposition 3.11 Let d be a discriminant and Sm = m
P d
n=1 ( n ). Then
and By′′ (x) = B(y, x) respectively, are linear forms on V . 1
|Sm | ≤ 2 |d|.

Definition 1.34 A bilinear form B(x, y) on V is non-degenerate if, for Proof: We first prove the following: let a1 , . . . , ar , r = |d|, denote a
any fixed y 6= 0 in V , the linear form By′ , is not zero, i.e. B(x, y) 6= 0 complete system of residues modulo r, i.e. a system of integers which
for at least one x, and for any fixed x 6= 0, the linear form Bx′′ is not are congruentP to the integers 0, 1, . . . , r − 1 modulo r (in some order).
zero. Then S = ri=1 ( adi ) = 0. In fact, let n be a positive integer with (n, d) =
1, ( nd ) = −1; then the numbers na1 , . . . , nar also form a complete system
Let V be a vector space of dimension n over a field K. Then we have of residues modulo r. Now ( db ) = ( dc ) if b ≡ c( mod r). We have therefore
r   r
X d X d
S= =− ( ) = −S, so that S = 0.
Proposition 1.6 Let B(x, y) be a non-degenerate bilinear form on V. nai ai
i=1 i=1
Then for any base α1 , . . . , αn of V , there exists a base β1 , . . . , βn of V
such that B(αi , βj ) = δij (the Kronecker delta) for 1 ≤ i, j ≤ n. Given m > 0, let k be a positive integer for which |m − kr| is minimal.
Then we have |m − kr| ≤ 12 r; Hence |Sm − Skr | ≤ 21 ; but Skr is the
sum of k terms ri=1 ( adi ) where a1 , . . . , ar runs over a complete residue
P
Proof: Consider the mapping of V to V ∗ taking y ∈ V to the linear
form By′ in V ∗ . This is clearly a homomorphism of V into V ∗ . Since B is system modulo r, and is hence zero. Thus |Sm | ≤ 12 r.
non-degenerate, this mapping is injective. Since dim V = dim V ∗ = n,
it follows by Noether’s homomorphism theorem and Corollary 1.2 above 3.5 The Dirichlet class-number formula
that this mapping is onto V ∗ . Let α1 , . . . , αn be a base of V over K
and α∗1 , . . . , α∗n the corresponding dual base of V ∗ . Let β1 , . . . , βn be Let K be an algebraic number field of degree n. The group h of ideal
the elements of V which are mapped into α∗1 , . . . , α∗n respectively by the classes of K is a finite group of order h = h(K). We shall obtain, in this
homomorphism above. Then B(αi , βj ) = α∗j (αi ) = δij . section, a formula for h in the case when K is a quadratic field.
74 CHAPTER 3. QUADRATIC FIELDS 1.9. IDEALS AND QUOTIENT RINGS 19

Since 4 | d, the first factors coincide for m and n. The same is true 1.9 Ideals and quotient rings
also of the other two factors, in the case n ≡ m( mod d). But if n ≡
−m( mod d), they differ exactly by the factor sgn d′ = sgn d. Let R be a (commutative) ring (with 1). Then R can be regarded as a
(ii) Let a = 0. Consequently, d ≡ 1( mod 4). Then module over itself.

  
d d
  b    b  
d d 2 d Definition 1.35 By an ideal of R, we mean an R-submodule of R.
= b ′
= · =
n 2n 2 n′ d n′
Clearly an ideal I of R, is a subgroup of (R, +) such that for any
′ x ∈ I and a ∈ R, we have ax ∈ I.
since ( d2 ) = ( nd ) for d ≡ 1( mod 4). Further, by Proposition 3.8,
Example 1.22 R and {0} are ideals of a ring R.
n′
     ′
d ′ n
= (−1)(d−1)(n −1)/4 =
n′ d d Example 1.23 Subgroups of (Z, +) are ideals of the ring Z with the
usual addition and multiplication. Any ideal of Z is clearly of the form
since n′ is odd and d ≡ 1( mod 4). Thus ( nd ) = ( nd ). Further ( −1 d ) =
d
mZ, m ∈ Z.
sgn d. Therefore ( m ) = ( nd ) for m, n > 0 and m ≡ n( mod d) and
( nd ) = ( nd ) = ( −m m d
d ) = sgn d · ( d ) = ( m ) · sgn d if n ≡ −m( mod d). Example 1.24 Let R, R′ be two rings and f : R → R′ be a homomor-
phism. Then ker f is an ideal of R.
Remark 3.5 Thus, for positive integers n, ( nd ) represents a so-called
“residue class character” modulo d, i.e. ( m d
)( nd ) for m, n > 0, m ≡ An integral domain R 6= {0} is a field if and only if R and {0} are
n( mod d) and ( mnd
) = (md
)( nd ) for m, n > 0. In particular, if p1 , p2 are the only ideals of R, as can be easily proved.
two primes satisfying p1 ≡ p2 ( mod d) and pi ∤ d, then either both p1
√ Definition 1.36 If a, b are two ideals P of R, the product ab of a and b
and p2 split or both stay prime in Q( d).
is the set of all finite sums of the form i ai bi with ai ∈ a and bi ∈ b.
In what follows, a discriminant will stand for an integer d 6= 1 which
It is easy to check that ab is again an ideal of R. Clearly ab = ba.
is the discriminant of a quadratic field; in other words, it will denote
either a square-free integer d 6= 1 with d ≡ 1( mod 4) or d = 4d′ where Definition 1.37 An ideal a of a ring R divides an ideal b of R, if a ⊃ b.
d′ is square-free and d′ ≡ 2 or 3( mod 4). Whenever n, m > 0, n ≡
m( mod d), we have ( nd ) = ( m
d
). Definition 1.38 By a proper ideal of a ring R, we mean an ideal of R
different from R and {0}.
Remark 3.6 (1) Any discriminant d can be written in the form d =
d1 d2 , where d1 is 1 or a discriminant, d2 is an odd discriminant and the Definition 1.39 Let S be a subset of a ring R. An ideal a of R is
only prime divisor of d1 if 2 if |d1 | > 1. generated by S, if it is generated by S as an R-module. We say a is
(2) If a, b are integers with (a, b) = 1, and α, β are any integers, finitely generated, if it is finitely generated as an R-module.
there exists n > 0 with n ≡ α( mod a), n ≡ β( mod b); in fact, if
x, y are integers with xa ≡ β( mod b), yb ≡ α( mod a), we may take Definition 1.40 If an ideal a of a ring R is generated by a single el-
n = xa + yb + k|ab|, where k is a large integer. ement α ∈ R, then a is called a principal ideal of R. (We denote it by
αR in this case or just by (α).)
Proposition 3.10 If d is a discriminant, there exists n > 0, n ∈ Z
with ( nd ) = −1. Example 1.25 R and {0} are principal ideals of R.
20 CHAPTER 1. PRELIMINARIES 3.4. LAWS OF QUADRATIC RECIPROCITY 73

Definition 1.41 An integral domain R all of whose ideals are principal We now define the Kronecker symbol ( na ) for any integer a ≡ 0 or
ideals is called a principal ideal domain. 1( mod 4) as follows. First we define

Example 1.26 Z is a principal ideal domain (see Example 1.23, p. 19).      0 if a ≡ 0( mod 4)
a a
= = 1 if a ≡ 1( mod 8)
2 −2
Let a be an ideal of a ring R. The additive group of R/a (in words, −1 if a ≡ 5( mod 8)

R modulo a) is a ring called the the quotient ring of R by a, with
multiplication defined by (x + a)(y + a) = xy + a for x, y ∈ R. (Since a This agrees with our definition of ( d2 ) for the discriminant d of a
is an ideal of R, this multiplication is well defined.) The natural map quadratic field on page 63. By Proposition 3.8, ( a2 ) = ( a2 ), whenever
′ ′
q: r → R/a is a surjective homomorphism with kernel a. a ≡ 1( mod 4) Further, clearly ( a2 ) = ( a2 ) for a ≡ a′ ( mod 8) and ( aa2 ) =
′
( a2 )( a2 ). In general, if a ≡ 0 or 1( mod 4), we introduce the Kronecker
Example 1.27 Z/(m) is the quotient ring of Z by the ideal (m). This symbol ( na ) for arbitrary denominator n by setting ( 2ac ) = ( a2 )c and
ring is called the ring of residue classes modulo m. (See Example 1.12, ( na ) = ( na1 ) · ( 2ac ) where n = n1 · 2c , c ≥ 0 and n1 is odd. By the very
page 10.) It is a field if and only if m is a prime, by Proposition 1.2. definition, it is clear that ( xy a
) = ( xa )( ay ) for x, y ∈ Z.
d
For the discriminant d of a quadratic field, we see that ( xy ) = ( xd )( dy )
Let f : R → R′ be a homomorphism of a ring R onto a ring R′ and let
a = ker f. The homomorphism f induces a homomorphism f¯: R/a → R′ , for x, y ∈ Z. i.e. ( zd ) is multiplicative in z. We now prove another
by setting f¯(x + a) = f (x) for x ∈ R. Clearly f is an isomorphism of interesting property of ( nd ).
rings. (Fundamental theorem of homomorphisms for rings.)
Proposition 3.9 If d is the discriminant of a quadratic field and m, n
Remark 1.17 Let K be a field. Consider the ring homomorphism are positive integers, then
f : Z → K given by f (n) = n · 1(= 1 + · · · + 1, n times). By the funda-    
d d
mental theorem of homomorphisms referred to above, Z/ ker f ≃ f (Z). = for n ≡ m( mod d)
We have ker f = (p) for some p ≥ 0. in Z. We call p the characteristic of n m
K. Observe that p is a prime, if p > 0. For, if p = rs, 1 < r, s < p then    
f (r)f (s) = f (rs) = 0. But neither f (r) nor f (s) is zero, contradicting d d
= sgn d for n ≡ −m( mod d)
the fact that K is an integral domain. If p = 0, then K contains f (Z) n m
which is isomorphic to Z and hence contains a subfield isomorphic to Q
(see remark on 13). Thus every field contains a subfield isomorphic to Proof: Let d = 2a · d′ , n = 2b · n′ , m = 2c · m′ with odd d′ , n′ , m′ and
either Q or Z/(p) (for a prime p). The fields Q and Z/(p)(p prime) are a, b, c ≥ 0 in Z.
called prime fields. (i) Let a > 0. The case b > 0 is trivial, for then, by assumption,
c > 0, and both the symbols in this proposition are zero, by definition.
Definition 1.42 A proper ideal p of an integral domain R is called a Let then b = c = 0. By Proposition 3.8
prime ideal if, for a, b ∈ R, ab ∈ p implies that either a or b is in p.    a ′   a  ′   
d 2 ·d 2 d 2 n ′
= = = (−1)a(n −1)/8 ′ (−1)(n−1)(d −1)/4
Example 1.28 In Z a prime p generates a prime ideal and conversely n n n n d
every prime ideal of Z is generated by a prime p.
and similarly
Remark 1.18 A proper ideal p is a prime ideal of a ring R if and only    a ′
d 2 d 2
 
m ′
if R/p is an integral domain. = = (−1)a(m −1)/8 ′ (−1)(m−1)(d −1)/4
m m d
72 CHAPTER 3. QUADRATIC FIELDS 1.10. LINEAR MAPPINGS AND MATRICES 21

in view of (3.12). Finally, by Theorem 3.2 Remark 1.19 If a prime ideal p of a ring R divides the product of two
  Y  pi   Y  qj  ideals a and b, then p divides either a or b. For, if p divides neither a
P P P
= = (−1) i (pi −1)/2 j (qj −1)/2 nor b there exist a ∈ a, b ∈ b while ab ∈ ab ⊂ p which is a contradiction.
Q qj pi
1≤i≤r 1≤i≤r
1≤j≤s 1≤j≤s Definition 1.43 A proper ideal a of R is maximal if a is not contained
in any other proper ideal of R.
so that, by (3.11), we have
   
P Q Remark 1.20 An ideal a of a ring R is a maximal ideal, if and only if
= (−1)((P −1)/2((Q−1)/2 for odd P, Q > 0. (3.13) R/a is a field.
Q P
If P is not necessarily positive, Remark 1.21 A maximal ideal is clearly prime.
   
−1 −1
= = (−1)(|P |−1)/2= = (−1)(P −1)/2+( sgn P −1)/2
1.10 Linear mappings and matrices
P |P |
since P = |P | · sgn P and, by (3.9), Let V be a vector space of dimension n over a field K. Let φ be a
linear mapping (i.e. a homomorphism) P of V into V. Taking a fixed base
P −1 |P | − 1 sgn P − 1
≡ + ( mod 2) (3.14) e1 , . . . , en of V over K, let φ(ej ) = ni=1 aij ei (j = 1, . . . , n) with aij ∈ K.
2 2 2 To the linear mapping φ, we associate the ordered set
Hence, for odd P, Q, we have  
       a11 · · · a1n
P P sgn P |P |  ··· 
= =  
Q |Q| Q Q  ··· 
a1 n · · · ann
 
|P |
= (−1)(( sgn P −1)/2)(( sgn Q−1)/2+(( sgn P −1)/2(( sgn Q−1)/2))
Q
of the n2 elements a11 , a12 , . . . , ann , which will be referred to as the
Further, by (3.11) corresponding ‘matrix’(aij ). The elements apq (p = 1, 2, . . . , n) are said
      to constitute the pth ‘row’ of (aij and the elements apq (p = 1, 2, . . . , n)
|P | |P | |Q| constitute the qth ‘column’ of (aij . The matrix (aij ) has thus n rows and
= = (−1)((|P |−1)/2)((|Q|−1)/2)
Q |Q| |P | n columns and is called an n-rowed square matrix (or an n × n matrix)
 
|Q| with elements in K.
= (−1)((|P |−1)/2·((|Q|−1)/2)
P Conversely, given an n-rowed square matrix (aij ) with elements in
K, we can
P find a unique linear mapping φ of V into itself for which
 
Q
= (−1)( sgn Q−1)(P −1)/4+( sgn Q−1)( sgn P −1)/4+(|P |−1)(|Q|−1)/4 φ(ej ) = ni=1 aij ei . Thus, once a base of V is chosen, the linear mappings
P
(by (3.13)) of V into itself stand in one-one correspondence with the set Mn (K) of
n-rowed square matrices with elements in K. In Mn (K),we can introduce
 
Q
= (−1)(( sgn Q−1)/2·(|P |−1)/2+(|P |−1)/2(|Q|−1)/2 (by (3.12)) the structure of a ring as follows. If A = (aij ), B = (bij ) are in Mn (K)
P
define A + B to be the n-rowed square matrix (cij ) with cij ) = aij + bij
and the product AB to be the element (dij ) of Mn (K) with
Using (3.13) again, we have
    n
X
P Q dij = aip bpj .
= (−1)(P −1)(Q−1)/4+( sgn P −1)( sgn Q−1)/4
Q P p=1
22 CHAPTER 1. PRELIMINARIES 3.4. LAWS OF QUADRATIC RECIPROCITY 71

The usual laws of addition and multiplication for a ring can be verified Proposition 3.8 For odd integers P, Q, we have
to be true. However, this ring is not, in general, commutative. To the  
identity mapping IV : V → V corresponds the matrix I = In = (δij ) and −1
= (−1)(P −1)/2+( sgn P −1)/2
this serves as the unit element in the ring Mn (K). Note that if φ, ψ are P
 
two linear mappings of V into itself, and if A, B are the corresponding 2 2
= (−1)(P −1)/8
matrices, then φ + ψ corresponds to A + B and φ ◦ ψ to A · B. For P
Pn= (aij ) ∈ Mn (K), we define the trace Tr (A) of A to be the sum
A
   
P Q
aii . Clearly for A = (aij ), B = (bij ) in Mn (K), we have Tr (AB) = = (−1)(P −1)(Q−1)/4+( sgn P −1)( sgn Q−1)/4
Pi=1 Q P
n
i,j=1 aij bji = Tr (BA) and further Tr (I) = n. For A ∈ Mn (K), we x
denote by det A, the determinant of A. For A = I, det I = 1. where, for real x 6= 0, sgn x = |x| .
We shall not define the determinant. It has the following properties
which are the only ones we shall use. (See e.g. [2],) Proof: For odd a, b ∈ Z, we have (a − 1)(b − 1) ≡ 0( mod 4) i.e.
Let A = (aij ), 1 ≤ i, j ≤ n. Then ab − 1 ≡ a − 1 + b − 1( mod 4) i.e.

