603727

ABC CONJECTURE AND ITS IMPLICATIONS
by
Tuğçe Koç
B.S., Mathematics, Boğaziçi University, 2016
Submitted to the Institute for Graduate Studies in

Science and Engineering in partial fulfillment of
the requirements for the degree of
Master of Science
Graduate Program in Mathematics

Boğaziçi University
2019
ii
APPROVED BY:
Assoc. Prof. Ekin Özman ...................

(Thesis Supervisor)
Prof. K. Ilhan Ikeda ...................
Assist. Prof. Ayberk Zeytin ...................
DATE OF APPROVAL: 21.06.2019

iii
ACKNOWLEDGEMENTS
First of all, I would like to thank my advisor, Ekin Özman, for the continuous
support throughout my studies, for her patience, immense knowledge, and advice . Her
guidance helped me in all the time of research and writing of this thesis. I could not
have imagined having a better advisor and mentor.
I, especially, want to thank my best friend Ezgi Kan for her endless energy,
encouragement, and for the days she had been dragged to many places for me to study,
and, of course, for the tech tips. I would also like to thank my great friend Merve
Taşan for her positivity and moral support during the hard times. I could not have
done it if you two were not in my life.
I am deeply thankful for my lovely family, Hilal, Kadir and Burak, for always
being there for me. This thesis stands as a testament to your unconditional love and
encouragement.
My great friend Ayçin İplikçi deserves special thanks for her help and colleague-
ship.
I thankfully acknowledge the financial support given by TÜBİTAK PIA Bospho-

rus project 117F274 “Sonlu Cisimler Üzerinde Tanımlı Eğriler, Jakobyen Varyeteleri
Ve Abelyen Varyeteler” (PI Alp Bassa). Finally, I would like to thank İlhan İkeda and
Ayberk Zeytin for serving on my thesis defense committee.
iv
ABSTRACT
In this thesis, our aim is to state the importance of abc conjecture and prove the
strong results we obtain with the help of abc conjecture. First, we give necessary notions
and tools which are used throughout the thesis. Then we introduce Hall conjecture,
Fermat’s last theorem and Mordell conjecture, and their relations with abc conjecture.
In particular, we give the effective proof of Mordell conjecture using abc conjecture,
given in the article of Noam Elkies, [1], and also get another height bound by combining
with a different theorem. Finally, we give three examples where we use both of the
height bounds. This thesis was supported by TÜBİTAK Project 117F274.
v
ÖZET
ABC VARSAYIMI VE ONUN ÇIKARIMLARI
Bu tezde, amacımız abc varsayımının önemini anlatmak ve onu kullanarak birçok

güçlü ifadeyi kanıtlayabileceğimizi göstermektir. İlk olarak, tez boyunca kullanacağımız
bazı kavramlar ve araçlardan bahsediyoruz. Daha sonra Hall varsayımı, Fermat’ın
son teoremi, Mordell varsayımı ifadelerini tanıtıp abc varsayımından çıkan ispatlarını
veriyoruz. Özellikle, Noam Elkies’in makalesinde kanıtladığı [1], abc varsayımından
çıkan Mordell varsayımının efektif kanıtı üzerinde duruyoruz. Aynı zamanda, bu kanıtı
başka bir teoremle birlikte kullanarak farklı bir yükseklik sınırı elde ediyoruz. Sonunda
da bu iki farklı yükseklik sınırını kullanarak örnekler veriyoruz. Bu tez TÜBİTAK
117F274 projesi tarafından desteklenmiştir.
vi
TABLE OF CONTENTS
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
ÖZET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
LIST OF SYMBOLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
2. PRELIMINARIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1. Absolute Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2. Prime Ideals and Factorization . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2. Points and Maximal Ideals . . . . . . . . . . . . . . . . . . . . . 8
2.2.3. Morphisms of Curves . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.4. Factorization of Ideals . . . . . . . . . . . . . . . . . . . . . . . 11
2.3. Algebraic Curves Basics . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1. Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.2. Differentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.3. Riemann-Roch Theorem . . . . . . . . . . . . . . . . . . . . . . 24
2.3.4. Rational Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.5. Primes of Good Reduction . . . . . . . . . . . . . . . . . . . . . 29
2.4. Elliptic Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5. Height Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5.1. Heights on Projective Space . . . . . . . . . . . . . . . . . . . . 38
2.5.2. Heights on Curves . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.6. Belyi’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3. ABC CONJECTURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1. The Origin of Abc: Mason-Stothers Theorem . . . . . . . . . . . . . . 57
4. HALL CONJECTURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1. Abc Implies Weaker Hall . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2. Abc Implies Fermat for Large Exponents . . . . . . . . . . . . . . . . . 65
vii
5. MORDELL CONJECTURE . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6. ABC IMPLIES MORDELL . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.1. Adaptation and Examples . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2. Motivation for the Proof . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3. The Proof: Abc Implies Mordell . . . . . . . . . . . . . . . . . . . . . . 92
6.4. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
viii
LIST OF FIGURES
Figure 4.1. An example of Mordell curve . . . . . . . . . . . . . . . . . . . . . 61

ix
LIST OF SYMBOLS
An affine n space
An (K) the set of K rational points of An
bf (x) branch number of f in x ∈ C
C algebraic curve
Cf = K[x, y]/hf i, ring of continuous functions on the curve
Zf (K)
C(K) the set of K rational points of C
C(Q) the points on C with rational coordinates
C(Zf (K), K) the set of continuous functions from Zf (K) to K
D discriminant of a polynomial
|D| complete linear system of the divisor D
D1 ≥ D2 indication that D1 − D2 is a positive divisor
deg(D) degree of a divisor
deg(φ) degree of the map φ
discZ (OK ) discriminant of the ring of integers
Div(C) divisor group of a curve
Div 0 (C) group of divisors of degree 0 on a curve
DivK (C) group of divisors defined over K
div(f ) divisor of the function f
div0 (f ) divisor of zeros of f
div∞ (f ) divisor of poles of f
div(w) divisor of the differential form w
eφ (P ) ramification index of the map φ
Ê reduction of the elliptic curve E modulo p
E(K) group of K-rational points on the elliptic curve E/K
Etors torsion subgroup of the elliptic curve E
g genus
GK the absolute Galois group Gal(K/K)
GQ Galois group on Pn (Q)
x
Gal(K/K) the Galois group of K/K

h absolute logarithmic height
H absolute multiplicative height
hf height relative to a morphism f
H(P ) height of a point in Pn (Q)
HK (P ) relative height of a point in Pn (Q)
hV,D height on the variety V relative to the divisor D
H(α) abbreviation for H((x, 1))
K number field
K∗ the unit group of K
K an algebraic closure of K
KC canonical divisor on a curve C
K(C) function field of C/K
K[C]P local ring of C at P
K(P ) minimal field of definition of the point P
K[X] the polynomial ring K[X1 , . . . , Xn ]
K(Zf ) the field of fractions of the ring Cf
L(D) space of functions satisfying div(f ) ≥ −D
l(D) dimension of the space of functions L(D)
L(E, s) L-function of E
Lp (T ) local factor at p of the L-series
MK a complete set of inequivalent absolute values on K
MK ∞ the archimedean absolute values in K
MK 0 the nonarchimedean absolute values in K
MP ideal associated to the point P
M ax(Cf ) the set of maximal ideals of Cf
N norm
n0 (f ) number of distinct roots of f
Np number of points in the reduction of the curve modulo p
O the identity element on an elliptic curve
O(1) a bounded function
xi
OK the ring of integers of K

ordp the p-adic order of a rational number
ordP valuation attached to the prime ideal
ordP normalized valuation on K[C]P and on K(C)
p rational prime
P prime ideal
PP prime rational divisor
Pσ the action of σ ∈ Gal(K/K) on the point P
Pn projective n space
Pn (K) set of K rational points of Pn
P ic(C) the Picard group of C
P ic0 (C) Picard group of degree-zero divisor classes
RE rank of an elliptic curve E
Rad(n) product of distinct prime factors of n
rankan (E/Q) analytic rank of E/Q
Spec(A) the set of prime ideals of a ring A
Supp(D) support of the divisor D
V projective variety
v(x) = − log |x|v , for an absolute value v ∈ MK
w invariant differential on an elliptic curve
|x|p p-adic absolute value
|x|P P-adic absolute value associated to a prime ideal
|x|φ absolute value associated to an embedding in C
|x|∞ archimedean absolute values
Zf (K) the set of points of f with coordinates in K
∆E discriminant of an elliptic curve E

∆0E minimal discriminant of an elliptic curve E
φ∗ map of function fields induced by rational map of curves
φ∗ map of function fields induced by rational map of curves
φ∗ map induced on divisors by the rational map φ
xii
φD rational map associated to the divisor D

φD,F rational map associated to the divisor D and a basis F of
L(D)
ΩC the space of meromorphic differential forms on the curve C
⊕ group law on an elliptic curve

∼ linear equivalence of divisors
xiii
1. INTRODUCTION
In any sum of three relatively prime integers, it is not possible for all three terms
to be divisible by many high powers of primes. That is the basic intuition behind
the famous abc conjecture in number theory. It seems like a simple conjecture about
the sum of integers, yet nobody has been able to prove it. Goldfeld described abc
conjecture as “the most important unsolved problem in Diophantine analysis.” [2] It is
first proposed by Joseph Osterlee and David Masser in the 1980s.
The importance of abc conjecture mainly stems from two reasons. First, the
conjecture investigates the relation between two arithmetic operations, addition and
multiplication. Divisibility and primes are multiplicative notions. Abc conjecture sets
a subtle limit on how these notions can interact with addition. Since number the-
ory is mostly built over the relation between addition and multiplication, it is useful
to understand such constraints. Secondly, abc conjecture has really important conse-
quences such as “eventual” Fermat’s last theorem, Catalan’s conjecture for any large
parameters, Roth’s theorem, Mordell conjecture and Hall conjecture.
As it is not provable till the current day, it is also not disprovable. The exponent
1 + makes the conjecture very powerful, and thus hard to find a counterexample.
Abc conjecture got some media attention in 2012 when Professor Shinichi Mochizuki
published a possible proof for the conjecture, which is 500 pages long based on per-
haps another 1000 pages of previous technical papers. After 7 years of struggle, al-
most no one could completely understand his proof let alone approve, and its reason
is that Mochizuki developed a whole new type of mathematics called Inter-universal
Teichmüller theory. So mathematicians who want to read his proof first need to un-
derstand his new theory. Workshops and conferences were held on Mochizuki’s work
on abc conjecture. Brian Conrad, who is a math professor at Stanford and was one
of the participants at the Oxford workshop, summarizes the issue on his notes from
the workshop as: “The method as currently formulated by Mochizuki does not yield
explicit constants, so it cannot be used to establish an effective proof of the Mordell
1
Conjecture. But if correct it would nonetheless be a tremendous achievement, setting

many difficult open problems, and would yield a new proof of the Mordell conjecture
(as shown long ago by Noam Elkies).” [3]
In this thesis, we first give two equivalent statements of abc conjecture and a few
examples. Then, we state and give the proof of Mason-Stothers theorem, which is the
inspiration for abc conjecture, and an analogy of it in polynomials, hence a reason for
us to believe the truth of abc conjecture. Mason-Stothers theorem puts a bound on the
degrees of polynomials using the addition relation between them. The idea of changing
“polynomial” with “integer” led Masser and Osterle to abc conjecture. Indeed this is
just the tip of the iceberg, they developed the conjecture with deep considerations of
algebraic geometry and the theory of modular functions.
Later, as another important result of abc conjecture, we introduce Hall conjecture

and give the proof of Weak Hall conjecture using abc conjecture. Hall conjecture puts
a bound on the size of the integral points on a Mordell curve, in addition this bound
simply depends on a coefficient of the defining equation of the curve.
Finally, we give the proofs for the most famous applications of abc conjecture,
which are Fermat’s last theorem for large exponents and Mordell conjecture. Even
though Fermat’s last theorem was proved by Wiles, proving it from abc conjecture
provides a much simpler proof. The same is also true for Mordell conjecture, now
called Faltings’ theorem. Furthermore, the proof of Faltings is not effective, in other
words it proves that there are finitely many rational points on a curve of genus > 1,
but does not give a method for finding them or give an upper bound on the size of the
points. On the other hand, the proof using abc conjecture, given by Noam Elkies [1],
provides an effective upper bound on the size of the points, hence limits the search
criteria for finding the points. Moreover, the proof could be modified to get Siegel’s
theorem which states that an algebraic affine curve of genus ≥ 1, or of genus zero with
3 points at infinity and defined over a number field K has only finitely many integral
points. But it is not exciting as Mordell’s, since effective methods for this case are
already available.
2
2. PRELIMINARIES
We have a long Preliminaries section, including many tools and theorems, which
will be helpful for us to understand the proofs and examples in the following sections.
2.1. Absolute Values
In this section, we define the notion of an absolute value (or multiplicative valu-
ation) and give various examples. The majority of the content is taken from Algebraic
Number Theory Notes by Milne [4], in which one can find more details and the proofs.
Definition 2.1. An absolute value or multiplicative valuation on a field K is a real-

valued function K → R with x 7→ |x|v such that
(i) |x|v ≥ 0 with equality if and only if x = 0

(ii) |xy|v = |x|v |y|v
(iii) there is a Cv ∈ R such that |x + y|v ≤ Cv max(|x|v , |y|v ) for all x, y ∈ K.
Note that Cv ≥ 1, which can be easily seen by taking y = 0 in (iii). If we can

take Cv = 1, then | |v is called nonarchimedean; otherwise archimedean.
By the first two conditions, we see that | |v : K ∗ → R>0 is a homomorphism

where both of them are multiplicative groups. The valuation homomorphism | |v sends
all roots of unity in K ∗ to 1, and | − x|v = |x|v for all x ∈ K ∗ . We say | |v is discrete
when |K ∗ |v is a discrete subgroup of R>0 .
A discrete (additive) valuation ord on K determines an absolute value by |x|v =

e−ord(x) , any e > 1. When we feel like additive valuations, we can pass from multiplica-
tive by taking logs gives loge |x|v = −ord(x), or ord(x) = − loge |x|v .
Definition 2.2. We define the trivial absolute value as |a|v = 1 for all a ∈ K ∗ .
3
Proposition 2.3. Let | |v be a nontrivial nonarchimedean absolute value, and put

v(x) = − log |x|v , x 6= 0 (log to base e for any real e > 1). Then v : K ∗ 7→ R
satisfies the following conditions:
• v(xy) = v(x) + v(y),

• v(x + y) ≥ min{v(x), v(y)}.
If v(K ∗ ) is discrete in R, then v is a multiple of a discrete valuation ord : K ∗ ,→ Z ⊂ R.
Definition 2.4. Let K be a field. A discrete valuation on K is a nonzero homomor-

phism v : K ∗ 7→ Z such that v(a + b) ≥ min(v(a), v(b)). The image of v is of the form
mZ for some m ∈ Z. If m = 1, then v : K ∗ 7→ Z is surjective, and v is said to be
normalized.
Example 2.5. Let K be an arbitrary number field, then any (real or complex) em-
bedding gives rise to an archimedean absolute value with complex-conjugate embeddings
yielding the same absolute value because |a + bi| = |a − bi| in C. Moreover, the only
way for two archimedean embeddings to define the same absolute value is when they
come from a pair of complex-conjugate embeddings. Let r1 denote the number of real
embeddings of K, and r2 denote the number of pairs of complex-conjugate embeddings
of K. Then K has r1 + 2r2 many embeddings and r1 + r2 many archimedean absolute
values up to equivalence.
For a real embedding φ : K ,→ R, |a|φ := |φ(a)|∞ , with | |∞ usual real absolute

value.
For a pair of complex-conjugate embeddings (φ1 , φ2 ) : K ,→ C, |a|(φ1 ,φ2 ) =

|φ1 (a).φ2 (a)|∞ .
√ √ √
A concrete example: K = Q( 3) has two real embeddings φ1 : a+b 3 7→ a+b 3
√ √ √
and φ2 : a+b 3 7→ a−b 3, hence has two archimedean absolute values. So for 2+5 3,
√ √ √ √ √
we get |2 + 5 3|φ1 = |φ1 (2 + 5 3)|∞ = 2 + 5 3 and |2 + 5 3|φ2 = |φ2 (2 + 5 3)|∞ =
√
5 3 − 2.
4
Another concrete example: K = Q(i) has two complex embeddings, which are
complex-conjugate, φ1 : a + bi 7→ a + bi and φ2 : a + bi 7→ a − bi, hence K has only
one archimedean absolute value. Let φ = (φ1 , φ2 ). For 3 + 5i, we get |3 + 5i|φ =
(3 + 5i)(3 − 5i) = 34.
Example 2.6. With an additive discrete valuation ord : K ∗ → Z and a real number
e > 1, we define a nonarchimedean absolute value for nonzero a ∈ K : |a| = (1/e)ord(a) ,
and set |0| = 0. On the field of rational numbers Q, for every prime p, we get the p-adic
absolute value | |p : |a|p = (1/e)ordp (a) with e > 1 a real number. We take the base e = p
to normalize. Hence |a|p = (1/p)ordp (a) = 1/pr where a = a0 .pr with ordp (a0 ) = 0.
3 1 ord3 (3/5) 1 3 1 ord5 (3/5)

A concrete example for Q : = = and = = 5.

5 3 3 3 5 5 5
Let P be a nonzero prime ideal in OK . If α ∈ K ∗ and ordP (α) ≥ 0, then α = x/y

with x, y ∈ OK and ordP (y) = 0. Similarly, on a number field K, we get a normalized
p-adic absolute value for every nonzero prime ideal P by setting the base as the norm
of the prime ideal. We denote the norm by N. Then we have:
For α ∈ K ∗ and for a nonzero prime P, |a|P = (1/N(P))ordP (α) with (α) = αOK =
I
(P )ordP (α) . and P - IJ .
J
Fact 2.7. K number field, a ∈ K ∗ . The norm of a principal fractional ideal aOK is
given by N(aOK ) = |NK/Q (a)|.
√
A concrete example for K : Let K = Q( 2). The ideal (2) factorizes into prime
√ √ √ √
ideals in the ring of integers Z[ 2] as 2Z[ 2] = (2, 2)2 . Let P = (2, 2). Then
2
2 1 1
|2|P = (1/NP ) = = .
2 4
From an absolute value, one can construct a metric on K with d(a, b) = |a − b|,
and hence a topology on K.
5
Proposition 2.8 ( [4], Chapter 7, Proposition 7.8). Let | |v , | |w be nontrivial absolute

values on K. Then the followings are equivalent:
(i) | |v and | |w define the same topology on K;

(ii) If | |v < 1, then | |w < 1;
(iii) | |w = (| |v )a for some a > 0.
In this case | |v and | |w are said to be equivalent.
There is a classification theorem on valuations on the field of rational numbers:
Theorem 2.9 (Ostrowski Theorem, [5]). Let | |v be a nontrivial absolute value on Q.
• If | |v is archimedean, then | |v is equivalent to | |∞ .

• If | |v is nonarchimedean, then | |v is equivalent to | |p for exactly one prime p.
Note that | |∞ is the usual real absolute value, and we say that | |∞ is normalized.
Theorem 2.10 (Product Formula for Q). For p = 2, 3, 5, 7, . . . , ∞, let | |p be the

corresponding normalized absolute value on Q. For any nonzero rational number a,
Y
|a|p = 1.
p∈MQ
This is a simple reflection of the fact that Z has unique factorization.
There is also a classification theorem of valuations for number fields:
Theorem 2.11 (Ostrowski Theorem for K, [5]). Every nontrivial absolute value on
K is equivalent to a p-adic absolute value for a unique prime ideal P in OK or is
equivalent to an archimedean absolute value coming from a real or complex-conjugate
pair of embeddings.
6
Theorem 2.12 (Product Formula for K). Let a ∈ K ∗ . Then

Q
|a|v = 1, where v
ranges over all normalized valuations of K.
√
Example 2.13. Let K = Q( 5) and a = 3. Then K has two real embeddings φ1 :
√ √ √ √
a + b 5 7→ a + b 5 and φ2 : a + b 5 7→ a − b 5, hence has two archimedean absolute
values. For a = 3, we have |3|φ1 = |3|φ2 = 3. Moreover, the element 3 ∈ K forms the
principal ideal 3OK where it has the factorization 3OK = P for some prime ideal P
with norm 9. Then |3|P = (1/9). Therefore multiplying over all absolute values for the
element 3, we obtain 1.
2.2. Prime Ideals and Factorization
In the following four subsections, we introduce some key notions from algebraic
number theory such as plane curves, factorization of ideals, and ramified primes. It
is a summary of three chapters from An Invitation to Arithmetic Geometry by Dino
Lorenzini. For more details and proofs, see [6].
2.2.1. Introduction
Let K be a field and K be an algebraic closure of K. In the first subsection

of the sequence, we explain that a nonsingular plane curve, defined by an irreducible
polynomial f (x, y) in K[x, y], monic in y, corresponds to the integral closure of K[x]
in the function field of the given curve.
Let f (x, y) ∈ K[x, y] be a polynomial in two variables with coefficients in K and

let d be the degree of f. Let
Zf (K) := {(a, b) ∈ K × K|f (a, b) = 0}.
Definition 2.14. The set Zf (K) is called an affine plane curve. The set Zf (K) is
the set of points with coordinates in K of the affine plane curve of degree d defined by
f (x, y).
7
Lemma 2.15. Let f (x, y) ∈ K[x, y] be a polynomial of degree d > 0. Then Zf (K) is
an infinite set. In particular, Zf (K) is not empty.
We assume that Zf (K) and K are endowed with Zariski topology, which will
allow us to view the ring Cf := K[x, y]/hf i as a ring of continuous functions on the
curve Zf (K).
Let C(Zf (K), K) denote the set of continuous functions from Zf (K) to K. The
map of sets if : Cf → C(Zf (K), K) is injective. We call Cf the ring of algebraic
functions on Zf (K), or simply the ring of functions of Zf (K).
Definition 2.16. Let f ∈ K[x, y] be an irreducible polynomial, so that the ring Cf :=

K[x, y]/hf i is an integral domain. We denote by K(Zf ) the field of fractions of the
ring Cf and call it the field of rational functions of the affine curve defined by f. The
elements of K(Zf ) are called the rational functions of Zf (K).
Definition 2.17. Let f (x, y) ∈ K[x, y]. The degree of f in y, degy (f ), is the degree of
the polynomial f viewed as a polynomial in the variable y with coefficients in K[x].
Let f (x, y) = an (x)y n + · · · + a0 (x) be an irreducible polynomial in K[x, y]. The

natural map
φ : K[x] → Cf := K[x, y]/hf i
is injective if f (x, y) 6= cx + d, for all c, d ∈ K. In this case, it can be extended to an

injection K(x) → K(Zf ) with g(x)/h(x) 7→ φ(g(x))/φ(h(x)) of the field of fractions
K(x) of K[x] into the field of fractions K(Zf ) of Cf . The field K(Zf ) is isomorphic to
K(x)[y]/(f ) and, therefore, the extension K(Zf )/K(x) is of finite degree degy (f ). We
have obtained in this way a triple of K[x], K(x), K(Zf ). It is not always the case that
Cf is the integral closure of K[x] in K(Zf ).
Fact 2.18. Let A be a domain of dimension 1. Let B be a domain containing A and

such that each element of B is integral over A. Then B has dimension 1.
8
Fact 2.19. Let A ⊆ B be two rings. If A is noetherian and B is a finitely generated

A-module, then B is noetherian.
When an (x) = 1, every element of Cf is integral over K[x], and hence Cf is an

integral extension of K[x]. When an (x) = 1, the ring Cf has dimension 1 by Fact 2.18.
Furthermore, since Cf is generated as a ring over K[x] by the classes of the elements
1, y, . . . , y n−1 , we conclude from Fact 2.19 that Cf is noetherian. Therefore, the domain
Cf is integrally closed if and only if it is a Dedekind domain.
2.2.2. Points and Maximal Ideals
Here, we show the association between the set of points of the curve Zf (K) and
the set of maximal ideals M ax(Cf ) of the ring Cf := K[x, y]/(f ).
Proposition 2.20. Let M be a maximal ideal of K[x, y]. Then there exists a point
(a, b) ∈ A2 (K) such that M is generated by (x − a) and (y − b).
Corollary 2.21. Let f ∈ K[x, y] be an irreducible polynomial. An ideal M of the ring

Cf := K[x, y]/(f ) is maximal if and only if it can be generated by the images in Cf
of two elements of the form x − a and y − b in K[x, y], with f (a, b) = 0. Let If (a, b)
denote the ideal of Cf generated by the images of x − a and y − b in Cf . The map
If : (a, b) 7→ If (a, b) is a bijection from Zf (K) to M ax(Cf ).
2.2.3. Morphisms of Curves
We introduce the notion of morphism of curves and the relation between mor-
phism of curves and the homomorphism of rings of functions.
Let φ : Zf (K) → Zg (K) be any map between two curves. This map φ uniquely
determines two maps φ1 , φ2 : Zf (K) → K such that φ(a, b) := (φ1 (a, b), φ2 (a, b)).
Definition 2.22. A map φ : Zf (K) → Zg (K) between two plane curves is a morphism
of affine plane curves if there exist two polynomials α(x, y) and β(x, y) in K[x, y] such
that φ1 (a, b) = α1 (a, b) and φ2 (a, b) = β(a, b) for all (a, b) ∈ Zf (K).
9
Lemma 2.23. Let f, g ∈ K[x, y] be irreducible polynomials. Endow Zf (K) and Zg (K)
with the Zariski topology. Let φ : Zf (K) → Zg (K) be a morphism of curves. Then φ
is a continuous map.
Any such continuous map φ : Zf (K) → Zg (K) defines a map:
φC ∗ : C(Zg (K), K) → C(Zf (K), K)
with h 7→ h ◦ φ. Consider the following diagram:
Cg Cf
ig if .
φC ∗
C(Zg (K), K) C(Zf (K), K)
Lemma 2.24. Let φ : Zf (K) → Zg (K) be a continuous map between two algebraic
curves endowed with the Zariski topology. The map φ is a morphism of plane curves if
and only if (φC ∗ ◦ ig )(Cg ) ⊆ if (Cf ).
If φ is a morphism of curves, then the map φ∗C|i can be used to define a

g (Cg )
homomorphism of K-algebras between Cg and Cf , denoted by φ∗ , in such a way that

the following diagram is commutative:
φ∗
Cg Cf
ig if .
∗
φC
Let φ∗ : Cg → Cf be any homomorphism of K−algebras. The homomorphism