(a) det A is a homogeneous polynomial in the elements aij of degree ab − 1 a−1 b−1
≡ + ( mod 2). (3.9)
n; 2 2 2

(b) det A is linear considered as a function of any two of (aij ); it is Similarly, for a, b odd (a2 − 1)(b2 − 1) ≡ 0( mod 16) and therefore
linear also in the columns of (aij );
a2 b2 − 1 a2 − 1 b2 − 1
≡ + ( mod 2) (3.10)
(c) if A = (aij ) and At = (bij ) where bij = aji ,then det At = det A; 8 8 8

(d) if A = (aij ) and aij = 0 for i > j, then det A = a11 . . . ann ; in By iteration of (3.9) and (3.10) we see that for any r odd numbers
particular det I = 1; p1 , p2 , . . . , pr we have
r
(e) for any two n × n matrices A and B, we have det(AB) = det A · p1 p2 · · · pr − 1 X pi − 1
≡ ( mod 2) (3.11)
det B. 2 2
i=1

Let f1 , . . . , fn constitute another base of V over K, and let and

r
(p1 p2 · · · pr )2 − 1 X p2i − 1
n n ≡ ( mod 2) (3.12)
8 8
X X
fj = pij ei , ej = qij fi , j = 1, . . . , n, pij , qij ∈ K. i=1
i−1 i=1
Let us first suppose that P, Q > 0 and let P = p1 p2 · · · pr , Q =
Then to the linear mapping φ of V into V, taking ej to fj corresponds q1 q2 · · · qs , where p1 , p2 , . . . , pr , q1 , q2 , . . . , qs are odd primes. By the
the matrix P = (pij ). Let ψ be the linear mapping of V taking fj to ej . definition of ( ap ) and by Proposition 3.5, we have
Denote by Q the corresponding matrix (qij ). The composite mappings      
φ ◦ ψ and ψ ◦ φ are clearly both equal to the identity mapping of V into −1 −1 −1 Pr
i=1 (pi −1)/2
= ··· = (−1) = (−1)(P −1)/2
itself. The corresponding matrices are P Q and QP respectively. Hence P p1 pr
P Q = QP = In and Q is the inverse P −1 of P in Mn (K). in view of (3.11) and
Suppose now, that instead of the base e1 , . . . P
, en of V , we had referred
fj = ni=1 pij ei (j = 1, . . . , n).
everything to another base f1 , . . . , fn withP Pr p2i −1
 
2 2

Then the linear mapping ψ taking ej to ni=1 aij ei (j = 1, . . . , n) would = (−1) i=1 8 = (−1)(p −1)/8
P
70 CHAPTER 3. QUADRATIC FIELDS 1.11. POLYNOMIAL RINGS 23

take fj to ni=1 ( nk,l=1 qil alk pkj )fi and the corresponding matrix would
P P
the Law of Quadratic Reciprocity. We may suppose that p 6= q, since,
otherwise, the theorem is trivial. be QAP = P AP ∈ Mn (K). Now Tr (P −1 AP ) = Tr (AP P −1 ) =
−1

(i) Let p ≡ 1( mod 4). Taking r = p, s = q, we have Tr (A). Consequently, if φ is an endomorphism of V , and A the matrix
 
p
 
q corresponding to it with respect to a base of V, then Tr (A) is indepen-
=1⇒ = 1. dent of the base chosen, and we set Tr (φ) = Tr (A) and speak of the
q p
trace of the endomorphism. Similarly, since
Similarly, if q ≡ 1( mod 4), we have, by the symmetry between p and q.
    det(P −1 AP ) = det P −1 · det A · det P = det A · det P −1 · det P = det A
q p
=1⇒ = 1.
p q we may define det φ to be det A where A ∈ Mn (K) is a matrix corre-
sponding to φ with respect to a base of V . For φ = IV , det φ = 1.
Thus, if p ≡ q ≡ 1( mod 4), then ( pq ) = ( pq ).
(ii) Let p ≡ 1( mod 4), q ≡ 3( mod 4). By (i) we have first Remark 1.22 A one-one linear mapping φ of V into V is necessarily
onto V. For, dim φ(V ) = dim V by the homomorphism theorem. Since
   
p q
=1⇒ = 1. φ(V ) is a subspace of V of the same dimension as V, we have φ(V ) =
q p
V i.e. φ is onto V.
Conversely, let ( pq ) = 1. Then ( −1 p ) = 1 by Proposition 3.5, and therefore
( −q q −1
p ) = ( p )( p ) = 1. Taking r = −q, s = p in the foregoing, we have
Remark 1.23 A linear mapping φ of V into V is one-one if and only
( pq ) = ( |r|p
) = 1. Thus we have shown that if p ≡ 1( mod 4) and q ≡ if det φ 6= 0.
3( mod 4),then ( pq ) = ( pq ). Again, by the symmetry between p and q, it
Proof: If φ is one-one, then φ is onto V by Remark 1.22 above.
follows that for p ≡ 3( mod 4) and q ≡ 1( mod 4) · ( pq ) = ( pq ). Clearly there exists a linear mapping ψ of V into V such that φ◦ψ = IV .
(iii) Let p ≡ q ≡ 3( mod 4). Then, by√Proposition 3.7, either p Since det(φ ◦ ψ) = det φ · det ψ = det IV = 1, it follows that det φ 6= 0.
or q is the norm of an algebraic integer x+y 2 (pq) ≻ 0. Without loss of Conversely, if det φ 6= 0, let, if possible, φ(e1 ) = 0 for e1 6= 0. But, by
generality, let 4p = x2 − pqy 2 . This means that p | x, i.e. x = pu, u ∈ Z. Remark 16 on page 16, e1 can be completed to a base e1 , e2 , . . . , en of
Hence 4 = pu2 − qy 2 and −qy 2 ≡ 4( mod p). Now p ∤ y since p ∤ 4. V . For this base, the corresponding matrix A in Mn (K) has the form
Since Z/(p) if a field, −q is a quadratic residue modulo p i.e ( −q p ) = 1. (aij ) with ai1 = 0 for i = 1, . . . , n and therefore det φ = det A = 0,(since
Similarly, pu2 ≡ 4( mod q) gives us ( pq ) = 1. Since ( −1 p ) = −1 by det A is linear in the columns of A). We are thus led to a contradiction.
Proposition 3.5, we have ( pq ) = −( pq ) = (−1)(p−1)(q−1)/4 ( qp ). We now use
Legendre’s symbol to define the Jacobi symbol ( aq ) for odd composite 1.11 Polynomial rings
denominators q. For two integers a and n of which n is odd and positive,
we define the Jacobi symbol ( na ) by ( na ) = ( pa1 )r1 · · · ( pak )rk where n = Let R be a commutative ring (containing 1). Let M be the set of all
pr11 · · · prkk and p1 , . . . , pk are odd primes. Clearly, if any of p1 , p2 , . . . , pk mappings f : Z+ → R such that f (n) = 0 for all but finitely many n.
divides a, ( na ) = 0. Further it is easy to check that ( ab a b
n ) = ( n )( n ) for odd We introduce on M the structure of an R-module by defining, for
′
positive n and ( na ) = ( an ) if a ≡ a′ ( mod n). f, g ∈ M, a ∈ R, f + g, af by

Remark 3.4 For composite n, ( na ) = 1 does not, in general, ensure the (f + g)(n) = f (n) + g(n), (af )(n) = a · f (n), n ∈ Z+ .
existence of x ∈ Z for which x2 ≡ a( mod n).
We make M a ring by defining f · g by (f · g)(n) = ni=0 f (i)g(n − i).
P
For negative odd n, we define ( na ) to be the Jacobi symbol ( |n|
a
). For The map e for which e(0) = 1, e(n) = 0 for n > 0 is the unit of M. We
Jacobi symbols, we have the general laws of reciprocity given by have a map i: R → M defined by i(a)(0) = a, i(a)(n) = 0 for n > 0, i
24 CHAPTER 1. PRELIMINARIES 3.4. LAWS OF QUADRATIC RECIPROCITY 69

is an isomorphism of R onto a subring of M, so that we may identity R Proposition 3.6 An odd prime p is a sum of two squares of integers if
with i(R). Let X ∈ M denote the map for which X(1) = 1, X(n) = 0 and only if p ≡ 1( mod 4).
if n 6= 1. Then (with multiplication defined as above) X k is the map for
which X k (k) = 1, X k (n) = 0 if n 6= k. Hence any f ∈ M can be written Proof: By Proposition 3.5, ( −1 p ) = (−1)
(p−1)/2 . By the remark on
−1
uniquely in the form page 63 , ( p ) = 1 if and only if p is a sum of two squares of integers.
X Hence the proposition follows.
f= ak X k , ak ∈ R;
Proposition 3.7 Let K be a quadratic field of discriminant d = q1 q2
the sum is finite, i.e. ak = 0 for large k.
where q1 , q2 are distinct primes congruent to 3 modulo 4. Then, either
Definition 1.44 The ring M is denoted by R[X] and is called the poly- q1 or q2 is the norm of an element α ∈ K, α ≻ 0 (but not both).
nomial ring in one variable over R. The elements of R[X] are called
polynomials with coefficients in R or polynomials over R. Proof: Observe first that √ for any unit ǫ ∈ K, NK (ǫ) = 1. For if
x+y d
there exists α = with x, y ∈ Z and NK (α) = −1, then
Definition 1.45 If f = ni=1 ai X i ∈ R[X] and f 6= 0, we define the
P
2
degree n of f (in symbols deg f ) to be the largest integer i such that −4 ≡ x2 ( mod q1 q2 ), i.e. ( −1 −4
q1 ) = ( q1 ) = 1 which, by Proposition 3.5,
ai 6= 0. We call an the leading coefficient of f. If this an = 1, we say f contradicts our assumption that q1 ≡ 3( mod , 4).
is a monic polynomial. If f is of degree 1, we call f a linear polynomial. Now, if O is the ring of algebraic integers in K, then q1 O = q21 , q2 O =
If f = 0, we set deg f = −∞. q22 for prime ideals q1 q2 by Proposition 3.2. Then by Theorem 3.1, there
exist a1 , a2 which are equal to 0 or 1 and such that a1 + a2 > 0 and
Remark 1.24 If f, g ∈ R[X] we have qa11 qa22 ≈ O. If both a1 and a2 were equal to 1, then q1 q2 = αO with
√ √
deg(f + g) ≤ max(deg f, deg g). α ≻ 0. On the other hand dO = (q1 q2 )O = q1 q2 since d = q1 q2
2 2 √
and q1 q2 O = q1 q2 . Hence α = ǫ d for a unit ǫ. But then NK (ǫ) = −1,
If deg f 6= deg g, then deg(f + g) = max(deg f, deg g). whereas we have just proved that K contains no units of norm −1. Thus,
either a1 = 0 and a2 = 1, or a1 = 1 and a2 = 0. Then either q2 ≈ O or
Remark 1.25 If R is an integral domain and f, g ∈ K[X] with f 6=
q1 ≈ O. Since N (q1 ) = q1 , N (q2 ) = q2 we see that either q1 or q2 is the
0, g 6= 0, then f g 6= 0, i.e. R[X] is an integral domain. Further,
norm of α ∈ O with α ≻ 0.
deg(f g) = deg f + deg g.
We have now the necessary preliminaries for the proof of the cele-
Remark 1.26 Let K be a field and let f, g ∈ K[X] with deg g > 0. brated Law of Quadratic Reciprocity.
Then there exist h, j ∈ K[X] such that f = gh + j where deg j < deg g.
(Division algorithm in K[X].) Theorem 3.2 (Gauss.) For odd primes p and q.
   
Remark 1.27 Given any ideal a 6= {0} of the polynomial ring K[X] p q
= (−1)(p−1)(q−1)/4 .
over a field K, it is clear by Remark 1.26 that a is generated over K[X] q p
by a polynomial t in a of minimal positive degree. Thus K[X] is a
Proof: Let r be the discriminant of a quadratic field K such that ±r
principal ideal domain.
is an odd prime and let s be an odd prime different from |r|. Let ( rs ) = 1
′
Let R, R′ be two rings and φ: R → R′ be a homomorphism. Then Then, if O is the ring of algebraic integers in K, we
√ have sO = pp by
we can extend φ uniquely to a homomorphism φ of R[X] to R′ [X] by Proposition 3.2. Further, ph0 = αO where α = x+y2 r ≻ 0, and h0 is the
2 2
prescribing that φ(X) = X and, in general order of the restricted class group of K. Taking norms, sh0 = x −ry 4 .
X  X X But h0 is odd by Proposition 3.4. Hence 4s h0 and consequently s is a
φ ai X i = φ(ai )X i for ai X i ∈ R[X]. s

quadratic residue modulo |r| i.e. |r| = 1. We use this fact to prove
i i i
68 CHAPTER 3. QUADRATIC FIELDS 1.12. FACTORIAL RINGS 25

Proposition 3.4 If the discriminant d of a quadratic field K is divisible Let R be a ring with R ⊂ C and let R′ = C. For any α ∈ C we have a
only by one prime number then h0 is odd and so equal to the class number unique R-linear ring homomorphism P →i C such that ψ(X) = α;
ψ: R[X]
in fact we have only to set ψ( ai X i ) = aiP
P
h of K. In this case, if d > 0, then K contains a unit of norm −1. α , ai ∈ R. We denote the
image ofPR[X] under ψ by R[α] and for f = i ai X i ∈ R[X], we write
Proof: By Theorem 3.1, the restricted class containing O is the sole f (α) = i ai αi .
ambiguous class h0 i.e. h0 does not contain elements of order 2.
We shall now prove that ho is of odd order. For x ∈ h0 , S
let Ax denote Definition 1.46 A complex number α is a root of f ∈ R[X], if f ∈
the subset of h0 consisting of x and x−1 . Now, h0 = A1 ∪ x∈h 0 Ax and ker ψ, i.e. f (α) = 0.
x6=1
A1 ∩ Ax = ∅ if x 6= 1. Further, since no element of h0 is of order 2, Ax √
consists of 2 elements for every x 6= 1 in h0 . But A1 = {1}. Thus the Example 1.29 Take R = C. The complex numbers ± (−5) are roots
2
of the polynomial x + 5.
order h0 of h0 is odd.
By the remarks at the beginning of this section, we necessarily have
Remark 1.28 If R = K is a field and f ∈ K[X], then α ∈ K is a
h0 = h. It is clear that if d > 0, K contains a unit of norm −1.
root of f if and only if (X − α)|f. In fact, there is β ∈ K[X] of degree
We shall make use of the above results to deduce the well-known
0 (or − ∞), i.e. β ∈ K, such that f = q · (X − α) + β, q ∈ K[X]. Then
laws of quadratic reciprocity.
f (α) = β = 0.
Proposition 3.5 If p is an odd prime, then
    Definition 1.47 If R = K is a field, α ∈ K is called a repeated root of
−1 2 2
f ∈ K[X] if (X − α)2 |f.
(i) = (−1)(p−1)/2 , (ii) = (−1)(p −1)/8
p p
Definition 1.48 If f = i≥0 ai X i ∈ K[X], then the polynomial f ′ =
P
Proof: (i) If p ≡ 1( mod 4), then (−1)(p−1)/2 = 1. We shall prove √ i−1
√
P
that ( −1 a+b p i≥1 iai X is called the derivative of f.
p ) = 1 in this case. In fact, Q( p) contains a unit ǫ = ( 2 )
of norm −1, in view of Proposition 3.4. Hence a2 ≡ −4( mod p), i.e. Remark 1.29 It is easily seen that (f + g)′ = f ′ + g′ , (f g)′ = f g′ + gf ′
−1 2 2
1 = ( −4 −1 4 −1 2
p ) = ( p )( p ) = ( p )( p ) = ( p ). Conversely, if a ≡ −1( mod p) and if K has characteristic 0, that f ′ = 0 if and only if f ∈ K. When
where p ∤ a we obtain, by the remark on page 18 that 1 ≡ ap−1 ≡ K has characteristic p > 0, f ′ = 0 if and only if f ∈ K[X p ].
(a2 )(p−1)/2 ≡ (−1)(p−1)/2 ( mod p) so that p ≡ 1( mod 4).
(ii) Let, first ( 2p ) = 1. Since ( 8p ) = ( p4 )( 2p ) = ( 2p ) = 1. we see, by The quotient field of the polynomial ring K[X] over a field K is de-
√
Proposition 3.3, that p = pp′ in Q( 2). We have shown that h = 1 noted by K(X) and called the field of rational functions in one variable
√ √
for K = Q( 2) (page 63). Since 1 + 2 is a unit of norm −1 in over K.
√
Q( 2), h0 = h = 1. Hence p = x − 2y 2 for some x, y ∈ Z. If 2 | y,
2

then x2 ≡ p( mod 8), and if 2 ∤ y. p ≡ x2 − 2( mod 8). In any case,

1.12 Factorial rings
2 ∤ x so that p ≡ ±1( mod 8). Conversely, if p ≡ ±1( mod 8), consider
√
K = Q( q) where q = ±p and q ≡ 1( mod 8). The discriminant of Let R be an integral domain.
√
Q( q) is q. Since ( 2q ) = 1, we have 2O = pp′ for prime ideals p 6= p′ . By
Proposition 3.4, h0 is odd. Since ph0 = αO for α ∈ O with NK (α) > 0, Definition 1.49 An element a ∈ R∗ which is not a unit in R is called
we have, on taking norms, NK (α) = 2h0 , i.e. 2h0 +2 = x2 − qy 2 with irreducible, if, whenever a = bc for b, c ∈ R, either b or c is a unit in R.
x, y ∈ Z. Since h0 is odd, it follows that ( p2 ) = 1. Thus ( 2p ) = 1 if and
2
only if p ≡ ±1( mod 8) i.e. ( 2p ) = (−1)(p −1)/8 . Definition 1.50 An element p ∈ R∗ is said to be prime if it is not a
We deduce the well-known result due to Fermat, namely unit and whenever p divides ab, with a, b ∈ R, p divides either a or b.
26 CHAPTER 1. PRELIMINARIES 3.4. LAWS OF QUADRATIC RECIPROCITY 67