φ∗ defines in a natural way a morphism of curves φ : Zf (K) → Zg (K). Let φu (x, y) ∈
K[x, y] be a polynomial such that the class of φu (x, y) in Cf is φ∗ (class of u).
10
Similarly, let φv (x, y) ∈ K[x, y] be a polynomial such that the class of φv (x, y) in
Cf is φ∗ (class of u). We set
φ : Zf (K) → Zg (K)
with (a, b) 7→ (φu (a, b), φv (a, b)). The map φ is well-defined and does not depend on
the choices of φu (x, y) and φv (x, y).
Lemma 2.25. Let φ∗ : Cg → Cf be any homomorphism of K−algebras. Let φ :

Zf (K) → Zg (K) be the induced morphism of curves described above. Let
φC ∗ : C(Zg (K), K) → C(Zf (K), K)
with h 7→ h ◦ φ. The following diagram is commutative:
φ∗
Cg Cf
ig if .
φC ∗
Definition 2.26. The set of prime ideals of a ring A is called the spectrum of A and is
denoted by Spec(A). Given any ring homomorphism φ : A → B, we let Spec(φ) denote
the natural map:
Spec(φ) : Spec(B) → Spec(A)
with P 7→ φ−1 (P ).
Let Zf (K) be any plane curve. Recall that the map If : Zf (K) → M ax(Cf ),
with (a, b) 7→ hx − a, y − bi, is bijective.
By following the set-up in “An Invitation to Arithmetic Geometry” by Dino

Lorenzini, we denote the quotient as hx − a, y − bi by abuse of notation.
11
Lemma 2.27. Let Cf and Cg be the rings of functions of two affine plane curves
given by irreducible polynomials f (x, y) and g(u, v), respectively. Let φ∗ : Cg → Cf
be a homomorphism of K−algebras. Then the map Spec(φ∗ ) restricts to a map φ0 :
M ax(Cf ) → M ax(Cg ), with M 7→ (φ∗ )−1 (M ). Moreover, the following diagram is
commutative:
If
Zf (K) M ax(Cf )
φ φ0 .
Ig
Zg (K) M ax(Cg )
Theorem 2.28. Let f (x, y) ∈ K[x, y] be an irreducible polynomial. The ring Cf :=

K[x, y]/(f ) is integrally closed if and only if the curve Zf (K) is nonsingular.
Theorem 2.29. Let f ∈ K[x, y] be an irreducible polynomial. Every nonzero prime

ideal of K[x, y]/(f ) is a maximal ideal generated by the images of (x − a) and (y − b)
for some (a, b) ∈ K × K such that f (a, b) = 0.
In conclusion of the section, for an irreducible polynomial f ∈ K[x, y], the set
Zf (K) is in bijection with M ax(Cf ).
2.2.4. Factorization of Ideals
In the final subsection of the sequence, we describe the factorization of ideals and
ramification, then give several important field extensions such as Kummer extensions
and Eisenstein extensions.
Theorem 2.30. In a Dedekind domain A, any nonzero proper ideal has a unique
factorization as a product of nonzero prime ideals.
Lemma 2.31. Let A be a Dedekind domain and let K denote its field of fractions. Let
B be the integral closure of A in a finite extension L of K. Assume that B is a finitely
generated A−module and let P be a maximal ideal of A. Then PB 6= B.
12
Then in our set-up above, by Lemma 2.31, it follows that the ideal PB factors
in B into a product of prime ideals of B, since B is a Dedekind domain. Let
PB = M1e1 . . . Mses ,
for some positive integers e1 , . . . , es . The integer eMi /P := ei is called the ramification
index of Mi over P.
Since Mi ∩ A = P, the inclusion A ⊆ B induces injections
A/P → B/Mi ,
for i = 1, . . . , s. Since B is a finitely generated A−module, the field B/Mi is a finite

extension of the field A/P. Let fMi /P := [B/Mi : A/P] be the dimension of B/Mi as an
(A/P)−vector space. The field A/P is called the residue field of A at P. The integer
fMi /P is called the residual degree of Mi over P.
Remark 2.32. For a maximal ideal M of B, the prime ideal P := M ∩ A is maximal.
Theorem 2.33. Let A be a Dedekind domain with field of fractions K. Let L/K be a
finite extension. Let B denote the integral closure of A in L. If B is a finitely generated
A−module, then
X
[L : K] = eM/P fM/P .
M |PB
√
Example 2.34. Let A = Z. Let K = Q( d) where d is a square-free integer. Then
√ √
the ring of integers OK is equal to Z[ d] when d ≡ 2, 3 mod 4, and Z[(1 + d)/2]
when d ≡ 1 mod 4.
Let p ∈ Z be a prime number. The factorization of the primes of Z above in OK

is known for every case:
13
√ 2
• If p | d, then pOK = hp, di .
• If p = 2 and 2 - d, then

prime if d ≡ 5 mod 8



√ √

 * +* +

 1+ d 1− d
2OK = 2, 2, if d ≡ 1 mod 8 (2.1)

 2 2
√


h2, 1 + di2

if d ≡ 3 mod 4

• If p is odd and p - d, then

prime

if d is not a square mod p
pOK = √ √ (2.2)
hp, d + nihp, d − ni 2
if n ≡ d mod p

Proposition 2.35. Let A be any Dedekind domain. Let f (y) be an irreducible monic
polynomial in A[y]. Let α denote a root of f (y) in a fixed algebraic closure of K, the
field of fractions of A. Let M ∈ M ax(A[α]). Let P := M ∩ A. Factor the reduction
s
gi (y)ei , with gi (y), i = 1, . . . , s, distinct monic
Q
f (y) of f (y) in (A/P )[y] as f (y) =
i=1
s
gi (y)ei , with h(y) ∈ P A[y], and such
Q
irreducible polynomials. Write f (y) = h(y) +
i=1
that the reduction of gi (y) in (A/P )[y] is equal to gi (y), for all i = 1, . . . , s. Then the
ideal M is generated by the elements of P and by gi0 (α), for some unique i0 ∈ {1, . . . , s}.
Moreover, the following statements are equivalent:
• f 0 (α) ∈
/ M.
• ei0 = 1 and the extension A[α]/M is separable over A/P.
Definition 2.36. Let A be a subring of a ring B. The ring extension B/A is called
simple or monogenic if there exists an element α ∈ B such that B = A[α].
14
√ √
Example 2.37. Let K = Q( 2, 3), we find its ring of integers.
√ √
Let α ∈ ZK . Then α = a + b 2 + ci + d 2i for some a, b, c, d ∈ Q. Since α is an
√ √
algebraic integer, then indeed all of its conjugates are so: α2 = a − b 2 + ci − d 2i,
√ √ √ √
α3 = a+b 2−ci−d 2i, α4 = a−b 2−ci+d 2i. By the closedness under addition in
√
ZK , these are also algebraic integers: α + α2 = 2a + 2ci, α + α3 =√2a + 2b 2, α√+ α4 =
√ A + B 2 + Ci + D 2i
2a + 2d 2i and integral if 2a, 2b, 2c, 2d ∈ Z. Rewrite α = for
2
integers A = 2a, B = 2b, C = 2c, and D = 2d. By the closedness under multiplication,
A2 − 2B 2 − C 2 + 2D2 AC − 2BD
αα2 = a2 − 2b2 − c2 + 2d2 + 2aci − 4bdi = + i is also
4 2
integral. Therefore 4|A2 − 2B 2 − C 2 + 2D2 and 2|AC − 2BD, then 2|AC. If only A or C
were even, then the expression A2 − 2B 2 − C 2 + 2D2 would be odd, and this would give a
contradiction with divisibility to 4. Hence both of them are even. Using this in the first
expression, we get 4| − 2B 2 + 2D2 or 2|D2 − B 2 . So either both of them are even or both
√ √
of them are odd. We find that integers are all of the form α = a + b 2 + ci + d 2i for
some a, c ∈ Z and b and d both integral or both halves of odd integers. Such elements
√ 1+i
are integer linear combinations of 1, 2, i, and √ , thus elements of this form are
2
all integers.
We have showed what kind of elements ZK has. Now our claim is ZK = Z[γ], where
1+i
γ= √ :
2
Z[γ] ⊂ ZK is obvious, since γ ∈ ZK .
√
Using the relations γ 2 = i, γ − γ 3 = 2, each element in the integral basis can be
written as an element in Z[γ], so ZK ⊂ Z[γ] and we are done.
15
Proposition 2.38. Suppose that K is a number field, and that ZK = Z[γ]. Write
g(X) ∈ Z[X] for the minimal polynomial of γ over Z. Let p be a prime in Z, and let
g(X) = g1 (X)e1 · · · gr (X)er
be the factorization of the minimal polynomial g modulo p of γ into irreducibles. Then

pZK = P1e1 · · · Prer , for certain distinct ideals Pi of ZK ; the inertia degree of Pi is simply
given by the degree of gi (X).
Kummer extensions:
Let f (y) = y n − a be an irreducible polynomial in Z[y]. Let α be a root of f (y)

in C. The extension K = Q(α) is called a Kummer extension of Q.
Proposition 2.39. Let p be a prime. Let a = q1 · · · qs be a product of distinct primes.

√
Let α := p a be a root of y p − a. If p | a, then Z[α] is equal to the ring of integers of
Q(α). If p - a, then Z[α] is equal to the ring of integers of Q(α) if and only if p2 - ap −a.
Moreover, when Z[α] = OQ(α) , the primes p and qi , i = 1, . . . , s, are the only primes of
Z that ramify in Z[α].
In addition, a factorization in Z[α] of a prime q such that q | a, can be explicitly

determined: hqiZ[α] = hα, qip .
16
Example 2.40. Let K be any field and f (y) = y n − a be an irreducible polynomial

in K[x, y]. So a := a(x) is a polynomial in K[x]. Then Theorem 2.28 implies that
Cf := K[x, y]/hy n − a(x)i is integrally closed if and only if the curve Zf (K) has no
singular points. Note that for the polynomial f (x, y) = y n − a(x), we have fx = a0 (x)
and fy (x) = ny n−1 . If the characteristic of K is prime to n, then Zf (K) is nonsingular
if and only if a(x) is square-free. Assume that this is the case, i.e. Cf is a Dedekind
n
Q
domain. We write a(x) = c (x − ai ) with ai ∈ K and ai 6= aj if i 6= j. Consider
i=1
the map K[x] 7→ Cf . The maximal ideal hx − di ∈ K[x] factors as the product of n
distinct maximal ideals of Cf if and only if d 6= ai for all i = 1, . . . , n. We find that
hx − ai iCf = hy, x − ai in . In particular, the only ideals in K[x] that ramify in Cf are
the ideals hx − ai i, i = 1, . . . , r.
Example 2.41. Let f (x, z) = z n + xn + 1 and Cf := C[x, z]/hf i. Consider the map
π : Zf (C) → A1 (C), with [x : 1 : z] 7→ [x : 1], corresponding to the inclusion C[x] → Cf .
The degree of the map deg π = [Cf : π ∗ (C[x])] is equal to n, since z n +xn +1 is irreducible
and monic in x. All partial derivatives of the homogeneous polynomial of f vanish at
[0 : 0 : 0], but there is no such point in the projective plane. Hence the curve is smooth.
Since the curve Zf (C) is nonsingular, the ring Cf is the integral closure of the ring
n
C[x] in C(Zf ). By Example 2.40, a(x) = xn + 1 = (x − ξi )n where ξi , i = 1, . . . , n
Q
i=1
are the n-th roots of −1 and hx − ξi iCf are the only ramified ideals in Cf , which are
explicitly factorized as hx − ξi iCf = hx − ξi , zin . Therefore, from the bijection between
the nonsingular points of the curve and the maximal ideals of Cf , we say that there are
exactly n ramification points with ramification degree n.
Eisenstein extensions:
Definition 2.42. Let A be any domain. Let P ∈ M ax(A). Let f (y) = an y n +

an−1 y n−1 + · · · + a0 ∈ A[y] be a polynomial of positive degree such that an ∈
/ P, ai ∈ P,
/ P 2 . We call f (y) an Eisenstein polynomial, or a
for all i = 0, . . . , n − 1, and a0 ∈
P -Eisenstein polynomial, when the dependence on P must be specified.
17
n−1
Lemma 2.43. Let p be a maximal ideal of Z. Let f (y) = y n + ai y i be a p-Eisenstein
P
i=0
polynomial, then f (y) is irreducible in K[y]. Let α be a root of f (y) in C. Then A[α] is
a local principal ideal domain in K(α) and, in particular, A[α] is the integral closure
of A in K(α). Moreover, pA[α] = hαin .
Cyclotomic extensions:
Let p be a prime number and ξp denote a root of f (y) := 1 + y + · · · + y p−1 .
Proposition 2.44. The ring Z[ξp ] is the ring of integers of Q(ξp ). Let P denote the
ideal of Z[ξp ] generated by ξp − 1. Then P is a prime ideal and hpiZ[ξp ] = Pp−1 . The
prime p is the only prime of Z that ramifies in Z[ξp ].
Proposition 2.45. Let q 6= p be a prime of Z. Let fq denote the order of the class of
q in (Z/pZ)∗ . Then there exist exactly s := (p − 1)/fq distinct prime ideals L1 , . . . , Ls
of Z[ξp ] over hqi, and hqiZ[ξp ] = L1 · · · Ls . Moreover, the residual index fLi /q of each
prime Li is equal to fq .
Galois extensions:
Let L/K be a finite Galois extension with Galois group G = Gal(L/K).
Proposition 2.46. Let p be a maximal ideal of Z. Let Mi and Mj be two distinct

maximal ideals in π −1 (p). Then there exists σ ∈ G such that σ(Mi ) = Mj . Moreover,
eM1 /p = · · · = eMs /p = e, and fM1 /p = · · · = fMs /p = f. In particular,
pB = hM1 · · · Ms ie , with ef s = [K : Q].
Theorem 2.47 ( [7], Theorem 1.3). For a number field K, the primes which ramify
are those dividing the integer discZ (OK ).
18
2.3. Algebraic Curves Basics
In this section, we introduce some notions about algebraic curves such as divisors
and rational maps. We also give famous Riemann-Roch theorem and Riemann Hurwitz
formula. We mostly benefit from The Arithmetic of Elliptic Curves by Silverman and
Generalization of the ABC-conjecture by Nils Bruin. For more details, see [8] and [9].
2.3.1. Divisors
Throughout this section, we take K as a perfect field unless something different

is specified.
Definition 2.48. Let C be a curve defined over K with K = K. A divisor of C is a