Example 1.30 A prime number p in Z is prime in Z (Proposition 1.2). 1+ω

ω = , so that if a = (1 + ω)−1 b, we have a = a′ . By what we
1 + ω′
Remark 1.30 A prime element is always irreducible but not conversely, have shown above, a = rOpt11 · · · ptkk where r ∈ Q, ti = 0 or 1. Further
(in general). In Z, every irreducible element is (upto sign) a prime NK (1 + ω) = ω(1 + ω ′ )2 > 0 if d > 0. Hence b ≈ pt11 · · · ptkk where ti = 0
number, by the very definition of a prime number. or 1. Hence any ambiguous ideal class is equivalent, in the restricted
sense to one of the at most 2t ideal classes containing pt11 · · · ptkk , ti = 0
Remark 1.31 In a principal ideal domain, an irreducible element gen- or 1.
erates a maximum ideal. To complete the proof of the theorem we have therefore only to prove
the following:
Definition 1.51 An integral domain R is a factorial ring (or a unique there is exactly one relation of the form
factorization domain) if every non-unit a ∈ R∗ can be written in the
form a = q1 · · · qr where q1 , . . . , qr are irreducible elements of R deter- t
X
mined uniquely by a upto multiplication by units of R and upto order. pa11 · · · pat t ≈ O where ai = 0 or 1, ai > 0. (3.6)
the decomposition a = q1 · · · qr is called the factorization of a (into irre- i=1
ducible elements).
We prove first the uniqueness of a relation of this form. Let pa11 · · · pa−t
t
Remark 1.32 In a factorial ring R, every irreducible elements is prime = ρO with NK (ρ) > 0. Clearly ρO = ρ′ O, so that
(and hence, by Remark 1.30 above, the notions of ‘prime’ and ‘irre-
ducible’ coincide in such a ring). For, if an irreducible element a divides ρ = ηρ′ where η is a unit. (3.7)
bc, then a must occur in the factorization of bc and hence in that of at
least one of b, c. Further, if d > 0, we have η ≻ 0. Let ǫ be a generator of the group
of the totally positive units in K if d > 0, and of the group of all units
Remark 1.33 Let a = p1 · · · pr be a factorization of an element a in K if d < 0; we have always ǫ 6= 1. Replacing ρ by ρǫn for a suitable
into irreducible elements. Clubbing together the irreducible elements n ∈ Z we may suppose that, in (3.7)
which are associated, we can write a in the form upn1 1 · · · pns s , where
n1 , . . . , ns ∈ Z, ni > 0, p1 , . . . , ps are irreducible elements in R which ρ = ηρ′ with η = 1 or ǫ. (3.8)
are non-associate, and u is a unit.
We claim that η 6= 1. In fact, if η = 1, then ρ ∈ Z, hence O =
Definition Q 1.52 Let a1 , . . . , ar be elements of the factorial ring R and ρ−1 pa11 · · · pat t = p01 · · · p0t , contradicting the uniqueness asserted by (3.5).
n ,i √
ai = ui sj=1 pj j with nj,i ∈ Z+ and irreducible p1 , . . . , ps (some of Hence η = ǫ,and we have ρ = ǫρ′ . Let µ = d(1−ǫ)(6= 0). Then µ = ǫµ′ ,
the nj,i may be zero). TheQgreatest common divisor (briefly g.c.d.) of ρ ′ ρ ρ
α
a1 , . . . , ar is defined to be sj=1 pj j where αj = min(nj,i , . . . , nj,r ), j = so that = , and = r ∈ Q. Hence µO = r −1 pa11 · · · pat t and the
µ µ µ
1, . . . , s. Similarly, the least common multiple of a1 , . . . , ar is defined to uniqueness assertion in (??) shows that this determines a1 , . . . , at . (Note
β √
be sj=1 pj j where, for j = 1, . . . , s, βj = max(nj,i , . . . , nj,r ). These are
Q
that µ = d(1 − ǫ) is uniquely determined by the field K.)
√
uniquely determined upto units. To prove the existence of a relation (3.6), define µ = d(1 − ǫ) as
above. Then µO = (µO)′ , so that µO = cpa11 · · · pat t , where P ∈ Q, ai = 0
c
Proposition 1.7 (Fundamental theorem of arithmetic.) Z is a factorial or 1. Since µ ≻ 0 if d > 0, we have only to show that ti=1 ai > 0. If
ring. this were not the case, then µ = cζ, where ζ is a unit and ζ ≻ 0 if d > 0.
cǫ
Since µ = ǫµ′ this gives cζ = cǫζ ′ = so that ǫ = ζ 2 , contradicting the
Proof: The only units in Z are +1 and −1. It suffices to prove that Pt ζ
every positive integer can be written uniquely as a product of positive definition of ǫ. Thus i=1 ai > 0, and the theorem is proved.
66 CHAPTER 3. QUADRATIC FIELDS 1.12. FACTORIAL RINGS 27

Definition 3.5 A class of ideals C in h0 = ∆/Π0 is said to be ambigu- irreducible elements of Z, in other words, of prime numbers (in view of
ous if C 2 = 1 in h0 . Remark 1.30 above). If two prime numbers are associated, they must
be the same.
We shall now find the number of ambiguous classes in h0 . We shall first prove that if a factorization of a > 0 in Z into primes
The following theorem is of great importance by itself, although it exists it is unique. Let, in fact, for a > 0 in Z, a = p1 · · · pr = q1 · · · qs
may look a little out of place in our scheme. It is connected with the where p1 , . . . , pr , q1 , . . . , qs are primes. Now q1 divides p1 · · · pr and
so-called ‘genus characters’ in Gauss’ theory of binary quadratic forms hence divides one of the pi , say p1 . Then q1 = p1 . Now p2 · · · pr =
and we are not able to go into this here. We have, however, used it to q2 · · · qs . By repeating the argument above, q2 is equal to one of p2 , . . . , pr ,
deduce the laws of quadratic reciprocity. say q2 = p2 . In finitely many steps, we can thus prove that r = s and
that q1 , . . . , qs coincide with p1 , . . . , pr order.
Theorem 3.1 The number of ambiguous ideal classes in a quadratic We now prove the existence of a factorization for any a > 0 in Z by
field K of discriminant d is 2t−1 where t is the number of distinct prime induction. For a = 2, this is trivial. Assume that a factorization into
numbers dividing d. primes exists for all positive integers less than a. Now if a is a prime,
we have nothing to prove. if a is not prime, then a is divisible by b ∈ Z
Proof: Let |d| = pα1 1 p2 · · · pt , where α1 = 1 or α1 = 2 or 3 according with 1 < b < a. In other words, a = bc with b, c ∈ Z and 1 < b, c < a.
as d is odd or even. Then, because of the remarks at the end of Chapter By the induction hypothesis, b and c have a factorization into primes
3,§1 pi O = p2i where pi is a prime ideal in O of norm pi ; further, pi = p′i . and therefore a = bc admits a similar factorization too. The theorem is
The class of each pi , i = 1, . . . , t in h0 is ambiguous. thus completely proved.
Let a be any nonzero ideal with a = a′ . We assert that
a can be written uniquely in the form Proposition 1.8 If K is a field, the polynomial ring K[X] in one vari-
able is a factorial ring.
a=r · pa11 · · · pat t , r ∈ Q, r > 0, ai = 0 or 1. (3.5)
Proof: First, let us observe that the set of units in K[X] is precisely
Proof of (3.5): Let n > 0, n ∈ Z be such that na = b is integral. K ∗ . If this were not true, let, if possible, f = an X n + · · · + a0 with
Then b = b′ . Let b = q1 · · · ql be the factorization of b into prime ideals. an 6= 0, n ≥ 1 be a unit in K[X]. Then, there exists g = bm X m + · · · + b0
Since b = b′ , we have, for any i, q′i = qj for some j. If i = j, then, with bm 6= 0 such that f g = 1. Then 0 = deg 1 = deg(f g) = n + m > 1,
unless qi is one of the ideals pk we have qi = qi O, where qi is a rational which is a contradiction.
prime. If i 6= j, then qi qj = (N (qi ))O. Hence b = c′ · pα1 1 · · · pαt t , where We start with the existence of a factorization. Let f be a non-
c′ , α1 , . . . , αt ∈ Z, αi ≥ 0. Since p2i = pi O, b = c · pa11 · · · pat t , c ∈ Z, ai = constant polynomial in K[X]. If f is an irreducible element, then there
0 or 1. Since a = n−1 b, is nothing to prove. If not, f has a non-constant divisor g not associated
with f , i.e. f = gh with 0 < deg g, deg h < deg f. Employing induction
a = r · pa11 · · · pat t , r ∈ Q, r > 0. on the degree of elements in K[X], it follows that g, h have factorizations
into irreducible elements of K[X] and consequently f = gh also can be
If, furthermore a = r1 · pb11 · · · pbt t , bi = 0 or 1, r1 > 0, then N (a) =
so factorized.
r 2 ·pa11 · · · pat t = r12 ·pb11 · · · pbt t , so that ai ≡ bi ( mod 2). Since each of ai , bi
The uniqueness of factorization can be established exactly as in the
is 0 or 1, we must have ai = bi . Hence rO = r1 O, and since r > 0, r1 > 0,
proof of Proposition 1.7, provided we have the following
this implies that r = r1 .
Suppose that b is a fractional ideal with b2 ≈ O. By multiplying b Lemma 1.1 In K[X], every irreducible element is prime.
by an integer, we may suppose that b is integral. Since, further bb′ ≈ O
we conclude that b = ωb′ where ω ∈ K, NK (ω) > 0 and ω ≻ 0 If dj > 0. Proof: Let p be an irreducible element of K[X] and let p divide the
Moreover, since NK (ω) > 0, ωω ′ = NK (ω) = NK (b)/NK (b′ ) = 1. Hence product gh of two polynomials g, h in K[X], and suppose that p ∤ g.
28 CHAPTER 1. PRELIMINARIES 3.4. LAWS OF QUADRATIC RECIPROCITY 65
√
Consider the set a of polynomials of the form up+vg where u, v ∈ K[X]. respectively. To prove this, we proceed as follows. Let α = p + q d+2 d

Then a is a non-zero ideal of K[X] and, since K[X] is a principal ideal be a unit in K. Then
domain, there is t ∈ K[X] with a = (t). Thus t divides p, but since
qd 2 q 2
 
p is irreducible, it follows that either p = ct with c ∈ K ∗ , or t ∈ K ∗ . NK (α) = p + + |d| = 1 (since NK (α) = αᾱ > 0).
2 4
If p = ct, then since t divides every element of a and, in particular, g
2
it follows that p | g. This contradicts our assumption. Hence t ∈ K ∗ Thus (p + qd 2 4
2 ) ≤ 1 and q ≤ |d| . If d < −4, then, of necessity, q = 0
and therefore there exists u1 , v1 in K[X] such that u1 p + v1 g = 1. Since and therefore .α = p = ±1 are the only units in K. Thus w = 2 for
gh = pw where w ∈ K[X], we have h = h(u1 p + v1 g) = p(u1 h + v1 w), d < −4. If d = −4, the q = 0, 1 or −1. If q = 0, p = ±1. If q = 1, p = 2
i.e. p | h. √
and if q = −1, then p = −2. Hence in K = Q( −4), the only units are
√
±1, ±i so that w = 4. Take now K = Q( −3). Then q = 0, 1 or −1. If
Remark 1.34 It can be shown, by similar reasoning, that any principal
then p − 3q
q = 0, p = ±1.√ If q = ±1 √ 1
2 = ± 2 . Hence the only units here
ideal domain is a factorial ring. −3 −3
1 1
are ±1, ± 2 + 2 , ± 2 − 2 so that w = 6.
We now give an example of a ring R which is not a factorial ring.
Take R to be the subring of C consisting of numbers of the form a +
√ √ 3.4 Laws of quadratic reciprocity
b (−5) with a, b ∈ Z and (−5) being a root of the polynomial x2 + 5.
In R, there are two distinct factorizations of 6, viz. 6 = 2 · 3 = (1 +
√ √ In Chapter 2, §3 we introduced the class group of an algebraic number
(−5))(1− (−5)). (It is not hard to see that the four numbers occurring field and proved that it is of finite order h.
here are irreducible.) Thus R cannot be a factorial ring and in R, an Let K be a quadratic field of discriminant d and O the ring of alge-
irreducible element is not prime in general. braic integers in K. Let Π0 denote the group of principal ideals λO with
We shall see however that R belongs to a general class of rings ad- λ ∈ K for which NK (λ) = λλ′ > 0. The quotient group of ∆ (the group
mitting unique factorization of ideals into prime ideals, which will be of all non-zero fractional ideals in K) modulo Π0 is denoted by h0 and
the object of our study in Chapter 2. called the restricted class group of K. Now Π0 is of index at most 2 in
Π and the order h0 of h0 is equal to h or 2h according as the index of Π0
1.13 Characters of a finite abelian group in Π is 1 or 2. If d < 0, trivially Π0 = Π since for α 6= 0 in K, we always
have NK (α) > 0. If d > 0, then Π = Π0 if and only if there exists in K
√
Let G be a finite abelian group, of order h. A character χ of G is a a unit of norm −1. (For dO is in the same coset of ∆ modulo Π0 as
√ √
mapping χ: G → C such that χ 6≡ 0, and χ(ab) = χ(a)χ(b) for a, b ∈ G. O if and only if d = ǫρ with NK (ρ) > 0. But, since NK ( d) < 0, this
If a ∈ G is such that χ(a) 6= 0, then for any b ∈ G, χ(a) = can happen, if and only if the unit ǫ has norm −1).
χ(b)χ(ab−1 ), so that χ(b) 6= 0. Hence χ is a homomorphism of G into
C∗ . Further, since bh = e we have [χ(b)]h = 1 for any b. Since there are Definition 3.4 The fractional ideals a, b different from 0 are equivalent
only finitely many hth roots of unity in C, it follows that there are only in the restricted sense (in symbols, a ≈ b) if a and b belong to the same
finitely many characters of G. coset of ∆ modulo Π0 (i.e. a = ρOb with ρ ∈ K and NK (ρ) > 0).
If we define the product χ1 χ2 of two characters χ1 , χ2 by (χ1 χ2 )(a) = √
It is clear that when K = Q( d), d < 0 or when d > 0 and K
χ1 (a)χ2 (a), then the characters form a finite abelian group Ĝ. contains a unit of norm −1, this concept coincides with the concept of
Proposition 1.9 Let G be a finite abelian group and let a ∈ G, a 6= e. equivalence introduced in Chapter 2, §3. In the case when d > 0, Π0
Then there exists a character χ of G such that χ(a) 6= 1. may be also defined to be the group of principal ideals λO with λ ∈ K
and λ > 0, λ′ > 0. Such numbers of a real quadratic field are referred to
Proof: Let a0 = e, a1 = a, a3 , . . . , ah−1
P be the elements of G. Let V as totally positive numbers (in symbols, λ ≻ 0). Thus Π0 consists of all
be the set of formal linear combinations λi ai , λi ∈ C. Clearly, V is a principal fractional ideals in K generated by a totally positive number.
64 CHAPTER 3. QUADRATIC FIELDS 1.13. CHARACTERS OF A FINITE ABELIAN GROUP 29

for α, β, β 6= 0 in O there exists γ, δ ∈ O such that α = γβ + δ with vector space of dimension h over C, and the elements of G form a base
0 ≤ |N (δ)| < |N (β)|. This leads easily to the fact that O is a principal of V.
ideal domain so that the class number of K is 1. We shall use repeatedly the following remark.

Let p be an odd prime in Z. By Proposition 3.2, pO = pp′ , p 6= p′ Remark 1.35 Let W 6= {0} be a finite dimensional vector space over
√
in K = Q( −1), if and only if ( −4 4 −1 −1
p ) = ( p )( p ) = ( p ) is equal to 1. In C, and T : W → W an endomorphism. Then there exists x 6= 0, x ∈ W
√ √
Q( −1) every ideals is principal so that p = αO for α = a + b −1 with such that T x = λx, λ ∈ C.
a, b ∈ Z. Since NK (α) > 0, we have a2 + b2 = Nk (α) = N (p) = p. Thus
we have In fact, every polynomial over C has a root in C, we can find λ ∈ C
such that det(T − M ) = 0. I: W → W being the identity map. then
Remark 3.3 An odd prime p is a sum of two squares of integers if and T − λI cannot be one-one by Remark 1.23 on page 23. Hence there is
only if ( −1
p ) = 1. x 6= 0 in W with T x − λx = 0. Such a λ is called an eigenvalue of T.
Any element ai (i = 0, 1, . . . , h − 1) gives rise to a permutation of
the elements of G viz, the permutation given by the mapping x 7→ ai x
3.3 The group of units of G onto G. There is a uniquely determined linear mapping Ai (i =
Let K be a quadratic field of discriminant d. In the notation of Chapter 0, 1, . . . , h − 1) of V which maps any element x in G to ai x. Further, if
2 §4 we see that r1 = 2, r2 = 0 if d > 0 and r1 = 0, r2 = 1 for d < 0. the linear mappings Ai , Aj of V correspond in this way to ai , aj in G
In the case of a real quadratic field K, the only roots of unity in K are respectively, then clearly Ai Aj corresponds to ai aj . Moreover, since G is
real roots of unity, namely, 1 and −1. By Theorem 2.4, every unit ǫ in K abelian, we have Ai Aj = Aj Ai . In addition, A0 is the identity mapping
can be written in the form ±ǫn1 , n ∈ Z for a fixed unit ǫ1 in K. Further I of V and Ahi = I for i = 0, 1, . . . , h − 1. Let us write A for the linear
ǫ1 6= ±1. If ǫ1 has this property, so have ǫ−1 −1 mapping A1 corresponding to a1 = a.
1 , −ǫ1 , −ǫ1 . But among
ǫ1 , ǫ−1 , −ǫ , −ǫ −1
, exactly one of them is greater than 1. We denote it We first prove that the linear mapping A corresponding to a ∈ G
1 1 1
by η and call it the fundamental unit of K. It is uniquely determined has at least one eigenvalue λ 6= 1, i.e. there exists λ 6= 1 in C and x 6= 0
and every unit ǫ is of the form ±η n for n ∈ Z. in V with Ax = λx. Clearly A 6= I, so that the space W = {Ax − x, x ∈
√ V} = 6 {0}. Also A maps W into itself since A(Ax − x) = Ay − y where
Any unit ǫ ∈ K = Q( d) of discriminant d > 0 gives rise to a
solution of the Diophantine equation y = Ax. Hence by the remark above, there exist y0 6= 0 in W and λ in C
such that ay0 = λy0 . Suppose, if possible that λ = 1. Let y0 = Ax0 − x0 .
x2 − dy 2 = ±4, x, y ∈ Z, (3.4) Then Ak (Ax0 − x0 ) = Ax0 − x0 for any k ≥ 0 in Z. Since Ak = I, we
√ 2 2
have (I + A + A2 + · · · + Ah−1 )(A − I) = 0 so that
(x + y d) x − dy
since NK (ǫ) = NK = and NK (ǫ) = ±1 in view of ǫ
2 4 (I + A + A2 + · · · + Ah−1 )(Ax0 − x0 ) = 0.
being a unit in K. Conversely, if, for d > 0 in Z, there exist x, y in Z
√ √
satisfying (3.4), then (x ± y d)/2 is a unit in k = Q( d). In the case But, since Ak (Ax0 − x0 ) = Ax0 − x0 this gives us h(Ax0 − x0 ) = 0,
when d is the discriminant of a real quadratic field, we have by Theorem contrary to our assumption that y0 = Ax0 − x0 6= 0. Hence λ cannot be
4, a non-trivial solution of the Diophantine equation (3.4). This equation equal to 1 and our assertion is proved.
is commonly referred to as Pell’s equation nor did he find a non-trivial Let now λ1 6= 1 be an eigenvalue of A1 = A and let V0 = V and
solution of it. V1 = {x ∈ V0 |A1 x = λ1 x}. Then V1 6= {0} and further, the mapping
If d > 0, K is an imaginary quadratic field and r = 0. Thus every unit Ai of V corresponding to any ai ∈ G maps V1 into itself; in fact, if
in K is a root of unity, by Theorem 2.4. By Lemma 2.10, we see that x ∈ V1 , then A1 (Ai x) = Ai (A1 x) = Ai (λ1 x) = λ1 (Ai x) so that Ai x ∈ V1 .
the units in K form a finite cyclic group of order w. One can, however, Hence again by our earlier remark, there exist λ2 in C and x 6= 0 in
check directly that w = 2, 4 or 6 according as d < −4, d = −4 or d = −3 V1 with A2 x = λ2 x. Let V2 = {x ∈ V1 | A2 x = λ2 x}. Again, each
30 CHAPTER 1. PRELIMINARIES 3.2. FACTORIZATION OF RATIONAL PRIMES IN K 63

Ai maps V2 into itself and we may continue the process above. let y are both even which is impossible. If 4 | y, then 4 | (2x + yd) implying
Vi+1 = {x ∈ Vi | Ai+1 x = λi+1 x} for i = 0, 1, 2, . . . , h − 2, where λi+1 that 4 | 2x, i.e. 2 | x. This contradicts the fact π ∈
/ 2O. Therefore y has to
is an eigenvalue of Ai+1 |Vi . For any x 6= 0 in Vh−1 we have Ai x = be odd. Find y2 ∈ Z such that yy2 ≡ 1(mod 8). (We have only to choose
2 2
λi x, i = 0, 1, 2, . . . , h − 1. Clearly λ0 = 1. If we set χ(ai ) = λi for y2 = ±1 ± 5.) Then d ≡ (2x + yd) y2 ( mod 8). Since 2 ∤ d, (2x + yd)y2
i = 0, 1, 2, . . . , h − 1, we obtain a character of G in fact, since(Ai Aj )x = is odd so that d ≡ 1( mod 8) and consequently ( d2 ) = 1.
Ai (λj x) = λi λj x, we have χ(ai aj ) = χ(ai )χ(aj ). Further χ(a) = λ1 6= 1. (iii) The proof is trivial, if we use (i) and (ii) above.
This proves Proposition 1.9. As an application of the criteria for splitting of rational primes
given above, we shall determine the class number h of a quadratic field
√
Proposition 1.10 (Orthogonality relations.) We have Q( m), m being square-free in Z, for special values of m.
 For this, we need to determine explicitly the constant C of Lemma 2.8
X h if χ1 = χ2 ,
S= χ1 (a)χ¯2 (a) = for the special case of a quadratic field. We claim that C can be chosen
0 otherwise,
a∈G to be 1 + |m| if m ≡ 2, 3( mod 4) and 2 + |m−1| 4 if m ≡ 1( mod 4).
 For m ≡ 2, 3( mod 4), d = 4m and taking the regular representation
√
 
k = order of Ĝ if a = b, 1 0
X
Ŝ = χ(a)χ̄(b) = with respect to the integral base {1, m} of O the matrices
0 otherwise.  0  1
χ∈Ĝ
√
  
0 m 1 0
Here χ̄(a) denotes the complex conjugate of χ(a). and correspond to 1, m respectively and det α1 +
 1 0  0 1
0 m
−1 α2 = α21 − mα22 so that C can be chosen to be 1 + |m|. The
Proof:P Since |χ(a)| = 1,Pwe have χ̄(a) = χ(a) . If χ1 = χ2 , we clearly 1 0
have a∈G χ1 (a)χ¯1 (a) = a∈G 1 = h. If χ1 6=Pχ2 , let b ∈ G be such that case m ≡ 1( mod 4) is dealt with in a similar fashion.
χ1 (b) 6= χ2 (b). Then we have S · (χ1 χ¯2 )(b) = a∈G (χ1 χ¯2 )(ab) = S since √ √
(1) Consider K = Q( 2). Here, d = 8 and O = Z + Z 2. Further
ab runs over all the elements of G when a does so. Since χ1 χ¯2 (b) 6= 1, we may take C = 3 in this case. Following the proof of Theorem 2.3, in
we have S = 0. order to find the number of ideal classes in K, it suffices to consider the
Similarly, if a = b, clearly Ŝ = k. If a 6= b, let χ1 ∈ Ĝ be such that √
splitting of prime ideals of norm at most 3. Now 2O = ( 2O)2 , while 3O
χ1 (ab−1 ) 6= 1. (This exists by Proposition 1.9.) Then 8
is prime since ( 3 ) = −1. Thus prime ideals of norm ≤ 3 are principal.
Any integral ideal of norm ≤ 3 is therefore principal so that h = 1.
√ √
X
Ŝχ1 (ab−1 ) = χχ1 (ab−1 ) = Ŝ, (2) Consider K = Q( −1). Here d = −4 and O = Z + Z −1. The
χ∈Ĝ constant C may now be chosen to be 2. To find h, it suffices to investigate
√ √
the prime ideals of norm at most 2. But 2O = (1− −1)O·(1+ −1)O =
so that Ŝ = 0. 2 √
p where p = (1 + −1)O. Thus any integral ideal of norm ≤ 2 is
It can be proved that G and Ĝ are isomorphic, so that k = h. This
principal and consequently h = 1.
is, however, unnecessary for our purposes and we do not go into this √ √
(3) Take K = Q( −5). Here d = −20 and O = Z + Z −5 which is
question.
not a factorial ring as remarked on page 28. Thus the class number h of
K clearly cannot be 1. Now the constant C = 6 and from the relations
√ √ √
2O = ((2, 1 + −5)O)2 , 3O = (3, 1 + −5)O · (3, 1 − −5)O, 5 =
√ 2 √ √ √
( −5O) 3(2, 1 − −5)O) = (1 + −5)(3, 1 − −5)O it is easy to see
that h = 2 in this case.

Remark 3.2 In the first two examples above one can show directly that
the ring O of algebraic integers possesses a Euclidean algorithm namely,
62 CHAPTER 3. QUADRATIC FIELDS

(Recall that the discriminant d of K is congruent to 0 or 1 modulo

4.)

Proposition 3.3 Let O be the ring of algebraic integers in a quadratic

field of discriminant d. Then we have

(i) 2O = p2 , p prime if and only if ( d2 ) = 0, Chapter 2

(ii) 2O = pp′ , p 6= p′ , p prime if and only if ( d2 ) = +1

(iii) 2O = p prime if and only if ( d2 ) = −1, Algebraic Number Fields

where ( d2 ) is Kronecker’s quadratic residue symbol.
√ √
Proof: (i) Let ( d2 ) = 0. Let p = (2, 1 + 2d ) or (2, 2d ) according as 8 2.1 Algebraic numbers and algebraic integers
does not or does divide d. Clearly p2 = 2O and p is necessarily prime.
Conversely, let 2O = p2 , p prime. If d ≡ 0(mod 4), ( d2 ) = 0 and there
√ Definition 2.1 A complex number α is said to be algebraic if α is a
1+ d
is nothing to prove. Let then d ≡ 1(mod 4) so that O = Z + Z( ). root of a polynomial an X n + · · · + a0 , a0 , . . . , an in Q not all zero.
√ 2
Since p2 6= p, there exists π ∈ p, π ∈ / p2 = 2O. Let π = x + y 1+2 d with
If α is algebraic then the mapping from Q[X] to C taking ni=0 ai X i
P
x, y ∈ Z. We can assume that x and y are either 0 or 1 for along with P i
to ai α is a homomorphism with kernel 6= 0.
π, π + 2α for any α ∈ O also satisfies π + 2α ∈ p but ∈ / p2 .
If y = 0, then x 6= 0 since otherwise π = 0 and π ∈ p2 . Further Definition 2.2 A complex number α is called transcendental if it is not
x 6= 1, since then π = 1 and π ∈ / p. So y = 1 and x can be 0 or 1. Now algebraic.
 √ 2  √   
1+ d 1+ d
2
π = x+ 2 = a+b 2 for a, b ∈ Z. Then b = 2 x + 21 =
Let a be the set of all polynomials in Q[X] having a fixed algebraic
2x + 1 is necessarily odd. But since π 2 ∈ 2O, b must be even and so we
number α as a root. Clearly a is an ideal of Q[X] and, in fact, generated
arrive at a contradiction if we assume that d ≡ 1( mod 4). Consequently
over Q[X] by a non-constant polynomial, say t in view of Remark 1.27
d ≡ 0( mod 4) so that ( d2 ) = 0.
on page 24. Now t is necessarily irreducible. Otherwise, t = uv with
(ii)√Let ( d2 ) = 1. Then d ≡ √1( mod 4) necessarily, so that O =
√ 0 ≤ deg u, deg v < deg t and since 0 = t(α) = u(α)v(α) at least one
Z + 1+2 d Z. Defining p = 2O + 1+2 d O, we see that pp′ = (4, 1 + d, 1 − of u, v is in a, contradicting the minimality of deg t. The polynomial t
√ 1− d √  √ √ √ √
d, 4 . Since 2 = 1 + d + 1 − d and 4, 1 + d, 1 − d and 1−d 4 is clearly determined uniquely upto a constant factor. We may suppose
are all in 2O we have pp′ = 2O. (Here p 6= p′ since otherwise, we would therefore that the leading coefficient of t is 1; so normalized, it is called
have 2O = p2 and 4 | d by (ii).) the minimal polynomial of x.
Conversely, let 2O = √ pp′ , p 6= p′ , p prime. Then N (p) = 2. Further
Definition 2.3 By the degree of an algebraic number α, we mean the
there exists π = x + y d+2 d ∈ p, π ∈ / pp′ = 2O, so that x, y are integers
degree of the minimal polynomial of α.
which are not both even. Since 2 = N (p) divides N (πO) = |N (π)|
we have (2x + yd)2 ≡ y 2 d( mod 8). Now 2 ∤ d, since otherwise, by
p
Example 2.1 The “quadratic irrationality” (−5) is of degree 2.
(i), we would have 2O = p2 and then p = p′ . If y is even, let first
y = 2y1 , 2 ∤ y1 , y1 ∈ Z. Then 2|((x + y1 d)2 + y12 d), and 2 ∤ y1 d together Let α be an algebraic number of degree
P n. Consider the subring Q[α]
give us 2 ∤ (x + y1 d) and this in turn implies that 2 | x. But then x and of C consisting of numbers of the form m i
i=0 ai α with ai ∈ Q. We now

31
32 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 3.2. FACTORIZATION OF RATIONAL PRIMES IN K 61

assert that Q[α] is a field. Let, in fact, f be thePminimal polynomial of (i) pO = p2 , p prime if and only if ( dp ) = 0,
α. Consider the mapping q: Q[X] → C taking m i
Pm i
i=0 bi X to i=0 bi α .
This is a homomorphism onto Q[α] with kernel f Q[X] = (f ). By the (ii) pO = pp′ , p 6= p′ , p prime if and only if ( pd ) = +1,
homomorphism theorem Q[X]/(f ) is isomorphic to Q[α]. Let now g ∈
(iii) pO = p prime if and only if ( pd ) = −1.
Q[X]/(f ) be such that g 6= 0 i.e. f ∤ g. The ideal generated by f and
g in Q[X] is a principal ideal b generated by, say h. Since h divides f , where ( dp ) is the Legendre symbol.
we have h = cf, c ∈ Q∗ unless h ∈ Q∗ . The former case is impossible,
Proof:
since h | g and f ∤ g. Thus b = Q[X] and consequently there exist k, l in √
Q[X], such that kf + lg = 1, so that ḡ¯l = 1. This proves that Q[X]/(f ), (i) Let pO = p2 , p prime. Then there exists π = m + n d+2 d ∈ p, π ∈ /
and Q[α] is a field. pO, m, n ∈ Z. Now, since π 2 ∈ pO we see that p divides both
We denote the field Q[α] by Q(α). (2m + nd) + d · n2 and n(2m + nd). If now p | n, then p | (2m + nd).
Since p is odd, this would imply that p | m, but then pO divides
Definition 2.4 A subfield K of C is called an algebraic number field if π, which is a contradiction. Thus p | (2m + nd) and p ∤ n. Since,
its dimension as a vector space over Q is finite. The dimension of K further, p | dn2 , this implies that p | d, i.e. ( dp ) = 0.
over Q is called the degree of K, and is denoted by [K : Q]. √
Conversely, if ( dp ) = 0, consider p = pO + dO. Then p2 =
√ √
Example 2.2 Q and Q( (−5)) are algebraic number fields of degree 1 (p2 , p d, d)O = pO since p is the gcd of d and p2 . Further, p
and 2 respectively. is necessarily a prime ideal, since at most two prime ideals of O
can divide pO (see page 60.)
Remark 2.1 Any element α of an algebraic number field K is alge- (ii) Let ( pd ) = 1. Then there exists a ∈ Z such that a2 ≡ d(mod p).
braic. (For, if [K : Q] = n then 1, α, α2 , . . . , αn are necessarily linearly √
Let p be the ideal generated by p and a + d. Then clearly, pp′ =
dependent over Q.) √ √
(p , p(a + d), p(a − d), a − d)O = pO [p ∈ pp′ since p = g.c.d.
2 2

(p2 , 2ap)]. Hence p, p′ are prime ideals. However p + p′ = O and

Remark 2.2 If α is an algebraic number of degree n, then Q[α] is an therefore p 6= p′ .
algebraic number field of degree n(Q(α) = Q[α]).
Conversely, let pO = pp′ , p 6= p′ , p prime. Then N (p) = N (p′ )√=
Remark 2.3 Any monic irreducible polynomial f in Q[X] is the mini- p. There exists α ∈ p, α ∈ / pO = pp′ . Then α = x + y d+2 d
mal polynomial of any of its roots. For, if α is a root of f, then α is an with x, y ∈ Z, p not dividing both x and y. Since αO = pq with
algebraic number and its minimal polynomial φ certainly divides f since q ⊂ O, it follows, on taking norms, that p = N (p) divides N (αO) =
2
f (α) = 0. Now, φ cannot be in Q so that the irreducibility of f gives |NK (α)| = |(x + y d2 ) − y 2 d4 |. Hence (2x + dy)2 ≡ y 2 d(mod p). If
f = c · φ for c 6= 0 in Q. Since both f and φ are monic, it follows that p | y, then p|(2x + dy)2 , i.e. p | 2x. Since p is odd, p | x, i.e.
φ = f. p | (x, y). We have thus a contradiction. Hence p ∤ y and since
Z/(p) is a field, we have z 2 ≡ d(mod p) for z ∈ Z, i.e. ( pd ) = 1.
Remark 2.4 Let α be an algebraic number of degree n(≥ 1) and f =
Pn i (iii) The validity of (iii) is an immediate consequence of (i) and (ii).
i=0 ai X (an = 1) be its minimal polynomial in Q[X]. We now claim
that all the roots of f are distinct, i.e. f has no repeated roots. In fact, Before we consider the factorization of 2O in K, we define Kro-
suppose that f = (X − α)2 g. Then, the derivative f ′ of f is 2(X − α)g + necker’s quadratic residue symbol ( d2 ) with denominator 2 by
(X − α)2 g′ , so that f ′ (α) = 0. Since f is the minimal polynomial of 
   0 if d ≡ 0( mod 4)
α, this implies that f |f ′ . Since deg f ′ < deg f , this is impossible unless d
= 1 if d ≡ 1( mod 8)
f ′ = 0, so that f ∈ K, which is absurd. 2
−1 if d ≡ 5( mod 8)