P
formal sum D := nP P with nP ∈ Z and all but finitely many nP = 0. The set of
P ∈C
points P for which nP 6= 0 is called the support of D, denoted by Supp(D).
The divisors of C form a free abelian group under addition, the divisor group,
Div(C).
Now we generalize to the case where K does not need to be algebraically closed.
Let GK = Gal(K/K) be the absolute Galois group of K, i.e., the group of all auto-
morphisms of K that fix K.
P σ where
P
Definition 2.49. The prime divisors of C over K are defined by PP :=
σ∈GK
P ∈ C(K).
P
Definition 2.50. A divisor D = nP P ∈ Div(C) is defined over K if for all
P ∈C
σ
σ ∈ GK , we have Dσ = D, where Dσ = nP P σ . The subset of
P P
nP P =
P ∈C P ∈C
Div(C), comprising of those divisors that are defined over K, forms the subgroup of
K-rational divisors.
19
Note that D = Dσ does not necessarily imply that P = P σ for all points in the
support of D, i.e., it does not have to be fixed pointwise, but rather setwise. But if
D = Dσ for all σ ∈ GK , then nP σ = nP for all σ ∈ GK . Hence we can group the
terms of a K-rational divisor into GK -orbits with a single coefficient nP applied to all
the points in the orbit. Equivalently, we can view a K-rational divisor as a sum over
GK -orbits of points P ∈ C(K).
Remark 2.51. The orbit of any algebraic number is the set of all of its conjugates,
hence finite.
Definition 2.52. Let C be a curve defined over K. Then GK -orbits of C(K) are called
closed points, which we also denote by P. A rational divisor of C/K is a formal sum
P
D := nP P , where P ranges over the closed points of C/K, with nP ∈ Z and all but
finitely many nP = 0. The group of rational divisors on C is defined as the subgroup
generated by the prime rational divisors, and denoted by DivK (C).
The points in C(K) correspond to a proper subset of the set of closed points, the
trivial GK -orbits consisting of a single element. But every point in C(K) is contained
in a closed point of C/K.
Let C be a curve defined over a perfect field K, we denote by C/K. Let K(C)
be the function field of C over K, K[C]P be the local ring of C at P, and MP be the
maximal ideal of K[C]P .
Definition 2.53. Let P ∈ C a smooth point. The normalized valuation on K[C]P is

given by
ordP : K[C]P → {0, 1, 2, . . . } ∪ {∞},
ordP (f ) = sup{d ∈ Z : f ∈ MP d }.
20
Using ordP (f /g) = ordP (f ) − ordP (g), we can extend the valuation to K(C),
ordP : K(C) → Z ∪ ∞.
Let C be a smooth curve, and let f ∈ K(C)∗ , then we can associate f to a divisor
P
div(f ) = ordP (f )P . Since each ordP is a valuation, the map
P ∈C
div : K(C)∗ → Div(C)
is a homomorphism of abelian groups.
Definition 2.54. Let f ∈ K(C)∗ . The divisor of f is, divf :=

P
ordP (f )P , called
P ∈CF
a principal divisor.
Note that for any closed point, a function f ∈ K(C) vanishes at a point P ∈ C(K)
if and only if it vanishes on the whole GK -orbit of P.
P P
For a principal divisor divf = P, the divisors div0 f := nP P and div∞ f :=
P nP nP >0
−nP P are called the divisor of zeros and the divisor of poles, respectively. We have
nP <0
the equality divf = div0 f − div∞ f. Note that div0 f is sometimes called the zero divisor
of f, e.g. in the paper, which we study later, “Abc Implies Mordell” by Noam Elkies.
The Picard group of C, denoted by P ic(C), is the quotient of Div(C) by its

subgroup of principal divisors.
Definition 2.55. Let C/K be a curve. If P is a closed point of C/K, we define the
degree of P to be degP = [K(P ) : K], where K(P ) is the minimal field extension K 0 of
K in K such that P is K 0 -rational. Equivalently, degP is the cardinality of the closed
point P as a GK -orbit of points in C(K).
21
P
The degree of a divisor D = nP P in the group of K-rational divisors is degD =
P
nP degP . When K is algebraically closed, we have degP = 1 for all points P, so
P
degD = nP .
The divisors of degree 0 form a subgroup of Div(C), which we denote by Div 0 (C) =
{D ∈ Div(C) : degD = 0}.
Example 2.56. Let F5 be a curve over Q, where F5 is the zero locus of x5 +y 5 +z 5 = 0.

Let f : F5 (Q) → P1 (C) with (x, y, z) 7→ (−x5 , y 5 ). Let D be the zero divisor of f. Let
ξ5 6= −1 be a 5th root of −1. Then D = 5((0, 1, ξ5 ) + (0, 1, ξ5 2 ) + (0, 1, ξ5 3 ) + (0, 1, ξ5 4 )) +
5(0, 1, −1), where P1 := (0, 1, ξ5 ) + (0, 1, ξ5 2 ) + (0, 1, ξ5 3 ) + (0, 1, ξ5 4 ) and P2 := (0, 1, −1)
are prime rational divisors of degrees deg(P1 ) = 4 and deg(P2 ) = 1. Then deg(D) = 25.
Now we partially order divisors by defining the relation ≤ on DivK C by D1 ≤ D2

if and only if nP (D1 ) ≤ nP (D2 ) for all P where P ranges over all closed points of C/K.
Then we have some consequences:
• If D1 ≤ D2 , then D1 + E ≤ D2 + E for any divisor E

• If D1 ≤ D2 and E1 ≤ E2 , then D1 + E1 ≤ D2 + E2 .
Note that ≤ is not a total ordering on DivK C since most pairs of divisors are
incomparable.
P
Definition 2.57. A divisor D = nP P is called effective or positive if nP ≥ 0 for all
P. We denote by D ≥ 0.
Like principal divisors divf = div0 f −div∞ f, every divisor can be written uniquely
P
as a difference of two effective divisors, D = D0 − D∞ , where D0 := nP P and
P nP >0
D∞ := −nP P .
nP <0
Definition 2.58. Two divisors are linearly equivalent if their difference is the divisor
of a function.
Proposition 2.59. Let C be a smooth curve and let f ∈ K(C)∗ . Then deg(div(f )) = 0.
22
Definition 2.60. Let V1 and V2 ⊂ Pn be projective varieties. A rational map from

V1 to V2 is a map of the form f : V1 → V2 with f = [f0 , . . . , fn ], where the functions
f0 , . . . , fn ∈ K(V1 ) have the property that for every point P ∈ V1 at which f0 , . . . , fn are
all defined, f (P ) = [f0 (P ), . . . , fn (P )] ∈ V2 . Multiplying with an appropriate function
g, we can assume that all gfi are defined in x ∈ V1 without changing the value of f in
x. We say f is regular at x if there exists a representative (gf0 , . . . , gfn ) that is defined
in x. If f is regular for all x ∈ C, then we say that f is a morphism.
Let φ : C1 → C2 be a nonconstant map of smooth curves, then φ induces maps

on the function fields of C1 and C2 ,
φ∗ : K(C2 ) → K(C1 ) and φ∗ : K(C1 ) → K(C2 ).
We similarly define maps of divisor groups as follows:
X
φ∗ : Div(C2 ) → Div(C1 ) given by (Q) 7→ eφ (P )(P ),
P ∈φ−1 (Q)
φ∗ : Div(C1 ) → Div(C2 ) given by (P ) 7→ (φP )
and extend Z−linearly to arbitrary divisors.
Definition 2.61. The degree of a morphism of curves φ : C1 → C2 is the degree of the

corresponding extension of function fields deg(φ) = [K(C1 ) : φ∗ (K(C2 ))].
Proposition 2.62. Let φ : C1 → C2 be a nonconstant map of smooth curves.
• deg(φ∗ D) = (deg(φ))(degD) for all D ∈ Div(C2 ).

• φ∗ (div(f )) = div(φ∗ f ) for all f ∈ K(C2 )∗ .
• deg(φ∗ D) = degD for all D ∈ Div(C1 ).
• φ∗ (div(f )) = div(φ∗ f ) for all f ∈ K(C1 )∗ .
• φ∗ ◦ φ∗ acts as multiplication by deg(φ) on Div(C2 ).
23
2.3.2. Differentials
Definition 2.63. Let C be a curve. The space of differential forms on C, denoted by

ΩC , is the K−vector space generated by symbols of the form dx for x ∈ K(C), subject
to the usual relations:
• d(x + y) = dx + dy for all x, y ∈ K(C).

• d(xy) = xdy + ydx for all x, y ∈ K(C).
• da = 0 for all a ∈ K.
Let φ : C1 → C2 be a nonconstant map of curves, then associated function field

map φ∗ : K(C2 ) → K(C1 ) induces a map on differentials,
X X
φ∗ : ΩC2 → ΩC1 , given by φ∗ fi dxi = (φ∗ fi )d(φ∗ xi ).
Proposition 2.64. • ΩC is a 1-dimensional K(C)-vector space.

• Let x ∈ K(C). Then dx is a K(C)-basis for ΩC if and only if K(C)/K(x) is a
finite separable extension.
• φ is separable if and only if φ∗ : ΩC2 → ΩC1 is injective.
Definition 2.65. Let w ∈ ΩC . The divisor associated to w is
X
div(w) = ordP (w)P ∈ Div(C).
P ∈C
The differential w ∈ ΩC is holomorphic if ordP (w) ≥ 0 for all P ∈ C, and it is

nonvanishing if ordp (w) ≤ 0 for all P ∈ C.
For more details, please see Chapter 2 in Arithmetic of Elliptic Curves by Silver-
man [8].
If w1 , w2 ∈ ΩC are nonzero differentials, then there is a function f ∈ K(C)∗ such

that w1 = f w2 . Thus div(w1 ) = div(f ) + div(w2 ).
24
Definition 2.66. The canonical divisor class on C is the image in P ic(C) of div(w)
for any nonzero differential w ∈ ΩC . Any divisor in this class is called a canonical
divisor.
Proposition 2.67. Let E be an elliptic curve. Then the invariant differential w associ-
ated to a Weierstrass equation for E is holomorphic and nonvanishing, i.e., div(w) = 0.
2.3.3. Riemann-Roch Theorem
The partial order we put on divisors is useful for describing zeros and poles of
functions.
Definition 2.68. Let D ∈ Div(C). We associate to D the set of functions
L(D) = {f ∈ K(C)∗ : div(f ) ≥ −D} ∪ {0},
which is known as Riemann-Roch space associated to the divisor D. It is sometimes

called the linear system associated with D. Its dimension is denoted by l(D) = dimK L(D).
Proposition 2.69. Let D ∈ Div(C).
• If degD < 0, then L(D) = {0} and l(D) = 0.

• L(D) is a finite-dimensional K-vector space.
• If D0 ∈ Div(C) is linearly equivalent to D, then
L(D) ∼
= L(D0 ), and so l(D) = l(D0 ).
Theorem 2.70 (Riemann-Roch Theorem). Let C be a smooth curve and let KC be a

canonical divisor on C. There is an integer g ≥ 0, called the genus of C, such that for
every divisor D ∈ Div(C),
l(D) − l(KC − D) = degD − g + 1.