60 CHAPTER 3. QUADRATIC FIELDS 2.1. ALGEBRAIC NUMBERS AND ALGEBRAIC INTEGERS 33

In future, K will always stand for a quadratic field with discriminant d, Remark 2.5 Let α1 and α2 be two algebraic numbers with the same
K will be real quadratic or imaginary quadratic according as d > 0 or minimal polynomial in Q[X]. Then, for any g in Q[X], we see that
d < 0. g(α1 ) = 0 if and only if g(α2 ) = 0. It is easy now to deduce that the
√ √
The mapping taking α = x+ y d (x, y ∈ Q) to α′ = x− y d may be mapping φ: Q[α1 ] → Q[α2 ] defined by
seen to be an automorphism of K. An element α ∈ K satisfies α = α′ if
m m
and only if α ∈ Q. For any subset S of K, let us denote by S ′ the image
X  X
φ ai αi1 = ai αi2 (a0 , ai , . . . , am ∈ Q)
of S under this automorphism. Since O = O′ it is clear that for any
i=0 i=0
fractional ideal a, a′ is again a fractional ideal. Clearly N (a) = N (a′ ).
For any integral ideal a we claim that aa′ where n = N (a) ∈ Z. is an isomorphism of Q(α1 ) onto Q(α2 ). The mapping φ is the identity
Let p be any prime ideal in O. Now, p contains a unique prime number on Q and takes α1 to α2 . Conversely, let α1 be any algebraic number
p > 0, p ∈ Z by Remark 2.18 Further, p occurs in the factorization with minimal polynomial f and φ an isomorphism of Q(α1 ) into C such
p1 · · · pr of pO into prime ideals p1 , . . . , pr . By the corollary to Lemma 2.4 that φ(a) = a, for any a ∈ Q. Then for any g ∈ Q[X], g(α1 ) = 0 if and
and by Lemma 2.6, we have p2 = NK (p) = N (pO) = N (p1 ) · · · N (pr ). only if g(φ(α1 )) = 0. The set of all polynomials in Q[X] having φ(α1 )
Since Z is a factorial ring, we have r ≤ 2 and N (pi ) = p or p2 . Thus as a root is precisely the ideal f Q[X] and therefore φ(α1 ) is an algebraic
pO = pp′ or p. But if p divides pO so does p′ . Thus we have either number with f as its minimal polynomial.

pO = pp′ , p 6= p′ Definition 2.5 Two algebraic numbers α1 and α2 as in Remark 2.5

or above are said to be conjugates of each other (over Q).
pO = p = p′
or Remark 2.6 Let K be an algebraic number field of degree n. Then
2 ′ there exists θ ∈ K such that K = Q(θ). Observe that such a number θ
pO = p , p = p (3.3) is an algebraic number of degree n, (by Remarks 2.1 and 2.2 above.)
In any case, pp′ = pO or p2 O i.e. pp′ = N (p)O. By Lemma 2.6, aa′ =
N (a)O for any integral ideal a. For proving this, we remark first that the intersection of any family of
subfields of a fixed field is again a field, and denote by Q(α1 , . . . , αp ) the
intersection of all subfields of C containing Q and the complex num-
3.2 Factorization of rational primes in K bers α1 , α2 , . . . , αp . (This field is also referred to as the subfield of C
generated by α1 , α2 , . . . , αp over Q.) Since [K : Q] = n, there exist
Let K be a quadratic field of discriminant d and O the ring of algebraic ω1 , ω2 , . . . , ωq (q ≤ n in K such that K = Q(ω1 , . . . , ωq ). By Remarks 2.1
integers in K. Let p be a prime number in Z. We start from the three and 2.4 above, ω1 , . . . , ωq . are all “separably algebraic” over Q i.e. the
possibilities given by 3.3. minimal polynomial of any ωi has no repeated roots. if g = 1, there is
nothing to prove. Let first q = 2. We shall prove that K = Q(ω1 , ω2 )
Definition 3.3 If pO = pp′ , p 6= p′ , we say that p splits in K. If p = p′ ,
contains θ1 such that K = Q(θ1 ). For simplicity of notation, let us de-
then either pO = p in which case we say that p stays prime in K or
note ω1 , ω2 by γ, δ respectively and let f and φ be their respective min-
pO = p2 in which case we say that p is ramified in K.
imal polynomials. By the ‘fundamental theorem of algebra’, we have
If p is an odd prime, the following proposition gives criteria for the f = (X − γ1 ) · · · (X − γr ) and φ = (X − δ1 ) · · · (X − δs ) where we may
factorization of pO in K. assume, without loss of generality, that γ = γ1 , δ = δ1 . Further, by Re-
mark 2.4 above, δ1 , δ2 , . . . , δs are pairwise distinct. Since Q is infinite, we
Proposition 3.2 For an odd prime p and a quadratic field of discrim- can find λ 6= 0 in Q such that γi +λδj 6= γi 1 +λδj 1 unless j = j1 and i = i1
inant d, we have (We have just to choose λ in Q different from 0 and −(γi − γi 1 )/(δj − δj 1
34 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 3.1. GENERALITIES 59

√
for j 6= j1 .) Set θ1 = γ +λδ = γ1 +λδ1 and denote Q(θ1 ) by L; obviously, O = Z + Z m.
L ⊂ K. The polynomial ψ(X) = f (θ1 − λX) ∈ L[X] has δ1 as a root,
Thus we have
since ψ(δ1 ) = f (θ1 − λδ1 ) = f (γ1 ) = 0. Further, for i 6= 1, ψ(δi ) 6= 0 √
since, otherwise, f (θ1 − λδi ) = 0 would give us θ1 − λδi = γj for some j Z + Z 1+2 m , for m ≡ 1(mod 4),
(
i.e. γ1 + λδ1 = γj + λδi for i 6= 1 contrary to our choice of λ. Thus φ and O= √ (3.1)
ψ have exactly one root γ1 in common. Let χ be the greatest common Z + Z m, for m ≡ 2, 3(mod 4)
divisor of φ and ψ in L[X]. Then every root of χ in C is a common (Since m is square-free, the case m ≡ 0(mod 4) does not arise.) Observe
root of φ and ψ. Thus χ is necessarily of degree 1 and hence of the form √ √
that if α = a + b m ∈ O, so does α′ = a − b m.
µ(X − γ1 ). In other words, µ, µγ1 ∈ L. i.e. γ1 = θ1 − λδ1 ∈ L(= Q(θ1 )). Let a be an integral ideal, and (α1 , α2 ) an integral base of a. We
Therefore, and δ1 ∈ L, i.e. K ⊂ L. This implies that K = Q(θ1 ). Let define the discriminant of a, written
 ∆(a),to be the square ∆(α1 , α2 )
now q ≥ 3. Assume by induction, that every algebraic number field of α1 α2
the form Q(α1 , α2 . . . , αr ) with r ≤ q − 1 contains a number α such of the determinant of the matrix (the prime are conjugates
α′1 α′2
that Q(α1 , α2 , . . . , αr ) = Q(α). Then K= 1Q(ω1 , ω2 , . . . , ωq−1 ) = Q(θ1 ) as defined above) i.e.
for some θ1 ∈ K1 . Further K = K1 (ωq ) = Q(θ1 , ωq ) = Q(θ) for some
2
θ ∈ K(because of the special case q = 2 established above). ∆(a) = ∆(α1 , α2 ) = (α1 α′2 − α′1 α2 )

Remark 2.7 Let K be an algebraic number field of degree n. Then there If (β1 , β2 ) is another base of a then β1= pα1+ qα2 , β2 = rα1 + sα2
p q
exist precisely n distinct isomorphisms σ1 , σ2 , . . . σn of K into C which where, we have p, q, r, s ∈ Z. If P = , we have det P = ±1.
r s
are the identity on Q. By Remark 2.6 above K = Q(θ) for a number 2
θ ∈ K whose minimal polynomial f in Q[X] is of degree n. Let θ1 (= It follows that ∆(β1 , β2 ) = ∆(α1 , α2 )(det P ) = ∆(α1 , α2 ) so that the
θ), θ2 , . . . , θn be all the distinct root of f. Then, by Remark 2.5, above, above definition is independent of the integral base of a. If a = O, we
there exists, for each θi (i P = 1, 2, . . . , n) an write d = d(K) = ∆(O), and call it the discriminant of the field K.
j Pmisomorphism
j
σi of Q(θ1 ) onto
Using (3.1) we find that
Q(θi ) ⊂ C defined by σi ( m j=0 aj θ1 ) = j=0 aj θi for a0 , a1 , . . . am ∈ Q.
By definition σi (a) = a for all a ∈ Q, σ1 is the identity isomorphism of 
m for m ≡ 1(mod 4)
K. Since θi 6= θj for i 6= j, the isomorphism σ1 , σ2 , . . . , σn are all distinct. d=
4m for m ≡ 2, 3(mod 4)
On the other hand, let σ be any isomorphism of K = Q(θ1 ) into C which
and therefore d is always congruent to 0 or 1 modulo 4. We have thus
is the identity on Q. By Remark 2.5 above, σ(θ P1 ) = θi jfor some Pm i(1 ≤ j proved
i ≤ n) and therefore for a0 , a1 , . . . , am ∈ Q, σ( m j=0 aj θ1 ) = j=0 aj θi .
Thus σ is necessarily one of the n isomorphisms σ1 , σ2 , . . . , σn .
Proposition 3.1 For a quadratic
√ field K with discriminant d, we have
√
K = Q( d) and further 1, d+2 d is an integral base of the ring O of
Let K be an algebraic number field of degree n, and σ1 , σ2 , σn the n
algebraic integers in K.
distinct isomorphism of K into C. We denote the image σi (K) of K by
K (i) and, for α ∈ K, σi (α) by α(i) . Let σ1 be the identity isomorphism of Corollary 3.1 The discriminant uniquely determines a quadratic field.
K; we have then K (1) = K and α(1) = α for any α ∈ K. Since each σi is
an isomorphism which is the identity on Q, it follows that K (1) , . . . , K (n) Remark 3.1 Let {α1 , α2 } be an integral base of an integral ideal a cho-
are again algebraic number fields of degree n. They are referred to as sen as in Proposition 3.1, i.e. α1 = p11 ω1 + p12 ω2 , α2 = p22 ω2 , p11 , p12 ,
the “conjugates” of K. If K (i) ⊂ R, we call it a real conjugate of K p22 ∈ Z, p11 , p22 > 0 and O = Zω1 + Zω2 . Then it is clear that
and if K (i) 6⊂ R, then we call it a complex conjugate of K. We now ∆(α1 , α2 ) = p211 p222 = p211 p222 · d. But p11 p22 is precisely N (a). Thus
claim that the complex conjugates of K occur in pairs, i.e. the distinct
isomorphisms σi with σi (K) 6⊂ R occur in pairs σ, ρ with ρ = σ̄, where ∆(a) = (N (a))2 d. (3.2)
58 CHAPTER 3. QUADRATIC FIELDS 2.1. ALGEBRAIC NUMBERS AND ALGEBRAIC INTEGERS 35

√
Any α ∈ K is of the form p + q m, p, q ∈ Q; define the conjugate σ̄(α) = σ(α) is the complex conjugate of σ(α). For, by Remark 2.6 above,
√
α′ of α by α′ = p − q m. It is clear that α is a root of the polynomial K = Q(θ) for some θ ∈ K. If K (i) = σi (K) is a complex conjugate of
(X −α)(X −α′ ) = X 2 −(α+α′ )X +αα′ = X 2 −2pX +p2 −q 2 m ∈ Q[X]. K, then necessarily θ(i) = σi (θ) is a complex number which is not real
It follows that α′ is the conjugate of α in the sense of Chapter II. Taking Now θ (i) is a root of the minimal polynomial ni=0 ai X i of θ and, since
P
√
the regular representation  of K with
 respect to the base (1, m) of K a0 , a1 , . . . , an ∈PQ, it follows that the complex conjugate θ (i) of θ (i) is
p qm also a root of ai X i . Hence by Remark 2.7 above, Q(θ (i) ) too occurs
over Q, the matrix A = corresponds to α and the polynomial
q p   among the conjugates K (i) , . . . , K (n) . Let r1 be the number of complex
X − p −qm conjugates of K. By the foregoing, s = 2r2 for r2 ∈ Z+ . Further we have
above is merely det(XI2 − A) = det . Observe that
−q X −p r1 + 2r2 = n.
for α ∈ K, Tr K (α) = Tr (A) = 2p = α + α′ and NK (α) = det A =
p2 − q 2 m = αα′ . If K is imaginary quadratic, then α′ is the complex Remark 2.8 Let K be an algebraic number field and ω1 , ω2 , . . . , ωn be
conjugate of α ∈ K, so that, for any α 6= 0 in an imaginary quadratic a base of K over Q. With the notation introduced above, let Ω denote
field K, the norm NK (α) is always positive. (i) (i) (i)
the n-rowed complex square matrix (ωj ) with (ω1 , ω (i) , . . . , ωn ) as its
Let O be the ring of algebraic integers in K. Any α ∈ O is of the i-th row. Then Ω has an inverse in Mn (C).
√
form p + q m for some p, q ∈ Q. If the minimal polynomial of α is of
degree 1, then by Proposition 2.1, it is necessarily of the form, X − a In fact, K = Q(θ) for some algebraic number in K of degree n, by Re-
for a ∈ Q so that p = a ∈ Z and q = 0. Thus α = α′ − 2p = 2a and mark 2.6, above. Further, if σ1 , σ2 , . . . , σn are the distinct isomorphisms
αα′ = p2 −q 2 m = a2 are both in Z. Let now the minimal polynomial of α of K into C, then θ (1) = σ1 (θ) = θ, θ (2) = σ2 (θ), . . . , θ (n) = σn (θ) are
which is a monic polynomial in Z[X], be of degree 2, say X 2 +cX +d with all conjugates of θ (by Remark 2.5, above). Moreover, they are distinct,
c, d ∈ Z. Since α is a root of the polynomial X 2 −2pX +p2 −q 2 m ∈ Q[X], (by Remark 2.4). Now α1 = 1, α2 = θ, . . . , αn = θ n−1 form a base
we have necessarily X 2 − 2pX + p2 − q 2 m ≡ X 2 + cX + d i.e −c = 2p = (i)
of K over Q. Let A = (αj ) be the matrix for the base α1 , α2 , . . . , αn ,
α + α′ = Tr K (α) and d = p2 − q 2 m = αα′ = NK (α). Conversely, for built as Ω was from the base ω1 , . . . , ωn . Then,Qdet A is the well-known
√
p, q ∈ Q, if 2p and p2 − q 2 m are in Z, then α = p + q m ∈ O. Thus, Vandermonde determinant and is equal to ± 1≤i<j≤n (θ (i) − θ (j)), so
√
for α = p + q m ∈ K to belong to O, it is necessary and sufficient that that det A 6= 0. If ω1 , ω2 , . . . , ωn is any base of K over Q, then clearly,
Tr K (α) = 2p and NK (α) = p2 − q 2 m are both in Z. We use this to (i) (i)
for 1 ≤ i, j ≤ n, we have ωj = nk=1 pjk αk with pjk ∈ Q. Since both
P
construct, explicitly, an integral base of O. {α1 , α2 , . . . , αn } and {ω1 , ω2 , . . . , ωn } form bases of K, it follows that
√
For p, q ∈ Q, let α = p + q m be in O. Then a = 2p, b = p2 − q 2 m the n-rowed matrix P = (pjk ) with (p1k , p2k , . . . , pnk ) as its kth row has
a2 −4q 2 m
belong to Z. Hence 4 ∈ Z. In particular, 4q 2 m ∈ Z. Since m an inverse in Mn (Q). Clearly, Ω = AP so that det Ω = det A · P 6= 0.
is square-free, it follows that q = f /2 with f ∈ Z. Now a2 − f 2 m ≡ Thus Ω has in inverse in Mn (C).
0(mod 4). We have to distinguish between two cases. In what follows, we shall prove a few of the important theorems
concerning algebraic number fields.
(1) Let m ≡ 1(mod 4). Then a2 ≡ f 2 (mod 4) i.e. f and a are both √
even or both odd. In this case, it is clear that α = a + b 1+2 m Definition 2.6 A complex number α is said to be an algebraic integer
√
with a, b ∈ Z, i.e. O = Z + Z 1+2 m . Note that, if m ≡ if α is a root of a monic polynomial in Z[X].
√

1+ m Remark 2.9 An algebraic integer is an algebraic number.
1(mod 4), 2 ∈O .
Remark 2.10 An element of Z is an algebraic number.
(2) Let m ≡ 2, 3(mod 4). Then a2 ≡ f 2 m( mod 4) if and only if a and
√
f are both even showing that α = a′ + b′ m with a′ , b′ ∈ Z, i.e Remark 2.11 If α ∈ Q is an algebraic integer, then α ∈ Z.
36 CHAPTER 2. ALGEBRAIC NUMBER FIELDS

Remark 2.12 For any algebraic number α, there exists m 6= 0 in

Z such that mα is an algebraic integer. (For, if αn + an−1 αn−1 +
· · · + a0 = 0, with a0 , a1 , . . . , an−1 ∈ Q then take m ∈ Z such that
ma0 , ma1 , . . . , man−1 ∈ Z.)