25
Corollary 2.71. (i) l(KC ) = g.

(ii) degKC = 2g − 2.
(iii) If degD > 2g − 2, then l(D) = degD − g + 1.
Proof. (i) Using Theorem 2.70 with D = 0, we get l(KC ) = l(0) + g − 1. If D = 0,

then L(D) is the set of functions that have no poles at all, i.e., div(f ) ≥ 0. Then
the rational map defined by f , f : C → P1 , is a constant map, hence f ∈ K and
L(0) ∼
= K. So we have l(0) = 1 and l(KC ) = g.
(ii) Using Theorem 2.70 with D = KC and Part (i), we get l(KC ) = 2g − 2.
(iii) If degD > 2g − 2, then by Part (ii), deg(KC − D) < 0. Take an arbitrary nonzero
f ∈ L(KC − D), then 0 = deg(div(f )) ≥ −deg(KC − D), so deg(KC − D) ≥ 0.
Hence L(KC − D) = 0, and l(KC − D) = 0. Now applying Theorem 2.70, we have
l(D) = degD − g + 1.
Proposition 2.72. Let C/K be a smooth curve and let D ∈ DivK (C). Then L(D)
has a basis consisting of functions in K(C).
Proposition 2.73. If an elliptic curve E given by a Weierstrass equation is singular,

then there exists a rational map φ : E → P1 of degree one.
Lastly, we look at the behavior of a map in the neighborhood of a point. Let C

be a curve over some number field K and f ∈ Q(C) be a rational function. We define
the branch number of f in x ∈ C by
bf (x) = ordx (f − f (x)) − 1 if f (x) is finite, and

1
bf (x) = ordx − 1 if f has a pole at x.
f
26
For α ∈ K ∪ {∞} we define the ramification of f above α by
X
bf (α) = bf (x).
x:f (x)=α
We have deg(f ) = |f −1 ({α})| + bf (α).
We give a theorem on how the genera of curves linked by a nonconstant map are
related.
Theorem 2.74 (Hurwitz, [8]). Let φ : C1 → C2 be a nonconstant separable map of

smooth curves of genera g1 and g2 , respectively. Then
X
2g1 − 2 ≥ (deg(φ))(2g2 − 2) + (eφ (P ) − 1).
P ∈C1
Further, equality holds if and only if one of the following two conditions is true:
(i) char(K) = 0.
(ii) char(K) = p > 0, where p is prime, and p does not divide eφ (P ) for all P ∈ C1 .
It turns out that the total ramification of a rational function on a curve is closely
connected with the genus of that curve.
Remark 2.75. Let C be a curve of genus g over an algebraically closed field K and
let f be a nonconstant rational function on that curve. Then
X
2deg(f ) = bf (α) + 2 − 2g.
α∈K
If C is a curve over a number field K, then this need not be true. However, since the
branch number bf (α) is nonzero for only finitely many α ∈ Q, the theorem is true for
some finite extension of K, dependent on f.
27
Definition 2.76. The Fermat curve Fn is the algebraic curve in the complex projective
plane defined in homogeneous coordinates (x, y, z) by the Fermat equation xn +y n +z n =
0.
Example 2.77. Let f (x, z) = z n + xn + 1 be an irreducible polynomial with degz (f ) =

n > 0 and monic in z. We consider the map π : Zf (C) → A0 (C). In Example 2.41, we
showed that there are exactly n ramification points with ramification index n, and the
degree of the map π is n. Note that the projective line P1 (C) has no handles and hence
its genus is 0. Using Hurwitz Theorem 2.74, we see that the genus of the Fermat curve
is
(n − 1)(n − 2)
g= .
2
Hence for n ≥ 4, the curve has genus ≥ 2, hence has finitely many rational points by
Mordell conjecture 5.1.
2.3.4. Rational Maps
Let C be a curve over K and let f0 , . . . , fn ∈ K(C). Then f : C → Pn with

x 7→ (f0 (x), . . . , fn (x)) is a rational map from C into Pn , which is defined for x ∈ C if
not all fi (x) = 0 and no fi has a pole in x. Since C is a curve, we can construct for
each x ∈ C a function g with a pole or a zero at x of arbitrary order. Therefore, if f
is regular at some point, i.e., f 6= (0, . . . , 0), then f is a morphism.
Definition 2.78. Let C be a smooth projective curve and D ∈ Div(C). Let F =

{f0 , . . . , fn } be a basis for L(D), then φD,F : C → Pn with x 7→ (f0 (x), . . . , fn (x)) is a
rational map. If this map is a nonconstant morphism, then L(D) is said to be without
base point.
Remark 2.79. For curves, L(D) is without base point once l(D) ≥ 2 because the
associated map φD must be nonconstant and regular somewhere.
Definition 2.80. If φD is injective and regular, then D is called very ample. If some
multiple of D is very ample, then D is called ample.
28
There is no canonical basis for L(D). Let G = {g0 , . . . , gn } be another basis for
n
P
L(D), then gi = aij fj for i = 0, . . . , n, and φD,G = M ◦ φD,F where M is the linear
j=0
transformation of Pn determined by the coefficients aij . Therefore, we can associate a
rational map φD with a divisor D defined up to linear transformation.
For any linearly equivalent divisors D1 ∼ D2 , there exists a function g ∈ K(C)

such that D1 = D2 +div(g). Then we have L(D1 ) = {f ∈ K(C) : div(f )+D2 +div(g) ≥
0} = {f ∈ K(C) : div(f g) + D2 ≥ 0} and L(D2 ) = gL(D1 ). So if {f0 , . . . , fn } is a basis
for L(D1 ), then {gf0 , . . . , gfn } is a basis for L(D2 ). Thus, we have φD1 = φD2 since we
map into projective space.
Therefore, if D ∈ P ic(C), we can define φD : C → Pn up to linear transformation.
Fact 2.81. If L(D) and L(D0 ) are two spaces without base points with bases {f0 , . . . , fn }
and {g0 , . . . , fm }, then L(D + D0 ) is spanned by {fi gj }, where i = 1, . . . , n and j =
1, . . . , m.
Definition 2.82. A complete linear system on a variety is defined as the set of all
effective divisors linearly equivalent to some given divisor D. It is denoted by |D|.
Definition 2.83. Suppose that |D| is a complete linear system of divisors on some
variety V. The set of base points of |D| is the intersection of supports of all divisors in
L(D).
The rational map φD associated to divisor D is defined outside of the base points
of |D|.
Definition 2.84. A divisor is called base point free if |D| has no base points.
Fact 2.85 ( [10], Fact 9.15). Let D be any divisor and H be an ample divisor. Then
D + mH is base point free for m big enough. Furthermore, if D is base point free, then
D + H is very ample.
Corollary 2.86. Every divisor D can be written as a difference of two base point free
(very) ample divisors.
29
Proof. Choose a very ample divisor, say H. By Fact 2.85, for m big enough, D + mH
and mH are base point free very ample divisors. Let D = (D + mH) − mH and we
are done.
Fact 2.87. Every divisor on a curve of sufficiently high degree is very ample by Corol-
lary 2.71 (iii).
2.3.5. Primes of Good Reduction
Let K be a number field. Let C be a curve consisting of points (x0 , . . . , xn ) ∈ Pn

satisfying the homogeneous equations h1 (x0 , . . . , xn ) = 0, . . . , hk (x0 , . . . , xn ) = 0 with
k ≥ n − 1 and hi irreducible. Let C be defined over K, i.e., the coefficients of h1 , . . . , hk
lie in K. Also let f : C → P1 be a map defined over K, where f = (f0 , f1 ) and fi
homogeneous polynomials of the same degree with coefficients in K. Multiply all hi ’s
and fi ’s by a common denominator so that all polynomials have integral coefficients.
Then in the reduction of a finite prime P, we take each coefficient modulo P. We set
an algebraic number α to 0 if ordP (α) > 0, and to ∞ if ordP < 0.
In the reduction of some primes, some hi may become indeterminate or the degree
of f may decrease or f may map a point to (0, 0). Also, given a divisor D ∈ DivK (C),
some points in D may coalesce or may become indeterminate. We call such primes as
bad primes.
We give a procedure for eliminating these finitely many bad primes, which is taken
from Abc Implies Roth’s Theorem and Mordell Conjecture by Machiel van Franken-
huysen, [11]: Take an arbitrary point a ∈ P1 (K) such that f is not ramified over a
and take another point b ∈ P1 (K) different from a. If we have a = 0 or b = 0 or
a = b in the reduction, we exclude such primes. Also if the points in the divisors
P P
eP (f )(P ) and eP (f )(P ) coalesce or become indeterminate, we exclude
P ∈f −1 (a) P ∈f −1 (b)
such primes, too. Now for the remaining primes, we have a 6= b, so f is not constant,
∗
and deg(f ) = deg(f (a)) = deg(f ). Still f may map a point of C to an indeterminate
point, but this happens for only finitely many primes.
30
2.4. Elliptic Curves
In this section, we introduce elliptic curves and then give some important proper-
ties which are necessary for the Section 3, Hall conjecture. For more details and proofs,
see [8] and [12].
Definition 2.88. Let K be a field. An elliptic curve over K is a smooth cubic projective
curve defined over K that has genus 1 and a distinguished rational point O, namely the
point at infinity.
Proposition 2.89 ( [8], Chapter 3, Proposition 3.1). Let K be a perfect field and E
be an elliptic curve defined over K with the point at infinity O.
(i) There exist functions x, y ∈ K(E) such that the map φ : E → P2 with φ = [x, y, 1]
gives an isomorphism of E/K onto a curve given by a Weierstrass equation
C : y 2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6
with coefficients a1 , . . . , a6 ∈ K; and such that φ(O) = [0, 1, 0].

(ii) Conversely, every smooth cubic curve C given by a Weierstrass equation as above
is an elliptic curve defined over K with origin O = [0, 1, 0].
Proof. (i) First we look at the vector spaces L(n(O)) for n ∈ Z+ . E has genus 1, and
deg(n(O)) > 2g−2 = 0, then by Corollary 2.71 (iii), l(n(O)) = deg(n(O)) = n for
all n ≥ 1. Thus by Proposition 2.72, L(n(O)) has a basis consisting of functions
in K(E). The set {1, x} is a basis for L(2(O)) and {1, x, y} is a basis for L(3(O)).
The Riemann Roch space L(2(O)) is the vector space of functions such that
div(f ) ≥ −2(O), so functions in L(2(O)) have poles at the point of infinity of
order at most 2. Note that x must have a pole of exact order 2 at O, otherwise it
would be in L(1(O)). By similar reasoning, y must have a pole of exact order 3 at
O. Observe that the seven functions 1, x, y, x2 , xy, x3 , y 2 are all in L(6(O)), since
they satisfy div(f ) ≥ −6O. However, l(6(O)) = 6, so there must be a K-linear
relation between them.
31
Say A1 + A2 x + A3 y + A4 x2 + A5 xy + A6 y 2 + A7 x3 = 0 with Ai ∈ K. Notice that

first five terms in the sequence 1, x, y, x2 , xy, x3 , y 2 have poles of order 0, 2, 3, 4, 5,
respectively, but x3 and y 2 both have poles of order 6. Thus leaving out one
of them would give a basis for L(6(O)), so the coefficients of x3 and y 2 in the
dependence relation must be nonzero, i.e., A6 .A7 6= 0.
Replacing x and y by −A6 A7 x and A6 A7 2 y, then dividing by A6 3 A7 4 , we get
C : y 2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6 with coefficients a1 , . . . , a6 ∈ K.
This gives a map φ : E → C ⊂ P2 with P 7→ [x(P ), y(P ), 1], which is a morphism

since E is smooth and surjective since the map is not constant. Moreover, φ maps
O to [0, 1, 0] because y has a higher order pole at point at infinity than x.
Now we want to show that φ has degree one, or equivalently K(E) ∼
= K(x, y).
Consider the map [x, 1] : E → P1 . Since x has a double pole at O, using the
definition of the degree of a map by ramification, this map has degree 2. Hence
[K(E) : K(x)] = 2. In a similar way, the map [y, 1] : E → P1 has degree 3, and
so [K(E) : K(y)] = 3. By tower rule, [K(E) : K(x, y)] must divide both 2 and 3,
hence [K(E) : K(x, y)] = 1, and K(E) ∼
= K(x, y).
In addition, we want to show that C is smooth. So assume for a contradiction C
is singular. Then from Proposition 2.73, there is a rational map ϕ : C → P1 of
degree 1. Taking the composition of two maps, we get that ϕ ◦ φ : E → P1 is a
map of degree 1, but then E ∼
= P1 . We get a contradiction since E has genus 1
and P1 has genus 0. In the end, we have a map φ : E → C of degree 1 between
two smooth curves. Hence E ∼
= C.
(ii) Let E be given by a nonsingular Weierstrass equation. Using Proposition 2.67,
dx
we know that the differential w = ∈ ΩE does not have any zeros or
2y + a1 x + a3
poles, so div(w) = 0. Then, by Theorem 2.70, we get 2g − 2 = deg(div(w)) = 0,
where g is the genus of E, and E has genus 1. Also taking [0, 1, 0] as the base
point makes E into an elliptic curve.
32
Proposition 2.89 above provides us the general Weierstrass form. Then we use
some linear change of variables to obtain short Weierstrass form.
Consider the equation C : y 2 +axy+by = x3 +cx2 +dx+e. We use the substitution