Definition 2.7 A polynomial f = an X n + · · · + a0 ∈ Z[X] is said to be

primitive if the gcd(a0 , a1 , . . . , an ) of a0 , a1 , . . . , an is 1. Chapter 3
In particular, a monic polynomial in Z[X] is primitive. It is clear
that any polynomial f ∈ Z[X] can be written in the form cg where c ∈ Z
and g is primitive.
Quadratic Fields
Remark 2.13 Any polynomial f ∈ Q[X] can be written in the form
(a/b)g where g is primitive and a, b ∈ b are such that (a, b) = 1. (a, b)
stands for the g.c.d. of a and b. 3.1 Generalities
Lemma 2.1 (Gauss.) The product of two primitive polynomials in Definition 3.1 By a quadratic field we mean an algebraic number field
Z[X] is primitive. of degree 2.

Proof: Let φ = an X n + · · ·+ a0 , ψ = bm X m + · · ·+ b0 be in Z[X] with Let K be a quadratic field and let α 6= 0 be in K. Since [K : Q] =
(a0 , a1 , . . . , an ) = 1 = (b0 , b1 , . . . , bm ). Suppose that p is a prime dividing 2, 1, α, α2 are linearly dependent over Q, i.e a0 + a1 α + a2 α2 = 0 for
all the coefficients of f = φψ. Consider the natural map η: Z → Z/(p); a0 , a1 , a2 in Q not all zero. Thus, any α in K is a root of an irreducible
this induces a homomorphism (which we denote again by) η: Z[X] → polynomial in Q[X] of degree at most 2. But K should contain at
Z/(p)[X]. We have η(f ) = 0, while, since the gcd of the coefficients least one element β whose irreducible polynomial in Q[X] is of degree
of φ, ψ is 1, we have η(φ) 6= 0, η(ψ) 6= 0, so that since Z/(p) is a 2, since, otherwise, K = Q. Then 1, β form a base of K over Q i.e.
field Z/(p)[X] is an integral domain, and η(f ) = η(φ) · η(ψ) 6= 0, a K = Q(β). Let a2 β 2 + a1 β + a0 = 0 where, without loss of generality, we
contradiction. Hence f is primitive. may suppose that a0 , a1 , a2 ∈ Z, a2 6= 0. Multiplying by 4a2 , we have
(2a2 β + a1 )2 = a21 − 4a0 a2 . Setting γ = 2a2 β + a1 we have K = Q(γ).
Proposition 2.1 The following statements are equivalent. √ √
Denoting a21 − 4a0 a2 by m ∈ Z we see that K = Q( m) where by m
we mean the positive square root of m if m > 0 and the square root
(i) α is an algebraic integer.
of m with positive imaginary part if m < 0. We could further suppose,
(ii) The minimal polynomial of α is a (monic) polynomial in Z[X]. without loss of generality, that m is square-free (i.e m 6= 1 and m is not
divisible by the square of any prime).
(iii) Z[α] is a finitely generated Z-module.

(iv) There exists a finitely generated Z-submodule M 6= {0} of C such Definition 3.2 A quadratic field K is called a real or an imaginary
that αM ⊂ M. quadratic field according as K ⊂ R or not.
√
Proof: (i)⇒ (ii). Let αn + an−1 αn−1 + · · · + a0 = 0 with ai ∈ Z and A quadratic field K is real if and only if K = Q( m) with square
φ = X n + an−1 X n−1 + · · · + a0 . Let f be the minimal polynomial of free m > 1 in Z. Note that if K is an imaginary quadratic field, then
α in Q[X]. By definition, φ = f ψ, where ψ ∈ Q[X]. Further, by the K ∩ R = Q.

57
56 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.1. ALGEBRAIC NUMBERS AND ALGEBRAIC INTEGERS 37

Remark 2.13 on page 36, let f = (a/b)f1 , ψ = (c/d)ψ1 with primitive f1

and ψ1 and a, b, c, d ∈ Z such that (a, b) = (c, d) = 1. Now bdφ = acf1 ψ1 .
By Gauss’ Lemma, f1 ψ1 is primitive. Let us compare the g.c.d. of the
coefficients on both sides. Since φ is monic and hence primitive, we have
ac = ±bd. Thus φ = ±f1 ψ1 ; comparison of leading coefficients implies
that the leading coefficient of f1 is ±1; since f1 (α) = 0, it follows from
the definition of the minimal polynomial that f = ±f1 , so that f ∈ Z[X].
(ii)⇒ (iii): Let φ = X n +an−1 X n−1 +· · ·+a0 ∈ Z[X] be a polynomial
with φ(α) = 0. Then clearly Z[α] is generated by 1, α, . . . , αn−1 over Z.
(iii)⇒ (iv): It is obvious that αZ[α] ⊂ Z[α]. Take M = Z[α].
(iv)⇒ (v): Let M = Zv1 + · · · + Zvn ⊂PC be a finitely generated
Z-module such that αM ⊂ M. Then αvi = nj=1 aij vj (i = P1, 2, . . . , n)
with aij ∈ Z. Let A = (aij ) and B = αIn −A = (bij ). Then nj=1 bij vj =
0 for i = 1, 2, . . . , n. Let V be a vector space of dimension n over
C and e1 , e2 , . . . ,P en a base of V. The linear mapping φ of V into it-
n
self
Pn taking e j to i=1 bij ei (j = 1, 2, . . . , n) takes the non-zero element
j=1 v j ej to 0. By the remark 1.23 on page 23, det φ = det B = 0. Ex-
panding det B = det(αIn − A), we see that αn + an−1 αn−1 + · · · + a0 = 0
for a0 , a1 , . . . , an−1 ∈ Z.
Proposition 2.1 is now completely proved.
Let K be an algebraic number field. Denote by O the set of algebraic
integers in K. If α, β ∈ O, then Z[α], Z[βa] are finitely generated Z-
modules by Proposition 2.1. Hence the ring M = Z[α, β], is a finitely
generated Z-module. Since if γ is one of α ± β, αβ we clearly have
γM ⊂ M , it follows from Proposition 2.1 that α ± β, αβ are in O so
that O is a ring. By Remark 2.12 on page 36, K is the quotient field of
O.
If α, β are algebraic, there exists m ∈ Z such that mα, mβ are
algebraic integers, by Remark 2.12. Since m(α ± β) and m2 αβ are
algebraic integers by the considerations above, it follows that α ± β and
αβ are algebraic. Further, if α 6= 0 is algebraic, so is α−1 . Thus, algebraic
numbers form a subfield of C.
Let K be an algebraic number field of degree n and ω1 , . . . , ωn be
a base of K over Q. For any α ∈ K,the mapping x 7→ αx is a linear
mapping of K into itself considered as a vector space over Q. We define
the trace Tr (α) = Tr K (α), and norm N (α) = NK (α) of the element
α to respectively, the trace and determinant of this linear mapping. If
αwj = ni=1 aij wi , j = 1, 2, . . . , n, and A = (aij ), we have Tr K (α) =
P
Tr (A), NK (α) = det A; hence Tr K (α), NK (α) ∈ Q.
38 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.4. THE GROUP OF UNITS 55

(k) Pn (k)
For α ∈ K, (αwj )(k) = α(k) wj = i=1 aij wi for j = 1, 2, . . . , n. · · · + cr log |ǫ(r) | = 0 for all ǫ ∈ U. We may suppose without loss of
Denote the n-rowed square matrix
(k)
(wj )
by Ω, as on page 35, and the generality, that at least one ci < 0. Let A be the set of k ≤ r − 1 with
c2 < 0, and B the complement of A in the set E of integers ≤ r + 1.
n-rowed square matrix (α(i) δij ) by A0 where δij = 1 for i = j and δij = 0
Clearly A ∩ B = ∅, A ∪ B = E, and A and B are nonempty (B contains
for i 6= j. Then we have A0 Ω = ΩA. Since, by Remark 2.8 on page 35,
r+1.) By Lemma 2.14, there is ǫ ∈ U with |ǫ(k) | < 1 for k ∈ A, |ǫ(k) | > 1
Ω has an inverse Ω−1 , we have A0 = ΩAΩ−1 . Hence we have (k)
for k ∈ B. PrBut then,(k)for all k, ck log |ǫ | ≥ 0, and is zero only if ck = 0,
NK (α) = det A = det(ΩAΩ−1 ) = det A0 = α(1) · · · α(n) . (2.1) so that 1 ck log |ǫ | > 0. This contradiction proves that t = r, and
−1 (1) (n)
Theorem 2.4 is completely established.
Tr K (α) = Tr (A) = Tr (ΩAΩ ) = Tr (A0 ) = α + ··· + α .
Further, if A corresponds to α ∈ K, B to β ∈ K, then to α + β Remark 2.23 The above proof of Theorem 2.4 follows, in essentials a
corresponds A + B and to αβ, AB Hence the mapping α 7→ A is a proof given by C.L. Siegel in a course of lectures in Göttingen. It does
homomorphism of K into Mn (Q). It is called a regular representation not seem to be available in the literature.
of K (viz. that corresponding to the base w1 , . . . , wn of K.) We verify
immediately that if α, β ∈ K, we have
Tr K (α + β) = Tr K (α) + Tr K (β) and NK (αβ) = NK (α)NK (β).
Let α be an algebraic integer in K. By Proposition 2.1, we can sup-
pose that the minimal polynomial of α is X m + am−1 X m−1 + · · · +
a0 where a0 , a1 , . . . , am−1 ∈ Z. Now α is of degree m and Q(α) has
1, α, . . . , αm−1 as a base over Q. (Clearly m ≤ n.) Let A ∈ Mn (Q) corre-
spond to α in the regular representation of Q(α), with respect to the base
1, α, . . . , αm−1 of Q(α). Let β1 , β2 , . . . , βl be a base of K considered as a
vector space over Q(α). Then β1 , β1 α, . . . , β1 αm−1 , β2 , β2 α, . . . , β2 αm−1 ,
. . . , βl , . . . , βl αm−1 constitute a base of K over Q. (Incidentally l ·m = n
and so m | n.) Let A1 correspond to α in the regular representation of
K with respect to this Q-base. Then
 
A 0 ··· 0
A1 = 0 A · · · 0  ∈ Mn (Q) and Tr (A1 ) = l · Tr (A).

0 0 ··· A
Since α is an algebraic integer, all the elements of A are in Z so that
Tr (A) and Tr (A1 ) = l · Tr (A) = lam−1 are integers. Thus, for an
algebraic integer α ∈ K, Tr K (α) is an integer. Similarly, NK (α) =
det A1 = (det A)l ∈ Z.
The mapping α 7→ Tr K (α) is clearly a Q-linear mapping of K into
Q. We define a bilinear form B(x, y) on the Q-vector space K by setting
B(x, y) = Tr K (xy) for x, y ∈ K.

Proposition 2.2 The bilinear form B(x, y) = Tr K (xy) for x, y ∈ K is

non-degenerate.
54 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.1. ALGEBRAIC NUMBERS AND ALGEBRAIC INTEGERS 39

so that f (γ− ri=2 ni γi ) = 0. This means that γ− ri=2 ni γi = λ1 γ1 , λ1 ∈ Proof: Let x 6= 0 be in K. Then Bx′′ (y) = Tr K (xy) is not identically
P P
R. Since P clearly this is an element of Γ we must have λ1 = n1 ∈ Z, so zero in y, since, for y = x−1 , Bx′ (x−1 ) = Tr K (1) = n. Similarly for y 6= 0
that γ = ri=1 ni γi . in K, By′ (x) is not identically zero in x.
With assertions (a) and (b), Lemma 2.15 is completely proved. If we apply Proposition 1.6 to this bilinear form on V = K, we obtain
the
Theorem 2.4 Let r1 be the number of real conjugates of K, 2r2 the
number of complex conjugates, and let r = r1 + r2 − 1. Then there exist Corollary 2.1 To any Q-base w1 , . . . , wn of K, there corresponds a base
ǫ1 , . . . , ǫr and a root of unity ζ in K such that any unit in K can be w1′ , . . . , wn′ such that Tr K (wi , wj′ ) = δij , 1 ≤ i, j ≤ n.
written in the form
If R is a subset of K and if a ∈ K then, by definition,
ǫ = ζ k ǫk11 · · · ǫkr r , k, k1 , . . . , k ∈ Z.
aR = {ar | r ∈ R}.
The ki , i ≥ q1 are uniquely determined, and k is uniquely determined
modulo w where w is the order of the group Z of roots of unity in K. The following theorem gives more information concerning the struc-
ture of the ring of algebraic integers in a given algebraic number field.
Proof: Let U be the group of units in K. Consider the homomorphism
Theorem 2.1 Let K be an algebraic number field of degree n and O the
f : U → Rr defined by
ring of algebraic integers in K. Then there exists a Q-base ω1 , . . . , ωn of
f (ǫ) = (log |ǫ(1) |, . . . , log |ǫ(r) |), r = r1 + r2 − 1. K such that ωi ∈ O and O = Zω1 + · · · + Zωn .

We assert that (a) the kernel of f is Z and that (b) f (U) = Γ has (Elements ω1 , . . . , ωn with this property are said to form an integral
the property that any bounded subset of Rr contains only finitely many base of O.)
elements of Γ. Proof: Let w1 , . . . , wn be a Q-base of K. By Remark 2.12 on page 36
Proof of (a): If f (ǫ) = 0, then |ǫ(1) | = 1, . . . , |ǫ(r) | =Q
1. This implies there exists m 6= 0 in Z such that mw1 , . . . , mwn are in O. We can thus
that |ǫ(r1 +r2 +1) | = 1, . . . , |ǫr2 +2r2 −1 | = 1. Since further nk=1 |ǫ(k) | = 1, assume without loss of generality that w1 , . . . , wn are already in O. Let
we conclude that |ǫ(r+1) | = 1. The integers α in O for which |α(i) | = w1′ , . . . , wn′ be a base of K for which
1, i = 1, 2, . . . , n, form a finite group, by Lemma 2.9. Hence αk = 1 for
Tr K (wi , Wj′ ) = δij(1 ≤ i, j ≤ n). (2.2)
any such α for some k ∈ Z. Thus ǫk + 1, i.e., ǫ is a root of unity.
Proof of (b): If −M < log |ǫ(i) | < M, i = 1, . . . , r, then e−M < |ǫ(i) | < Pn ′
We know that for any z ∈ O, z = i=1 ai wi with a1 , . . . , an ∈ Q.
eM for i 6= r + 1, n. (since ǫ(ī) = ǫ(i) ). Since, further, n1 |ǫ(k) | = 1, this
Q
Since zwi ∈ O for 1 ≤ i ≤ n, we have, because of (2.2), ai = Tr K (zwi ) ∈
implies that |ǫ(r+1) | < enm , |ǫ(n) | < enm , so that, by Lemma 2.9, the Z. Thus we obtain
number of such ǫ is finite. O ⊂ Zw1′ + · · · + Zwn′ .
By Lemma 2.15, there are units ǫ1 , . . . , ǫt , t ≤ r such that f (U) is
By Proposition 1.4 (Chapter 1) there exist ω1 , . . . , ωm ∈ O, m ≤ n,
generated by f (ǫ1 ), . . . , f (ǫt ) which are independent over R. This means
such that
that if ǫ ∈ U there are uniquely determined integers k1 , . . . , kt so that
O = Zω1 + · · · + Zωm .
ǫ · ǫ−k
1
1
· · · ǫ−k
t
t
∈ Z. Since, by Lemma 2.10, Z is a cyclic group of order
w, the theorem will be proved if we show that t = r. We claim that, necessarily, m = n. In fact, if m < n, then, the Q-
We now prove that t = r. Suppose, if possible, that t < r. Then, subspace of K generated by ω1 , . . . , ωm , which is clearly K itself, would
the subspace V of Rr generated by f (ǫ1 ), . . . , f (ǫt ) has dimension t ≤ have dimension ≤ m < n over Q, contrary to our assumption that K is
r − 1. Hence there are real numbers c1 , . . . , cr not all zero such that if of degree n. Further, we see that ω1 , . . . , ωn are Q-independent, so that,
(x1 , . . . , xr ) ∈ V then c1 x1 + · · · + cr xr = 0; in particular, c1 log |ǫ(1) | + necessarily, the sum above is direct. This proves Theorem 2.1.
40 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.4. THE GROUP OF UNITS 53