a b
y 7→ y − x − to eliminate xy and y terms.
2 2
2
a2 b2

a b a b a b ab
y− x− + ax y − x − +b y− x− = y 2 − x2 − x −
2 2 2 2 2 2 4 2 4
a2

2 3 2 ab
Combining with the right hand side, we get y = x + c + x + d+ x+
4 2
b2 a2 b2

ab
e+ . Let a0 := c + , b0 := d + , and c0 := e + , then y 2 = x3 + a0 x2 + b0 x + c0 .
4 4 2 4
a0
Finally, use the substitution x 7→ x − to eliminate x2 .
3
(a0 )2 (a0 )3 2(a0 )2 (a0 )3 a0 b 0

y 2 = x 3 − a0 x 2 + x− + a0 x 2 − x+ + b0 x − + c0
3 27 3 9 3
(a0 )2 2(a0 )3 a0 b0
So we get y 2 = x3 + Ax + B with A = b0 − , B = c0 + − .
3 27 3
Elliptic curves have a special group structure in their geometric nature. Let E be
an elliptic curve defined over Q. Then (E(Q), +) is a group with the point at infinity
O as the identity element. The addition is defined as follows: Let P and Q be two
rational points in E(Q) and let L be the line joining P and Q. (If P = Q, then L is the
tangent line to E at P.) Since E is cubic, the line L and E intersect at a third point,
say R. Next we join O and R by a line and take the third intersection point with E to
be P + Q. The addition law is commutative, hence (E(Q), +) is an abelian group.
Theorem 2.90 (Mordell-Weil Theorem, [12]). E(Q) is a finitely generated abelian

group. In other words, there are points P1 , . . . , Pn such that any other point Q in E(Q)
can be expressed as a linear combination
Q = a1 P1 + a2 P2 + · · · + an Pn for some ai ∈ Z.
33
Thanks to Mordell-Weil theorem, we know the general structure of the group

completely:
E(Q) ∼
= E(Q)tors ⊕ ZRE
where E(Q)tors = {P ∈ E(Q) : ∃n ∈ N such that nP = O} and RE is the rank of E.

The first part of the decomposition is a finite group formed by the torsion points of E,
i.e., points of finite order. The second part is called the free part, and generated by
points of infinite order. The minimum number of generators gives the rank.
There is a simple algorithm to determine E(Q)tors :
Theorem 2.91 (Nagell-Lutz). Let E/Q be an elliptic curve with Weierstrass equation
y 2 = x3 + Ax + B, A, B ∈ Z. Then, every torsion point P 6= O of E satisfies:
(i) The coordinates of P are integers, i.e. x(P ), y(P ) ∈ Z.

(ii) If P is a point of order n ≥ 3, then 4A3 + 27B 2 is divisible by y(P )2 .
(iii) If P is of order 2, then y(P ) = 0 and x(P )3 + Ax(P ) + B = 0.
Let E/Q be an elliptic curve given by a Weierstrass equation y 2 = x3 + Ax + B

with A, B ∈ Z. Let p be a prime. If we reduce A, B modulo p, then we obtain the
equation of a curve Ê given by a cubic curve and defined over the field Fp . Even though
E is smooth as a curve over Q, the curve Ê may be singular over Fp .
Definition 2.92. Let E be an elliptic curve given by y 2 = x3 + Ax + B, with A, B ∈ Q.
(i) We define ∆E , the discriminant of E, by ∆E = −16(4A3 + 27B 2 ).

(ii) Let S be the set of all elliptic curves E 0 that are isomorphic to E over Q and
such that the discriminant of E 0 is an integer. The minimal discriminant of E
is the integer ∆0E that attains the minimum of the set {|∆0E | : E 0 ∈ S}. If E 0 is
the minimal model for E with minimal discriminant, we say that E 0 is a minimal
model for E.
34
Let Ê be a cubic curve over a field K with Weierstrass equation f (x, y) = 0,

where f (x, y) = y 2 + a1 xy + a3 y − x3 − a2 x2 − a4 x − a6 , and suppose that Ê has a
singular point P = (x0 , y0 ), i.e., fx (P ) = fy (P ) = 0. Thus, one can write the Taylor
expansion of f (x, y) around (x0 , y0 ) :
f (x, y) − f (x0 , y0 ) = ((y − y0 ) − α.(x − x0 )).((y − y0 ) − β.(x − x0 )) − (x − x0 )3
for some α, β ∈ K.
Definition 2.93. The singular point P ∈ Ê is a node if α 6= β. In this case there are
two different tangent lines to E at P, namely
y − y0 = α.(x − x0 ), y − y0 = β.(x − x0 ).
If α = β, then we say that P is a cusp, and there is a unique tangent line at P.
Definition 2.94. Let E/Q be an elliptic curve given by a minimal model, let p ≥ 2
be a prime and let Ê be the reduction curve of E modulo p. We say that E/Q has
good reduction modulo p if Ê is smooth and hence is an elliptic curve over Fp . If Ê is
singular at a point P ∈ E(Fp ), then we say that E/Q has bad reduction at p and we
distinguish two cases:
(i) If Ê has a cusp at P, then we say that E has additive reduction.

(ii) If Ê has a node at P, then we say that E has multiplicative reduction. If the
slopes of the tangent lines, α and β, are in Fp , then the reduction is said to be
split multiplicative; and non-split otherwise.
Proposition 2.95. Let K be a field and let E/K be a cubic curve given by y 2 = f (x),
where f (x) is a monic cubic polynomial in K[x]. Suppose that f (x) = (x − α)(x −
β)(x − γ) with α, β, γ ∈ K and put
D = (α − β)2 (α − γ)2 (β − γ)2 .
Then E is nonsingular if and only if D 6= 0.

35
The quantity D in Proposition 2.95 is the discriminant of the polynomial f (x).

The discriminant of E/Q, ∆E as in Definition 2.92, is equal to 16D. This fact with
Proposition 2.95 yields:
Corollary 2.96. Let E/Q be an elliptic curve with coefficients in Z. Let p ≥ 2 be a

prime. If E has bad reduction at p, then p|∆E . In fact, if E is given by a minimal
model, then p|∆E if and only if E has bad reduction at p.
Let E be an elliptic curve over Q given by a minimal model
y 2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6
with integer coefficients. For a prime p of good reduction for E/Q, we define Np as the
number of points in the reduction of the curve modulo p, i.e., the number of points in
E(Fp ). Also, let ap = p + 1 − Np . We define the local factor at p of the L-series to be
Lp (T ) =:

1 − ap T + pT 2 , if E has good reduction at p,






1 − T, if E has split multiplicative reduction at p,

(2.3)
1 + T, if E has non-split multiplicative reduction at p,








1, if E has additive reduction at p.
Definition 2.97. The L-function of the elliptic curve E is defined to be
Y 1
L(E, s) = −s )
,
p≥2
L p (p
where the product is over all primes p ≥ 2 and Lp (T ) is the local factor defined above.
36
Proposition 2.98. Let E/Q be an elliptic curve, and let L(E, s) be its L-function.
Define Fourier coefficients an for all n ≥ 1 as follows. Let a1 = 1. If p ≥ 2 is prime,
we define ap =:




 p + 1 − Np , if E has good reduction at p,



1, if E has split multiplicative reduction at p,

(2.4)
−1, if E has non-split multiplicative reduction at p,








0, if E has additive reduction at p.
If n = pr for some r ≥ 1, we define apr recursively using the relation
ap .apr = apr+1 + p.apr−1
if E/Q has good reduction at p and apr = (ap )r if E/Q has bad reduction at p. Finally,
if (m, n) = 1, then we define amn = am .an . Then the L-function of E can be written as
the series
X an
L(E, s) = .
n≥1
ns
Conjecture 2.99 (First part of Birch and Swinnerton-Dyer conjecture). Let E be an

elliptic curve over Q, and let L(E, s) be the L-function attached to E. Then L(E, s)
has a zero at s = 1 of order equal to the rank RE of E(Q). In other words, the Taylor
expansion of L(E, s) at s = 1 is of the form
L(E, s) = c0 .(s − 1)RE + c1 .(s − 1)RE +1 + . . .
where c0 is a non-zero constant.

37
Theorem 2.100 (Gross-Zagier, Kolyvagin, [12], Chapter 5, Theorem 5.2.8). Let E/Q
be an elliptic curve of algebraic rank RE . Suppose that the analytic rank of E/Q is ≤ 1,
i.e., ords=1 L(E, s) ≤ 1. Then the first part of Birch and Swinnerton-Dyer conjecture
holds for E/Q, i.e.,
RE = rank(E(Q)) = rankan (E/Q) = ords=1 L(E, s).
Example 2.101. Let E : y 2 = x3 − 13. Then E(Q) does not have any torsion point of
order 2 by Theorem 2.91, (ii). If (x, y) is a torsion point of E, then by Theorem 2.91, y
is an integer and y divides 4A3 +27B 2 = 27.132 . Then y(P ) = ±1, ±3, ±13, ±39, which
implies x(P )3 = 14, 22, 182, 1534, respectively. Since x(P ) is an integer, E(Q)tors is
trivial. Furthermore, the discriminant of E is ∆E = −73008 = 24 .33 .132 and E is a
minimal model since powers of the primes in the discriminant do not exceed 12. Then
by Corollary 2.96, E has bad reduction at primes 2, 3, 13.
(i) E has additive reduction at p = 2, since modulo 2 we can write the equation as
((y − 1) − 0.(x − 0))2 − x3 and the unique slope is 0.
(ii) E has additive reduction at p = 3, since modulo 3 we can write the equation as
((y − 0) − 0.(x − 1))2 − (x − 1)3 and the unique slope is 0.
(iii) E has additive reduction at p = 13, since modulo 13 we can write the equation as
((y − 0) − 0.(x − 0))2 − x3 and the unique slope is 0.
Q 1 P an a5 a7 a11 a17
Then L(E, s) = = = 1 + s + s + s + s + ...
p6=2,3,13 1 − ap p−s + p1−2s n≥1 n
s 5 7 11 17
Computing a few terms of the Taylor expansion of L-function about s = 1 and

evaluating L(E, 1) and derivatives at s = 1 in Magma, we see that L(E, 1) = 0 and the
first derivative is non-zero. This suggests that L(E, s) has a zero of order 1 at s = 1.
Then by Conjecture 2.99 and Theorem 2.100, the rank of E is 1.
38
2.5. Height Function
For the study of rational and integral points on an algebraic variety, it is required
to have a good size function to measure the “size” of a point. By good, we actually
mean two essential properties for these functions. First, the function should have
a finiteness property, in other words there should exist only finitely many points of
bounded size. Second, it should reflect the underlying geometry of the variety and the
arithmetic nature of the point. We introduce height functions here as an example of
a good size function. We mostly benefit from Generalization of the ABC-conjecture
by Nils Bruin and Diophantine Geometry: An Introduction by Silverman and Hindry.
For more details and proofs, see [9] and [13].
We start with defining height functions for points in a projective space.
2.5.1. Heights on Projective Space
Let P ∈ Pn (Q). Then we can write P = (x0 , x1 , . . . , xn ) with x0 , x1 , . . . , xn ∈ Z

and gcd(x0 , x1 , . . . , xn ) = 1. We define the height of P as H(P)=max{|x0 |, |x1 |, . . . , |xn |}.
For any M, the set {P ∈ Pn (Q)|H(P ) < M } is finite, since there are finitely
many integers x ∈ Z satisfying |x| ≤ M.
We fix K as a number field for the rest and generalize in the following way:
Definition 2.102. Let P = (x0 , x1 , . . . , xn ) ∈ Pn (K). The height of P, relative to K,

is
Y
HK (P ) = max{|x0 |v , |x1 |v , . . . , |xn |v },
v∈MK
where v ranges over all normalized valuations of K, denoted as MK .
We will denote HK (P ) as H(P ) since we fixed K.