Remark 2.14 Any set of elements ω1 , . . . , ωn as above forms a Q-base We have

(k)
of K. We have only to repeat the argument in the last few lines above. |αµ | n < 1 if k ∈ A
|ǫ(k) | = (k)
|αv | > 1 if k ∈ B.
Remark 2.15 Let a be any non-zero ideal in O. Then a ∩ Z 6= {0}. In
fact if α 6= 0 is an algebraic integer in a and if αr +ar−1αr−1 +· · ·+a0 = 0 A subset S of Rm is said to be bounded if for all (x1 , . . . , xm ) ∈ S,
where ai ∈ Z, a0 6= 0, then a0 = −α(a1 + · · · + ar−1) ∈ Z. It follows that we have |xi | ≤ M for a constant M depending only on S.
for any α ∈ K, there is a ∈ Z, a 6= 0 such that aα ∈ a.
Lemma 2.15 Let Γ be an (additive) subgroup of Rm such that any
If ω1 , . . . , ωn are as in Theorem 2.1, then a ⊂ O = Zω1 + · · · + Zωn . bounded subset of Rm contains only finitely many elements of Γ. Then
By Proposition 1.4 (Chapter 1), there exist α1 , . . . , αm ∈ a, m ≤ n, such there exist r ≤ m elements γ1 , . . . , γr of Γ which are linearly independent
that over R and which generate Γ as a group.
a = Zα1 + · · · + Zαm .
As in Remark 2.14,we must have m = n. The αi are said to constitute Proof: Let γ1′ be a non-zero element of Γ. Consider the set A1 of all
an integral base of a. λ ∈ R for which λγ1′ ∈ Γ·A1 contains a smallest positive element µ1 , 0 <
µ1 ≤ 1 since there are, by assumption, only finitely many λ, |λ| ≤ 1, with
P
Further, we may choose the αi so that αi = j≥i pij ωj , pij ∈ Z.
λγ1′ ∈ Γ and γ1′ ∈ Γ. Let γ1 = µ1 γ1′ . If γ = λγ1 ∈ Γ, λ ∈ R, and if n is an
Remark 2.16 As in Remark 2.14, any elements α1 , . . . , αn such that integer with 0 ≤ λ − n < 1 then γ − nγ1 = (λ − n)γ1 = (λ − n)µ1 γ1′ ∈ Γ,
a = Zα1 + · · · + Zαn form a Q-base of K. In particular, any non-zero which is not possible unless λ = n,(by minimality of µ1 ) Hence if λγ1 ∈ Γ
ideal of a contains nPelements which are linearly independent over Q. It for λ ∈ R, then λ ∈ Z. In particular, the lemma is proved if m = 1.
follows that if αi = j≥i pij ωj ,then pii 6= 0. Suppose the lemma already proved for subgroups of Rm−1 , and let
e2 , . . . , em ∈ Rm be such that (γ1 , e2 , . . . em ) form m
P a base of R . Let f :
Remark 2.17 If a 6= {0} is an ideal of O, then by Remark 2.15, there
Rm → Rm−1 be the map defined by f (λ1 γ1 + i≥2 λi ei ) = (λ2 , . . . , λm )
exists 0 6= a ∈ Z such that aO ⊂ a ⊂ O. Now if O = Zω1 + · · · + Zωn ,
which linear over R, and let Γ1 be the image of Γ under f. We claim
then aO = Zaω1 + · · · + Zaωn so that O/aO is of order an . Therefore
that any bounded subset of Rm−1 contains only finitely many elements
O/a is also finite.
of Γ1 . In fact, let γ ′ = (λ2 , . . . , λm )P
∈ Γ1 , |λi | ≤ M, i ≥ 2. Then, there
Remark 2.18 If p is a prime ideal in O then p contains exactly one exists λ1 ∈ R such that γ = λ1 γ1 + i≤2 λi ei ∈ Γ. Since γ1 ∈ Γ, we may
prime number p > 0 of Z. suppose that 0 ≤ λ1 < 1. But then |λi | ≤ M + 1, i > 1, so that there
are only finitely many such γ. This proves our claim.
In fact, let a = p1 · · · pk in p ∩ Z with primes p1 , . . . , pk ∈ Z. Since p By induction, there exist γ2′ , . . . , γr′ ∈ Γ1 , r ≤ m, which are linearly
is prime, at least one pi ∈ p. If p, q are distinct primes in p, we can find independent over R and generate Γ1 . Let γ2 , . . . , γr ∈ Γ be such that
integers x, y with xp + yq = 1; then 1 ∈ p and p = O, a contradiction. f (γi ) = γi′ , i ≥ 2. We assert that γ1 , . . . , γr (a) are linearly independent
over R and (b) generate Γ.
Definition 2.8 For any ideal a 6= {0} of O, the number of elements in Proof of (a):
the residue class ring O/a is called the norm of a and denoted by N (a); If ri=1 λi γi = 0 then ri=1 λi f (γi ) = ri=2 λi γi′ = 0; hence λ2 =
P P P
if a = {0} we set N (a) = 0. · · · = λr = 0, hence λ1 γ1 = 0, hence λ1 = 0.
Proof of (b):
Clearly N (O) = 1. For a proper ideal a of O, N (a) > 1.
If γ ∈ Γ, then f (γ) ∈ Γ1 , hence
Proposition 2.3 Let a P 6= {0} be an ideal of O = Zω1 + · · · + Zωn . r
n
Then there exist αi = j=1 pij ωj , pii > 0, pij ∈ Z such that a =
X
f (γ) = ni γi′ , ni ∈ Z,
Zα1 + · · · + Zαn . Further N (a) = p11 p22 · · · pnn . i=2
52 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.2. UNIQUE FACTORIZATION THEOREM 41

Further Proof: By Remark 2.15 above, a has an integral base α1 , . . . , αn of

Y Y the required form.
(k) (l)
1 ≤ |N (α)| = |α | |α | ≤ cm−1
1 t(m−1)(1−n/m) ln−m |α(k) | We claim that the p11 p22 · · · pnn numbers
k∈A∪Ā l∈B∪B̄
n
= cm−1
1 t−(1−n/m) |α(k) |, X
η = η(x1 , . . . , xn ) = xi ω i , 0 ≤ xi < pii , x ∈ Z
i=1
for any k ∈ A. This proves Lemma 2.12.
In the following two lemmas, A, and B have the same significance form a complete system of residue of O modulo a, In fact, if
as in Lemma 2.12. n
X
ξ= ci ωi ∈ O, ci ∈ Z,
Lemma 2.13 There exists a sequence of non-zero integers αv ∈ O v =
i=1
1, 2, . . . , with
(k) (k)
and we set Oi = O∩(Zωi+1 +· · ·+Zωn), then there exist x1 , 0 ≤ x1 < p11
|α(k) (k)
ν | > |αν+1 |, k ∈ A, |αv | < |αν+1 |, k ∈ B and m1 ∈ Z with
ξ1 = ξ − x1 ω1 − m1 α1 ∈ O1 ;
and
|NK (αν )| ≤ cm
1 . further, x1 , m1 are determined by the relation

Proof: Let tν+1 = M tν , where M is a suitable constant, and let c1 = m1 p11 + x1 .

αν ∈ O satisfy
We can, in the same way, find m2 ∈ Z, 0 ≤ x2 < p22 with
c−m+1
1 t1−n/m
ν ≤ |α(k) 1−n/m
ν | ≤ c1 tν , k ∈ A,
ξ2 = ξ1 − x2 ω2 − m2 α2 ∈ O2 ;
c−mt
1 tν ≤ |α(k)
v | ≤ tν , k ∈ B.
n/m−1 > cm . Then we have c−m+1 t 1−n/m continuing in this way, we find that
Suppose that M > cm 1 , M 1 1 v >
1−n/m (k) (k) n
c1 tν+1 , so that |αν+1 | < |αv |, for k ∈ a, while tv < c−m
1 tν+1 , so that
X
(k) (k) ξ= (mi αi + xi ωi ), mi ∈ Z, 0 ≤ xi < pii ,
|αv+1 | > |αv |,
k ∈ B. It is trivial that i=1
Y Y
|NK (α)| = |α(j)
v | |α(j) m
ν | ≤ c1 · t
m(1−n/m)
· tn−m = cm
1 .
so that the η generate O modulo a.
Now, ifP ni=1 xi ωi =P ni=1 mi αi , where mi ∈ Z, 0 ≤ x < pii , we
P P
i∈A∪Ā j∈B∪M̄
note that i=1 mi αi = ni=1 ci ωi , where c1 = mi p11 . Since the ωi are
n

Lemma 2.14 There exists a unit ǫ with linearly independent, we have x1 = m1 p11 and since 0 ≤ x1 < p11 , we
conclude that m1 = x1 = 0. This implies that c2 = m2 p22 , which in turn
|ǫ(k) | < 1, k ∈ A, |ǫ(k) | > 1, k ∈ B.
implies that x2 = m2 = 0, and so on; hence xi = 0. This proves that the
η’s are distinct modulo a and with it, the proposition.
Proof: Let {αν } be a sequence of integers as in Lemma 2.13, and let
aν be the principal ideal (αν ). Then by Lemma 2.4 and 2.13 , N (aν ) =
|NK (αv )| ≤ cm
1 . Hence, there exist v, µ, v < µ with av = aµ (since, 2.2 Unique Factorization Theorem
by Lemma 2.7 the number of integral ideals of norm ≤ const. is finite).
This means that Let K be an algebraic number field of degree n and O the ring algebraic
αν = ǫαν , ǫ a unit. integers in K . Let p be a prime ideal in O. Then O/p is finite and indeed
42 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.4. THE GROUP OF UNITS 51

a commutative integral domain with 1, by Remark 1.18 on page 20. By Lemma 2.12 Let A and B be two nonempty subsets of E with A ∩ B =
Example 1.13, page 10, O/p is a field. Thus: ∅, A ∪ B = E. Let m be the number of elements in A ∪ Ā. Then, there
(D1 ) Every prime ideal of O is maximal. exists a constant c1 depending only on K such that for any integer t > t0 ,
We say that an element α ∈ C is integral over O if there exists a there exists α ∈ O for which
monic polynomial f with coefficients in O such that f (α) = 0. As in
Proposition 2.1, one can prove that α ∈ C is integral over O if and c−m+1
1 t1−n/m ≤ |α(k) | ≤ c1 t1−n/m , k ∈ A,
only if there is a non-zero finitely generated O-module M ⊂ C with (∗)
αM ⊂ M (alternatively, if and only if O[α] is finitely generated over c−m (k)
1 t ≤ |α | ≤ t, k ∈ B
O.) By Theorem 2.1, such a module M is finitely generated over Z, and
consequently an element α ∈ C integral over O is an algebraic integer. Proof: Let ω1 , . . . , ωn be P a set of n integers of K which P are indepen-
It follows at once that, if K is an algebraic number field and O the dent over Q. Then, if α = xj ωj , we have α(k) = ni=1 xj ω (k) . Let
ring of algebraic integers in K, then any α ∈ K which is integral over O k1 , . . . , ku be the elements of A with k̄i = ki , l1 , . . . , lv those with l¯i 6= li .
(l )
belongs to O. If we define, for any integral domain R the integral closure Then .m = u + 2v. We set aij = ω (ki ) , i ≤ u, au+i,j = Re ωj i , 1 ≤ i ≤
(l )
of R in its quotient field as the set of all elements of the quotient field v, au+v+i,j = Im ωj i , 1 ≤ i ≤ v. By Lemma 2.11, it follows that there
of R which are roots of monic polynomials with coefficients in R, we can are integers xj , not all zero, |xj | ≤ t, with
therefore assert the following:
n
(D2 ) The integral closure of O in K is O itself. X
To each ideal a 6= {0}, associate its norm N (a) > 0 in Z. The map- |α(k) | ≤ 2ct1−n/m , α = xj ωj .
j=1
ping N : a 7→ N (a) is, in general, not one-one. However, if a ⊂ b and
a 6= b, then N (a) > N (b). [For, let f : (O/a, +) → (O/b, +) be the map Thus, there exists c′ > 0, such that, for t > 1, there is α ∈ O, α 6= 0
defined by f (x + a) = x + b. Then f is well-defined, onto (O/b, +) but with
is not one-one since there exists y ∈ b, y 6⊂ a. We deduce at once the |α(k) | ≤ c′ t1−n/m , k ∈ A, |α(l) | ≤ c′ t for all l.
following statements.
(N1) If a1 ⊂ a2 ⊂ · · · ⊂ an ⊂ an+1 ⊂ · · · is an increasing sequence Replacing t by t/c′ , we see that for t > t0 (= c′ ), there is an integer
of ideals in O then am = am+1 for m ≥ m0 ∈ Z+ . α ∈ O, α 6= 0 with
(N2) Any non-empty set S of ideals in O contains a maximal ele-
ment, i.e. an ideal a such that a 6⊂ b for any b ∈ S, b 6= a. (For, any set |α(k) | ≤ c1 t1−n/m , k ∈ A, |α(l) | ≤ t for all l.
of positive integers contains a least element.)
(N3) Any ideal a in O with a 6= O is contained in a maximal ideal Note that |α(k) | ≤ c1 t1−n/m for k ∈ Ā, since α(k̄) = α(k) .
of O. We assert that we have the inequalities (∗) for this α. In fact, since
(Take for S in (N2) the set of all proper ideals b with a ⊂ b ⊂ O.) α 6= 0, NK (α) is a rational integer 6= 0. Using (1.1) we have
n
Remark 2.19 A ring R is noetherian if the statement (N 1) is true of Y Y Y
ideals in R. One can show that R is noetherian if and only if statement I ≤ |NK (α) = |α(l) = |α(k) | |α(l) |
l=1 l∈A∪Ā l∈B∪B̄
(N 2) is true of ideals in R. This is further equivalent to the statement
that any ideal of R is finitely generated (as an R-module). Statement ≤ cm
1 t
m(1−n/m
· |α(l) | · tn−m−1 = cm −1 (l)
1 t |α |.
(N 3) is true in arbitrary (even non-commutative) rings with unity, and,
in this generality, is due to Krull. for any l ∈ B (since there are m elements in A ∪ Ā and n − m in B ∪ B̄);
i.e.
We have thus proved that |α(l) | ≥ c−m
1 t, l ∈ B.
50 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.2. UNIQUE FACTORIZATION THEOREM 43

Definition 2.12 A complex number α is called a root of unity if αm = 1 (D3 ) O is noetherian.

for some m ∈ Z, m 6= 0. If ρ is a root of unity in K, then ρm = 1 for Any commutative integral domain R satisfying the conditions (D1 ),
some m ∈ Z, m 6= 0, so that |ρ(i) | = 1 for i = 1, 2, . . . , n. (D2 ), (D3 ) is known as a Dedekind domain and these conditions are
themselves known as axioms of classical ideal-theory. The ring of al-
Every root of unity in K in is a unit but not conversely. For example, gebraic integers in an algebraic number field is therefore a Dedekind
√ √
in K = Q( 2), 1 + 2 is a unit but not a root of unity. domain.
Taking C = 1, we deduce from Lemma 2.9 the following.

Corollary 2.5 The number of roots of unity in K is finite. Definition 2.9 By a fractional ideal in K, we mean an O submodule a
of K for which there exists m 6= 0 in Z such that ma ⊂ O.
Let Z be the group of roots of unity in K, and let ζk = e2πipk /qk , k =
1, . . . , w be the elements of Z. Let q0 = q1 · · · qw and let A be the sub- Any ideal in O is trivially a fractional ideal. By analogy with Z we
group of Z consisting of integers p for which e2πip/q0 ∈ Z. A is of the may call an ideal in O an integral ideal in K. Any fractional ideal a can be
form p0 Z, p0 > 0. Clearly Z is generated by ζ = e2πipo /qo . Thus we have written as a−1 b for a 6= 0 in Z and an integral ideal b. If c is an integral
Lemma 2.10 The roots of unity in K form a finite cyclic group. ideal, then for any b 6= 0 in Z, b−1 c is clearly a fractional ideal in K. If
c, d, are fractional ideals in K, then for a suitable c ∈ Z, c 6= 0, cc, cb are
We denote the order of this group by w. both integral ideals and the sum c + d = c−1 (cc + cd) and the product
cd = c−2 (cc · cd) are again fractional ideals in K.
Lemma 2.11 Let m and n be positive integers, with 0 < m < n. Let We now prove the important
(aij , 1 ≤ i ≤ m, 1 ≤ j ≤ n be real numbers. Then, for any integer
t > 1, there exist integers P x1 , . . . , xn not all zero, |xj | ≤ t, such that
Theorem 2.2 (Dedekind.) Any proper ideal of the ring O of algebraic
|yi | ≤ ct1−n/m , where yi = nj=1 aij xj and c is constant depending only
integers in an algebraic number field K can be written as the product of
on (aij ).
prime ideals in O determined uniquely upto order.
Proof: Let a = maxi nj=1 |aij |. Then, for |xj | ≤ t, we have |yi | ≤ at.
P

Consider the cube −at ≤ yi ≤ at, i = 1, . . . , m in Rm , and divide it For the proof of the theorem, we need two lemmas.
into hm smaller cubes of side 2at/h(h being an integer≥ 1.) If we assign
to the xj the values 0, 1, . . . , t, the (t + 1)n points (y1 , . . . , ym ) lie in Lemma 2.2 Any proper ideal a ⊂ O contains a product of prime ideals
the big cube, so that, if hm < (t + 1)n , at least two of them lie in the in O.
same cube of side 2at/h; let these points correspond to (x′1 , . . . , x′n ) and
(x′′1 , . . . , x′′n ); 0 ≤ x′j , x′′j ≤ t. If xj = x′j − x′′j , then |yi | ≤ 2at/h, and Proof: Let S be the set of proper integral ideals not containing a
|xj | ≤ t, not all the xj are zero. Now, since t > 1, n/m > 1, we have product of prime ideals. If s 6= ∅, then by statement (N2), S contains
(t + 1)n/m − tn/m > 1; hence, there is an integer h with a maximal element, say a. Clearly a cannot be prime. Thus there exist
x1 , x2 ∈ O, x1 x2 ∈ a but x1 , x2 ∈ / a. Let ai (i = 1, 2) be the ideal
tn/m < h < [(t + 1)]n/m .
generated by a and xi . Then a1 and a2 contain a properly. By the
Hence, there exist integers xj , not all zero, |xj | ≤ t, with |yi | < 2at1−n/m . maximality of a in s, a1 ∈ / S, a2 ∈ / S. Hence a1 ⊃ p1 p2 · pr and a2 ⊃
Let K = K (1) , . . . , K (r1 ) be the real conjugates of K, and let K r1 +1 , q1 · · · qs where p1 , . . . pr , q1 , . . . , qs are prime ideals in O. Since a1 a2 ⊂ a,
. . . , K r1 +r2 be the distinct complex conjugates of K, and let K (r1 +r2 +i) = we have p1 · · · pr q1 · · · q3 ⊂ a giving us a contradiction. Hence S = ∅.
K̄ r1 +i) , 1 ≤ i ≤ r2 . Consider the set E of integers 1 ≤ k ≤ r1 + r2 ; for
any k ∈ E, we set k̄ = k if k ≤ r1 , and k̄ = k + r2 if r1 < k ≤ r2 , and, Lemma 2.3 Any prime ideal p ⊂ O is invertible, i.e. there exists a
for any subset A of E, we set Ā = {k̄ | k ∈ A}. fractional ideal p−1 such that pp−1 = O.
44 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.4. THE GROUP OF UNITS 49

Proof: Let p−1 be the set of x ∈ K such that xp ⊂ O. Clearly p−1 is Let O be the ring of the algebraic integers in K.
an O-module containing O. Since p contains n 6= 0 in Z (see Remark 2.15
on page 40), we have np−1 ⊂ p−1 p ⊂ O. Hence p−1 is a fractional ideal. Definition 2.11 A non-zero element α in O is called a unit of K if
Now p ⊂ pp−1 ⊂ O. Since p is maximal, either pp−1 = O in which case α−1 ∈ O.
our lemma will be proved or p = pp−1 .
Clearly the units of K form a subgroup U of K ∗ .
If p = pp−1 , then every x ∈ p−1 satisfies xp ⊂ p. Since p is a finitely
We observe that if α ∈ O is a unit then NK (α) = ±1. For if α is a
generated Z-module (vide Remark 2.15 on page 40), we see that x ∈ O
unit, there is β ∈ O with αβ = 1. Thus 1 = NK (αβ) = NK (α)NK (β).
in view of Proposition 2.1. Hence p−1 ⊂ O i.e. p−1 = O. This, as we
Since both NK (α) and NK (β) are in Z, we must have NK (α) = ±1.
now show, is impossible. Take x ∈ p such that xO 6= O. Then xO is a
[The converse is also true: if α ∈ O satisfies NK (α) = ±1 then α is a
proper integral ideal and by Lemma 2.2, p1 · · · pr ⊂ xO for prime ideals
unit, as follows from the fact that the norm of α is the product of the
p1 , . . . , pr . Assume r so chosen that xO does not contain a product of
conjugates of α.]
r − 1 prime ideals in O. Now p ⊃ xO ⊃ p1 · · · pr . By Remark 1.19 on
√ √
page 21 p divides one of p1 , . . . , pr , say p1 . But by Property (D1 ), p = Example 2.3 Let K = Q( 5). Then 1± 5
are units in K.
2
p1 . Now p2 · · · pr , is not contained in xO, by minimality of r. Hence
there exists b ∈ p2 · · · pr , b ∈ / xO, i.e. x−1 b ∈ / O. But since bx−1 p ⊂ Lemma 2.9 The number of integers α ∈ O such that |α(i) | ≤ C for
p2 · · · pr (x O)p = x O·p1 · · · pr ⊂ x O·xO = O, we have bx−1 ∈ p−1 .
−1 −1 −1
i = 1, 2, . . . , n, is finite.
But x−1 b ∈ O, i.e. p−1 6= O. Thus pp−1 = O.
We may now give the following Proof: Let ω1 , . . . , ωn be an integral base of O. Then, any α ∈ O can
Proof of Theorem 2.2. As in Proposition 1.7, the proof is split be written
into two parts. α = x1 ω1 + · · · + αn ωn , xi ∈ Z.
(i) Existence of a factorization. Let S be the set of proper ideals of O We then have
which cannot be factorized into prime ideals. We have to show that (i)
S = ∅. Suppose then that S 6= ∅. Then by (N2) S contains a maximal α(i) = x1 ω1 + · · · + xn ωn(i) , i = 1, 2, . . . , n;
element say a ⊂ O. Now obviously, cannot be prime. Hence by (N3)
this can be written
a ⊂ p, a 6= p where p is prime. By Lemma 2.3, there exists an ideal p−1
A = ΩX
such that pp−1 = O. Thus ap−1 is a proper ideal in O and contains a
α(1) x1
   
properly since p−1 contains O properly. But if ap−1 = p1 · · · pr for prime . .
ideals p1 , . . . , pr then a = pp1 · · · pr contradicting the assumption that where A is the column  ..  , X the column  ..  and Ω the matrix
α (n)
a ∈ S. Hence ap−1 ∈ S but this contradicts the maximality of a in S. xn
(k)
Thus S = ∅, i.e. every proper ideal of O can be factorized into prime (ωj ). Since Ω has an inverse Ω−1 in Mn (C), this gives
ideals.
(ii) Uniqueness of factorization. If possible, let a proper ideal a in O have X = Ω−1 A.
two factorizations a = p1 · · · pr = q1 · · · qr where p1 , . . . , pr , q1 , . . . , qs , are
By assumption, |α(i) | ≤ C. This clearly implies that
prime ideals. This means that q1 divides p1 · · · pr and by Remark 1.19, q1
divides one of the ideals p1 , . . . , pr say p1 . But since p1 is maximal,q1 = |xi | ≤ M C
p1 . Now by Lemma 2.3, q−1 −1
1 a = q1 q1 · · · qr = q2 · · · qr and q1 a =
−1
−1
p1 p1 · · · pr = p2 · · · pr . Thus p2 · · · pr = q2 · · · qs . By repeating the ar- where M depends only on the matrix Ω−1 , thus only on K. Since the
gument above, q2 is equal to one of the prime ideals p2 , . . . , pr , say p2 . number of rational integers (xi ) satisfying |xi | ≤ M C is finite the lemma
In finitely many steps, we can thus prove that r = s and that p1 , . . . , pr follows.
48 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.3. THE CLASS GROUP OF K 45

at most n prime ideals occurring in the factorization of pO into prime coincide with q1 , . . . , qr upto order.
ideals since (pO = qa11 · · · qas s , qi prime) implies by Lemma 2.6 and the
corollary to Lemma 2.4, that pn = N (pO) = N (q1 )a1 · · · N (qs )as leading Corollary 2.2 Any fractional ideal a can be written uniquely in the
to s ≤ n. Since p ≤ x, the lemma is completely proved. form
q1 · · · qr  1 
a= stands for p−1
Lemma 2.8 There exists a constant C depending only on K such that p1 · · · pr p
every integral ideala 6= {0} contains α 6= 0 with |NK (α)| ≤ CN (a). where the qi , pj are prime; and no qi is a pj .
Proof: By Theorem 2.2, O = Zω1 + · · · + Zωn . Let t denote the This follows immediately if we choose c 6= 0, c ∈ Z with b = ca ⊂ O,
largest integer ≤ (N (a))1/n . Then among the (t + 1)n numbers ni=1 ai ωi
P
and write (c) = p1 · · · pr , b = q1 · · · qs′ and ‘cancel’ equal qi and pj in
with ai ∈ Z, 0 ≤ ai ≤ t, i = 1, 2, . . . , n, there should exist at least pairs.
two distinct numbers whose difference is in a (since N (a) = order of
O/a < (t + 1)n ). Thus there exists α = a1 ω1 + · · · + an ωn 6= 0 in a with Corollary 2.3 Given any fractional ideal a 6= {0} in K, there exists a
ai ∈ Z, |ai | ≤ t, i = 1, 2, . . . , n. Let Ai be the n-rowed square matrix fractional ideal a−1 such that aa−1 = O.
with elements in Z, corresponding to ωi under the regular representation
with respect to the For proving this, it suffice to show that every integral ideal is in-
Pbase ω1 , . . . , ωn of K over Q. Hence to α corresponds vertible. But this is an immediate consequence of Theorem 2.2 and
the matrix A = N i=1 ai Ai all of whose elements are integers ≤ tµ (in
absolute value) where µ = µ(A1 , . . . , An ) depends only on ω1 , . . . , ωn i.e. Lemma 2.3.
only on K. It is now immediate that |NK (α)| = | det A| ≤ C·tn ≤ C·N (a)
Remark 2.20 Let a = pa11 · · · par r , b = pb11 · · · pbrr be integral ideals
for a constant C depending only on K.
p1 , . . . , pr being prime ideals a1 , . . . , ar , b1 , . . . , br ∈ Z+ . (We define p0i =
Remark 2.22 The constant C obtained in Lemma 2.8 is not the best O, i = 1, 2, . . . , r.) The greatest common divisor (a, b) of a and b is de-
possible. By using a theorem due to Minkowski, one can obtain a better fined to be pc11 · · · pcrr where ci = min(ai , bi ) i = 1, 2, . . . , r. Clearly ci is
constant. the largest integer c such that pci divides both a and b. But now if c, d are
integral ideals then c divides d if and only if d = cc1 for c1 ⊂ O. (For,
We have now all the material necessary to prove. if c ⊃ d, then c − i = dc−1 ⊂ O and conversely if d = cc1 with c1 ⊂ O
then c ⊃ d.) Thus the greatest common divisor of a and b is actually the
Theorem 2.3 The class number of K is finite.
smallest ideal dividing a and b which is nothing but a + b. For any set
Proof: We shall prove that in every class of ideals, there exists an of s ideals a1 , . . . , as the greatest common divisor of a1 , . . . , as may be
integral ideal of norm ≤ C where C is a constant depending only on K. similarly defined.
By Lemma 2.7, the number of ideal classes, i.e. h, will then be finite.
Let R be an ideal class. Take an ideal a in the inverse class R−1 (and we
can assume a to be integral without loss of generality). By Lemma 2.8,
2.3 The class group of K
there exists α ∈ a such that |N (α)| ≤ C ·N (a) for a constant C = C(K). Let K be an algebraic number field of degree n. By Corollary 2.3 to
But now b = (α)a−1 ∈ R and N (a)N (b) = N (ab) = |NK (α)| ≤ CN (a), Theorem 2.2, the non-zero fractional ideals in K form a multiplicative
which gives us N (b) ≤ C. Theorem 2.3 is completely proved. group which we denote by ∆. The ring O of algebraic integers is the
identity element of ∆.
2.4 The group of units A fractional ideal in K is said to be principal if it is of the form αO
with α ∈ K. Clearly the principal (fractional) ideals a 6= {0} form a
Let K be an algebraic number field of degree n and let K (1) (= K), K (2) , subgroup Π of ∆. The quotient group h = ∆/Π is called the group of
. . . , K (n) be all the conjugates of K. ideal classes in K or briefly the class group of K.
46 CHAPTER 2. ALGEBRAIC NUMBER FIELDS 2.3. THE CLASS GROUP OF K 47

Definition 2.10 The ideals a, b 6= {0} in K are in the same ideal class, Proof: Let a = pa11 · · · par r , b = pb11 · · · pbrr , ai , bi ∈ Z+ and p1 , . . . , pr
or are equivalent if and only if a = (α)b for some α ∈ K. are all the prime ideals dividing a and b. We can find an element πi ∈
ai−1 +1 ai ai+1 +1
pa11 +1 · · · pi−1 pi pi+1 · · · par r +1 but πi ∈/ pa11 +1 · · · pai i +1 · · · pan +1 −
We shall prove in this section that h is a finite group. The order n(since pi 6= O). Take ω = π1 + · · · + πn . Clearly pai i +1 divides πj (for
denoted by h ( = h(K)) is called the class number of K. j 6= i) and pai i is the highest power of pi dividing πi . Hence Q pi and no
ai

If h = 1, then O is a principal ideal domain. higher power of pi divides ω. Consequently (ab, (ω)) = ri=1 pai i = a.
For proving that h is finite, we need a few lemmas.
Let a be an integral ideal in K and let Remark 2.21 Given any integral ideal a there exists t 6= 0 in Z such
that b = ta−1 ⊂ O, i.e. ab = tO. By Lemma 2.5, a = (tO, ωO) = tO+ωO
a = Zα1 + · · · + Zαn = Zβ1 + · · · + Zβn with α1 , . . . , αn , β1 , . . . , βn ∈ a. by the remark on page 45. In other words, any integral ideal can be
Pn Pn generated, over O by two algebraic integers in K.
Clearly αi = j=1 qij βj , qij ∈ Z and βi = j=1 rij αj , rij ∈ Z
for i = 1, 2, . . . , n. Denoting the n-rowed square matrices (qijP),n (rij ) by Lemma 2.6 Let a and b be integral ideals. Then N (ab) = N (a)N (b).
Q, R respectively, we see that if QR = (sij ), then αi = j=1 sij αj .
By Remark 2.16 after Theorem 2.1, α1 , . . . , αn are linearly independent Proof: Let λ = N (a) and µ = N (b). Let ξ1 , . . . , ξλ and η1 , . . . , ηµ
over Q. Thus sij = δij (the Kronecker delta) for 1 ≤ i, j ≤ n i.e. be a complete set of representative of O modulo a and of O modulo b
QR = In . Taking determinants, it follows that det Q · det R = 1 and respectively. By Lemma 2.5, there is ω ∈ O such that (ab, (ω)) = a.
since det Q · det R ∈ Z, we have det Q = det R = ±1. We claim that the λµ elements ξi + ωηj (i = 1, 2, . . . , λ, j = 1, 2, . . . , η)
We now use the foregoing remarks to identity the norm N ((α)) of form a complete set of representatives of O modulo ab.(and this will
the principal ideal (α) generated by α 6= 0 in O with the absolute value prove the lemma). First, they are all distinct modulo ab. For, if ξi +
|NK (α)| of the norm NK (α). ωηj ≡ ξk + ωηl (mod ab), then ξi − ξk ∈ a and therefore i = k But
then ω(ηj − ηi ) ∈ ab and since (ab, (ω)) = a, it follows by Theorem 2.2
Lemma 2.4 For α 6= 0 in O, N ((α)) = |NK (α)|. that ηj − ηl ∈ b i.e. j = l. Given any x ∈ O, there exists a unique
ξi (1 ≤ i ≤ λ) such that x − ξi ∈ a. Now a = (ab, (ω)) = ab + (ω). Hence
Proof: Pn If O = Zω1 + · · · = Zωn , then by Proposition 2.3, there exist x − ξi = ωη + y with y ∈ ab. This gives us x − ξi ≡ ωηj (mod ab) for some
βi = j=1 pji ωj , i = 1, 2, . . . , n; pji ∈ Z, p11 , . . . , pnn > 0 such that ηj (1 ≤ j ≤ µ), since ω ∈ a. Thus for any x ∈ O, x ≡ ξi + ωηj (mod ab)
(α) = Zβ1 + · · · + Zβn and N ((α)) = p11 p22 · · · pnn . Let Q stand for the for some i and j and we are through.
n-rowed square matrix (qij ) where qij = 0 for 1 ≤ i < j ≤ n and qij = pij
otherwise. Since (α) = Zαω1 + · · · + Zαωn as well, we P have in view of Lemma 2.7 For any integer x > 0, the number of ideals a ⊂ O for
the remark immediately preceding this lemma, αωi = nj=1 rji βj , i = which N (a) ≤ x is finite.
1, . . . , n, and if R denotes the n-rowed square matrix (rij ), then det R =
±1. Let S be the matrix QR. Taking the regular representation with Proof: Let a = pλ1 1 · · · pλr r be any integral ideal with N (a) ≤ x,
respect to the base ω1 , . . . , ωn , we have NK (α) = det S. But (here p1 , . . . , pr are prime and λ1 , . . . , λr > 0 in Z). By Lemma 2.6,
(N (p1 ))λ1 · · · (N (pr ))λr ≤ x. Since N (pi ) ≥ 2, we have 2λi ≤ N (pi )λi ≤
det S = det Q · det R = ± det Q = ±p11 · · · pnn = ±N ((α)). N (p1 )λ1 · · · N (pr )λr ≤ x. The number of λi ’s is finite. To prove the
lemma, it therefore suffices to prove that the number of prime ideals in
Corollary 2.4 For t ∈ Z, N (tO) = |N (t)| = |tn |. O of norm ≤ x is finite. Now any prime ideal p contains exactly one
prime p ∈ Z (Remark 2.18) and hence p occurs in the factorization of
Lemma 2.5 For any two integral ideals a, b there exists ω ∈ O, such pO into prime ideals. Moreover, N (p) | (N (p)) = pn , so that, since
that gcd(ab, (ω)) is a. N (p) 6= 1,we have N (p) = pf , f ≥ 1, and p ≤ N (p) ≤ x. But there are