39
We also define logarithmic height h(P ) := logH(P ), also called as additive and
the height H above is also called as the multiplicative height.
Lemma 2.103. Let P = (x0 , x1 , . . . , xn ) ∈ Pn (K).
(i) The height H(P ) is well-defined, independent of the choice of homogeneous coor-
dinates for P.
(ii) H(P ) ≥ 1 for all P ∈ Pn (K).
Proof. (i) Take another choice of coordinates for P, which is in the form of
(cx0 , cx1 , . . . , cxn ) for some c ∈ K ∗ . Using the Product Formula 2.12 and the
Q
multiplicativity of absolute values, we get max{|cx0 |v , |cx1 |v , . . . , |cxn |v } =
v∈MK
! !
Y Y
|c|v max{|x0 |v , |x1 |v , . . . , |xn |v }
v∈MK v∈MK
Y
= max{|x0 |v , |x1 |v , . . . , |xn |v }.
v∈MK
(ii) Choose homogeneous coordinates for P so that some coordinate equals 1. Then
H(P ) ≥ 1 by definition.
Definition 2.104. The height of an element α ∈ K is defined as the height of the

corresponding projective point (α, 1) ∈ P1 (K) :
Y
H(α) = max{|α|v , 1}.
v∈MK
Remark 2.105. For all r ∈ K ∗ , we have H(r) = H(1/r) = H(r − 1) + O(1). The
first equation follows from Lemma 2.103 (i) and the second follows from Definition 2.1
(iii).
40
Remark 2.106. Let P ∈ Pn (Q) and let σ ∈ GQ , which is the Galois group on Pn (Q).
Then Hσ(K) (σ(P )) = HK (P ), since the absolute value σ(v) ∈ Mσ(K) is defined by
|σ(x)|σ(v) = |x|v .
The automorphism σ of Q defines an isomorphism between K and σ(K), and it

likewise identifies the sets of absolute values on K and σ(K).
Recall that the field of definition of a point P = (x0 , x1 , . . . , xn ) ∈ Pn (Q) is the

field Q(P ) = Q(x0 /xj , x1 /xj , . . . , xn /xj ) for any j with xj 6= 0.
Theorem 2.107. For any number M ≤ 0 and for any fixed number field K, the set
{P ∈ Pn (K)|H(P ) ≤ M } is finite.
Proof. Choose homogeneous coordinates for P = (x0 , x1 , . . . , xn ) so that some coordi-

nate equals 1. Then for any absolute value v, we have
max{|x0 |v , |x1 |v , . . . , |xn |v } ≥ max{|xi |v , 1}
for all 0 ≤ i ≤ n. Multiplying over all v ∈ MK , we get H(P ) ≥ H(xi ) for all 0 ≤ i ≤ n.
Hence it is enough to show that the set {x ∈ Q|H(x) ≤ M } is finite.
Let x ∈ Q have degree d and let K = Q(x). Denote the conjugates of x over Q
as x1 , x2 , . . . , xd . Let
d
Y
fx (t) = (t − xj )
j=1
be the minimal polynomial of x over Q. We can write this minimal polynomial fx (t)
so that its coefficients are composed of symmetric polynomials, say sr (x):
d
X
(−1)r sr (x)td−r .
r=0
41
Then for any absolute value v ∈ MK , using Definition 2.1 (iii), we have

X
|sr (x)|v = x i 1 . . . xi r

1≤i1 <···<ir ≤d v
≤ Cv max1≤i1 <···<ir ≤d |xi1 . . . xir |v ≤ Cv max(|xi1 |v , . . . , |xir |v )r ,
d

where Cv = 1 when v is nonarchimedean and Cv = r
when v is archimedean. Then
d
Y
max{|s0 (x)|v , . . . , |sd (x)|v } ≤ C max{|xi |v , 1}d ,
i=1
d

where we can take C = 2d ≥
P
r
. Multiplying over all absolute values v ∈ MK ,
1≤r≤d
we estimate
d
Y
H(s0 (x), . . . , sd (x)) ≤ 2 d
H(xi )d .
i=1
2
Moreover, we can write H(s0 (x), . . . , sd (x)) ≤ 2d H(x)d by the Remark 2.106.
Finally, let x ∈ {x ∈ Q|H(x) ≤ M }. Then x is the root of a polynomial fx (t) ∈

2
Q[t] whose coefficients s0 , . . . , sd satisfy HK (s0 (x), . . . , sd (x)) ≤ 2d M d . Since there are
only finitely many points of bounded height in Pn (Q), there are also finitely many
possibilities for the polynomial fx (t), which gives us only finitely many possibilities for
x. Hence the set {x ∈ Q|H(x) ≤ M } is finite.
This property of height functions is an essential key for the proofs of many funda-
mental finiteness theorems in Diophantine geometry such as Faltings’ theorem, Siegel’s
theorem and Mordell-Weil theorem.
42
Now we define a product. Let x = (x0 , . . . , xn ) ∈ Pn (K) and y = (y0 , . . . , ym ) ∈

Pm (K). We define
x ⊗ y := (x0 y0 , . . . , x0 ym , . . . , xn y0 , . . . , xn ym ) ∈ P(n+1)(m+1)−1 (K),
xr := (x0 r , . . . , xn r ) and x(r) := x ⊗ · · · ⊗ x.
Proposition 2.108. Let x = (x0 , . . . , xn ) ∈ Pn (K) and y = (y0 , . . . , ym ) ∈ Pm (K) be

projective points. Then we have the following properties:
(i) H(x ⊗ y) = H(x)H(y)

(ii) H(xr ) = H(x(r) ) = H(x)r
(iii) Let f : Pn (K) → Pm (K) be a rational map of degree d. Then there exists an
effectively computable constant C = Cf such that for all x ∈ Pn (K) where f is
regular, we have H(f (x)) ≤ CH(x)d .
Q Q
Proof. (i) H(x ⊗ y) = maxi,j |xi yj |v = maxi,j |xi |v |yj |v
v∈MK v∈MK
Y Y
= maxi |xi |v maxj |yj |v = H(x)H(y).
v∈MK v∈MK
(ii) Let |xi0 |v = maxi |xi |v , then |xi0 |v r = maxi1 ,...,ir |xi1 . . . xir |v . Hence H(xr ) =
H(x)r . The second equality H(x(r) ) = H(x)r follows from (i).
(iii) Let f = (f0 , . . . , fm ) be a representation of f where fi are homogeneous polyno-
mials of degree d. Then for all x ∈ Pn (K) where f is regular, we can find a fi so
that fi (x) 6= 0. We can write fi as
fi (x) = (ci,1 , . . . , ci,t ).x(d) ,
where t is the number of entries in the vector x(d) , and the product . is the
standard inner product of vectors. (We give an example below to understand
this representation better, please see Example 2.110.) Let ci := (ci,1 , . . . , ci,t ).
Then by Definition 2.1 (iii), since there exist t terms in the sum of monomials,
43
there exists a constant Cv such that
Y Y
H(f (x)) = maxi |ci .x(d) |v ≤ Cv t−1 maxi (maxj |ci,j |v |(x(d) )j |v ),
v∈MK v∈MK
where (x(d) )j denotes the j−th place. Then
!
Y Y
Cv t−1 maxi (maxj |ci,j |v |(x(d) )j |v ) ≤ Cv t−1 H(c0 , . . . , cm )H(x)d .
v∈MK v∈MK
Here we get Cv > 1 for only finitely many v ∈ MK , namely for archimedean
absolute values, hence it is a finite product and by Product Formula 2.12, we
Cv t−1 = 1. Letting C := H(c0 , . . . , cm ), we are done.
Q
have
v∈MK
Remark 2.109. Linear transformations of Pn (K) change the logarithmic height by

only a bounded function, i.e., h(f (x)) ≤ h(x) + O(1). In particular, logarithmic height
is only dependent on the choice of the coordinates by a bounded function.
Example 2.110. Let f : P2 (K) → P1 (K) and let f = (f0 , f1 ) with f0 (x0 , x1 , x2 ) =
x1 2 + x0 x2 and f1 (x0 , x1 , x2 ) = x2 2 . One can write f0 as
f0 (x) = (ci,1 , . . . , ci,9 ).x(2) =
(ci,1 , . . . , ci,9 ).(x0 x0 , x0 x1 , x0 x2 , x1 x0 , x1 x1 , x1 x2 , x2 x0 , x2 x1 , x2 x2 ),
where c03 = c05 = 1 and the other coefficients are 0.
Let x ∈ P2 (K) where f is regular, say f0 (x) 6= 0. Then we have H(f (x)) ≤
CH(x)2 for some constant C.
44
Proposition 2.111. Let F be a homogeneous polynomial over K of degree d in vari-

ables X0 , . . . , Xn with the coefficient of Xn is nonzero, i.e., the point (0, . . . , 0, 1) does
not satisfy the equation of F. Let π be the projection from (0, . . . , 0, 1) defined by
π : Pn → Pn−1 with (x0 , . . . , xn ) 7→ (x0 , . . . , xn−1 ). Then there exists an effectively
computable constant C dependent on F such that for any x ∈ Pn (K) on the hypersur-
face determined by F, we have
h(x) − C ≤ h(π(x)) ≤ h(x).
Proof. For any x ∈ Pn (K) on the hypersurface determined by F, we have F (x) = 0.

Then xn d can be written as a linear combination of other monomials. The point
x(d−1) ⊗ π(x) contains all those monomials. Indeed, it contains all monomials of degree
d in n + 1 variables but with the degree of xn strictly less than d. Since we can write
xn d using monomials in x(d−1) ⊗ π(x), there exists some linear map determined by F
sending x(d−1) ⊗ π(x) to x(d) . Then by Proposition 2.108 (iii), there exists a constant
C such that
h(x(d) ) ≤ h(x(d−1) ⊗ π(x)) + C.
Finally, using Proposition 2.108 (ii), we get h(x) ≤ h(π(x)) + C. The second inequality
h(π(x)) ≤ h(x) follows directly from the definition.
We can generalize Proposition 2.111 for any projection from a point not on the
hypersurface with a slight adjustment h(x) − C1 ≤ h(π(x)) ≤ h(x) + C2 , since the
logarithmic height depends on the choice of coordinates by a bounded function by
Remark 2.109.
Theorem 2.112. Let f : Pn (K) → Pm (K) be a morphism of degree d. Then there

exist effectively computable constants C1 , C2 such that for x ∈ Pn (K), we have
h(f (x)) − C1 ≤ dh(x) ≤ h(f (x)) + C2 .

45
Proof. We only need to show dh(x) ≤ h(f (x))+C2 , since we showed the first inequality
in Proposition 2.108. Let f = (f0 , . . . , fm ) be a representation of f where fi are
homogeneous polynomials of degree d. Let p : Pn (K) → Pt (K) with x 7→ x(d) . Then
fi (x) are linear combinations of the coordinates of p(x). Since fi ’s have no common
zero, we can write the morphism as f = π ◦ p where π is a combination of projections
from points outside f (Pn ) and linear transformations. Notice that we take projections
from points outside f (Pn ) in order to not lose any points, since in the end we project
into f (Pn ). Then by Proposition 2.111, we have h(π(p(x))) ≥ h(p(x)) − C2 for x ∈
Pn (K). Lastly, using Proposition 2.108 (ii) with h(p(x)) = h(x(d) ) = dh(x), we get
h(f (x)) = h(π(p(x))) + C2 ≥ dh(x).
2.5.2. Heights on Curves
Let V be a smooth projective variety defined over K. In order to satisfy the second
essential property of height functions, we begin by defining a height function for each
projective embedding of V. Then using the relationship between projective embeddings
and divisors, we get an equivalence class of height functions for each divisor class on
V. Therefore we get information about the rational points on V by taking from the
structure of its divisor class group. This construction is called Weil’s Height Machine.
We restrict ourselves to curves for the construction. Let C be a curve over K.

For every morphism f : C → Pn , we have a height on C defined by hf := h ◦ f where
h is the logarithmic height on Pn (K).
For every very ample divisor, we have such a morphism, defined up to a linear
transformation, since φD is regular and injective. Then by Theorem 2.112, we get
a height function for every very ample divisor, defined up to a bounded function,
hC,D = h ◦ φD + O(1). This determines the height function for very ample divisors.
Now since linearly equivalent divisors have the same associated morphism, they
have the same associated height: if D ∼ D0 , then hD = hD0 .
46
Moreover, by Fact 2.81, we have L(D1 +D2 ) = L(D1 )⊗L(D2 ) for divisors without
base points, particularly for very ample divisors. Then for the associated morphisms,
we have φD1 +D2 = π◦(φD1 ⊗φD2 ) for some linear map π, which is regular on the image of
C. Using Proposition 2.108 (i) and Proposition 2.111, we get hD1 +D2 = hD1 +hD2 +O(1),
the additivity property.
We associate equivalence classes of height functions, i.e., height functions modulo

bounded functions, first with all very ample divisors and then with all divisors by
enforcing additivity, since any divisor can be written as the difference of very ample
divisors by Fact 2.86.
Let D ∈ Div(C), then using Fact 2.86, we write D = D1 − D2 where D1 , D2 are

very ample divisors. Then we define hC,D (P ) = hC,D1 (P ) − hC,D2 (P ) for all P ∈ C(K).
This gives us a height function hC,D for every divisor D on every curve C.
Next we check the additivity property, which we already know for base point free
divisors. Let D, E be arbitrary divisors and we write them as differences D = D1 − D2
and E = E1 − E2 of base point free divisors using Fact 2.86. Then D1 + E1 and D2 + E2
are base point free, so hC,D+E = hC,D1 +E1 − hC,D2 +E2 + O(1) = hC,D1 + hC,E1 − hC,D2 −
hC,E2 + O(1) = hC,D + hC,E + O(1).
We are now ready to give Weil’s construction that associates a height function to
every divisor.
Theorem 2.113 (Weil’s Height Machine, [13]). Let K be a number field. For every
smooth projective variety V /K there exists a map
hV : Div(V ) → {functions V (K) → R}
with the following properties:

47
• (Normalization) Let H ⊂ Pn be a hyperplane, and let h(P ) be the absolute loga-

rithmic height on Pn . Then
hPn ,H (P ) = h(P ) + O(1) for all P ∈ Pn (K).
• (Functoriality) Let φ : V → W be a morphism and let D ∈ Div(W ). Then
hV,φ∗ D (P ) = hW,D (φ(P )) + O(1) for all P ∈ V (K).
• (Additivity) Let D, E ∈ Div(V ). Then
hV,D+E (P ) = hV,D (P ) + hV,E (P ) + O(1) for all P ∈ V (K).
• (Linear Equivalence) Let D, E ∈ Div(V ) with D linearly equivalent to E. Then
hV,D (P ) = hV,E (P ) + O(1) for all P ∈ V (K).
• (Positivity) Let D ∈ Div(V ) be an effective divisor, and let B be the base locus
of the linear system |D|. Then
hV,D (P ) ≥ O(1) for all P ∈ (V − B)(K).
• (Finiteness) Let D ∈ Div(V ) be ample. Then for every finite extension K 0 /K

and every constant M, the set
{P ∈ V (K 0 )|hV,D (P ) ≤ M }
is finite.
• (Uniqueness) The height functions hV,D are determined, up to O(1), by normal-
ization, functoriality just for embeddings φ : V ,→ Pn , and additivity.
48
Remark 2.114. If the variety V is not smooth, Weil’s Height Machine is still valid,
provided that we work with Cartier divisors instead of Weil divisors. For the theory of
Cartier divisors, see [13].
The height associated to the zero divisor of a rational function is essentially the
same as the height function induces itself.
Proposition 2.115. Let C be a curve over K and let f be a rational function on that
curve. Let (f )0 be the zero divisor of f. Then

1
h(f (x)) = h (x) = h(f )0 (x) + O(1).
f
Proof. Since (f )0 is positive, using Fact

2.112,
we choose an m such that m(f )0 is very
1 1
ample. We have m ∈ L(m(f )0 ), since = (f )0 . Moreover, 1 ∈ L(m(f )0 ), since
f f ∞
1
(f )0 is positive. Now we take a basis 1, m , g1 , . . . , gs for L(m(f )0 ). None of gi ’s has
f
1
a pole of higher order than m , then (0, 0, x1 , . . . , xs ) ∈
/ V for any xi if V is identified
f
with its embedding. By Proposition 2.111, we know that the projection onto the first
two coordinates changes the height by a bounded function. Therefore,

1 1
h(f (x)) = h 1, (x) = h 1, (x), g1 (x), . . . , gs (x) + O(1).
f f
Since the height associated to a divisor is defined via the projective embedding, we are
done.
Theorem 2.116 ( [14], Page 45). Let V be a nonsingular projective variety over K,
c ∈ P ic0 (V ), and c0 an element of P ic(V ) which is ample. Then on V (K),
p
hc ≤ O( hc0 ) + O(1).
49
Theorem 2.117 ( [15], Chapter 4, Corollary 3.5). Let C be a curve over K and let
D1 , D2 ∈ DivK (C). Then for every > 0, there exist constants C1 , C2 such that for all
P ∈ C, it holds that
(1 − )deg(D2 )hD1 (P ) + C1 ≤ deg(D1 )hD2 (P ) ≤ (1 + )deg(D2 )hD1 (P ) + C2 .

50
2.6. Belyi’s theorem
For curves over number fields, we have a theorem called Belyi showing that we
can control where a function is ramified. Moreover, the proof of Belyi is effective, in
other words, this function can be constructed explicitly.
Definition 2.118. A smooth algebraic curve C is defined over a subfield K of C if it

is isomorphic to the set of zeros, in some affine or projective space over C, of a finite
set of polynomials with coefficients in K. We call K a field of definition of C.
Example 2.119. The Fermat curve Fn , which is the zero locus of xn + y n + z n = 0, is

defined over Q for all n > 0.
If a curve C is defined over Q, then finitely many coefficients of the defining

polynomials of C all lie in some finite extension of Q, i.e., in a number field. This gives
us the lemma below.
Lemma 2.120. An algebraic curve is defined over a number field if and only if it is
defined over Q.
Theorem 2.121 (Belyi, [16], Theorem 2.6.1). Let C be an algebraic curve. Then C is
definable over Q if and only if there exists a morphism f : C → P1 unramified outside
{0, 1, ∞}.
For the proof of Mordell conjecture using abc conjecture, we merely need the “only
if” part of Belyi’s theorem, which results from a surprisingly simple construction.
Before proving Belyi, we prove another theorem, which is in fact Belyi restricted
to the case when C = P1 , and then Belyi easily follows.
Theorem 2.122. If S ⊆ P1 is a finite set, then there is a map φ : P1 → P1 , defined

over Q, and ramified only over 0, 1, ∞ such that φ(S) ⊆ {0, 1, ∞}.
51
Proof. Let S ⊆ P1 be a finite set and φ : P1 → P1 be a nonconstant morphism. Denote

the set Sφ := φ(S) ∪ {x ∈ P1 : φ is ramified over x}. Let α ∈ S be a nonrational point
in Q and let αn + a1 αn−1 + · · · + an = 0 be its minimal equation over Q. Assume that
n is the maximal degree over Q of the elements of S and the number of such elements
with degree n over Q is p. We take φ(z) = z n + a1 z n−1 + · · · + an , which is defined over
Q, and consider the set
Sφ = φ(S) ∪ {x ∈ P1 : φ is ramified over x} = (2.5)
φ(S) ∪ {∞} ∪ {φ(z) : φ0 (z) = 0} = φ(S) ∪ {∞} ∪ φ(S 0 ) (2.6)
where S 0 consists of z ∈ Q such that φ0 (z) = 0, hence S 0 contains elements of degree

≤ n − 1. The points of φ(S) have degree ≤ n and the number of these elements of
degree n is ≤ p − 1 since φ(α) = 0. Therefore, Sφ contains less elements of degree n
than S. By repeating this step, S will contain only rational points eventually. We may
1
assume that {0, 1, ∞} ⊆ S, because if a ∈ S, then z 7→ is unramified and it maps
z−a
a to ∞, in other words, it does not affect the critical values, just translate the value.
Similarly, if {a, ∞} ⊆ S, the map z 7→ z − a is unramified and it maps a, ∞ to 0, ∞.
Finally, if {a, 0, ∞} ⊆ S, then z 7→ z/a is unramified and it maps a, 0, ∞ to 1, 0, ∞.
Now let {0, 1, ∞, β} ⊆ S and |S| = n ≥ 4 with β be a rational number different

from 0, 1, ∞. We seek a map φ(z) = z A (1 − z)B with A, B ∈ Q, A, B, A + B 6= 0.
Then φ(0), φ(1), φ(∞) ∈ {0, ∞}. Ramified points of the map are the points z such that
φ0 (z) dz dz A
φ0 (z) = 0, or = A +B = 0 since x 6= 0, 1, ∞, which gives z = .
φ(z) z z−1 A+B
A
Now choose A, B integers so that α = . Then
A+B
Sφ = φ(S) ∪ {z ∈ P1 : φ is ramified over z} ⊆ {0, ∞, φ(α)} ∪ φ(S 0 )
where S 0 = S −{0, 1, ∞, α}. So φ(S 0 ) can have at most n−4 elements, and |Sφ | ≤ n−1.
Therefore, Sφ contains less elements than S. By repeating this step, S will only contain
{0, 1, ∞} eventually.
52
Proof of Belyi 2.121. Let f be any nonconstant rational function C → P1 defined over
a number field K ⊂ Q. Let S ⊂ P1 (Q) be the finite set of ramification points of f .
Using Theorem 2.122, if we take π = φ ◦ f, we are done.
Example 2.123. Let C : y 2 = x7 − 1 be defined over Q. We need a meromorphic

function on C, so we start with the projection
C → P1
(x, y) 7→ x
which is ramified at points (ξ7 k , 0), k = 0, 1, . . . , 6 and (∞, ∞). Indeed, in the affine
plane with z = 1, the map is only ramified where the polynomial of x vanishes, by
Example 2.40 of Kummer extension, which are the 7th roots of unity. On the other
hand, for the homogeneous polynomial Y 2 Z 5 = X 7 − Z 7 , when Z = 0, we have X 7 = 0.
Hence, we get the ramified point ∞ = (0, 1, 0) 7→ (1, 0) = ∞.
We need to eliminate the critical values ξ7 k , where k = 1, . . . , 6. So we decompose

this projection with the function x 7→ x6 + x5 + x4 + x3 + x2 + x + 1 which sends ξ7 k to
the rational number 0 for k = 1, . . . , 6, and ∞ 7→ ∞, 1 7→ 7. For the affine plane with
w = 1, we have the corresponding map C[x] → C[x]. Let M be a maximal ideal of C[x].
Using Proposition 2.35, M is ramified over C[x] if and only if f 0 (α) ∈ M where α is a
root of f. Here since f 0 = 1, and 1 ∈
/ M for any maximal ideal, the map is unramified.
In the other affine plane with x = 1, we get (1, w) 7→ (w6 +w5 +w4 +w3 +w2 +w+1, w6 ),
w6
or w 7→ 6 . When w = 0, we have 0 7→ 0 with w6 = 0.
w + w5 + w4 + w3 + w2 + w + 1
Hence the map P1 → P1 is only ramified over (1, 0) = ∞.
In order to deal with the resulting critical value 7, we now apply the function
x 7→ x/7, which is unramified, since the degree is 1, and 7, 0, ∞ 7→ 1, 0, ∞.
53
The composition
x 6 + x5 + x4 + x3 + x2 + x + 1
f : (x, y) 7→ x 7→ x6 + x5 + x4 + x3 + x2 + x + 1 7→
7
maps the first critical values 1, ∞, ξ7 k , with k = 1, . . . , 6 into {0, 1, ∞} as 1 7→ 1,

∞ 7→ ∞, ξ7 k 7→ 0. Hence we have a Belyi function f.
√ √
Example 2.124. Let C : (y − 2)2 = (x5 − 1) be defined over Q( 2). We need a
meromorphic function on C, so we start with the projection
C → P1
(x, y) 7→ x
√
which is ramified at points (ξ5 k , 2), k = 0, 1, . . . , 4 and (∞, ∞). Indeed, in the affine
plane with z = 1, the map is only ramified where the polynomial of x vanishes, by
Example 2.40 of Kummer extension, which are the 5th roots of unity. On the other
√
hand, for the homogeneous polynomial Y 2 Z 3 − 2 2Y Z 4 − X 5 + 3Z 5 = 0, when Z = 0,
we have X 5 = 0. Hence, we get the ramified point ∞ = (0, 1, 0) 7→ (1, 0) = ∞.
We need to eliminate the critical values ξ5 k , where k = 1, . . . , 4. So we decompose

this projection with the function x 7→ x4 + x3 + x2 + x + 1 which sends ξ5 k to the rational
number 0 for k = 1, . . . , 4, and ∞ 7→ ∞, 1 7→ 5. For the affine plane with w = 1, we
have the corresponding map C[x] → C[x]. Let M be a maximal ideal of C[x]. Using
Proposition 2.35, M is ramified over C[x] if and only if f 0 (α) ∈ M where α is a root
of f. Here since f 0 = 1, and 1 ∈
/ M for any maximal ideal, the map is unramified.
In the other affine plane with x = 1, we get (1, w) 7→ (w4 + w3 + w2 + w + 1, w4 ), or
w4
w 7→ 4 . When w = 0, we have 0 7→ 0 with w4 = 0. Hence the map
w + w3 + w2 + w + 1
P1 → P1 is only ramified over (1, 0) = ∞.
In order to deal with the resulting critical value 5, we now apply the function
x 7→ x/5, which is unramified, since the degree is 1, and 5, 0, ∞ 7→ 1, 0, ∞.
54
The composition
x4 + x3 + x2 + x + 1
f : (x, y) 7→ x 7→ x4 + x3 + x2 + x + 1 7→
5
maps the first critical values 1, ∞, ξ5 k , with k = 1, . . . , 4 into {0, 1, ∞} as 1 7→ 1,

∞ 7→ ∞, ξ5 k 7→ 0. Hence we have a Belyi function f.
55
3. ABC CONJECTURE
In the first chapter into the main subject, we start by giving two equivalent
statements of abc conjecture and a few examples of different cases.
Definition 3.1. The radical of a positive integer n is the product of its distinct prime
factors:
Y
Rad(n) = p.
p|n
Conjecture 3.2 (abc conjecture, 1, [17], Page 27). For all > 0, there exists a constant
λ > 0 such that, if a, b and c in Z+ are relatively prime and satisfy a + b = c, then
c = max(a, b, c) < λ Rad(abc)1+ . (3.1)
Conjecture 3.3 (abc conjecture, 2). For all > 0, there exist only finitely many triples
(a, b, c) of coprime positive integers with a + b = c satisfying
c = max(a, b, c) > Rad(abc)1+ .
Both statements can be easily obtained from each other.
Conjecture 3.3 implies conjecture 3.2: For a given , if there are only finitely many
(a, b, c) triples for which Rad(abc)1+ < c, then there are only finitely many values
of c/Rad(abc)1+ , and we can choose the maximum such value, call it m. Letting
λ := 1 + m, we are done.
Conjecture 3.2 implies conjecture 3.3: For a given , we have c < λ Rad(abc)1+ .

If λ < 1, it is trivial. If not, let 0 := . Then there exists a positive constant λ0
2
56
0
0 c1/1+ 1 1
such that c < λ0 Rad(abc)1+ ,or 0 1/1+0 < Rad(abc). Note that 0
= +
(λ) 1+ 1+
1
and = 0
since 0 = . Thus for c large
(1 + )(2 + ) (1 + )(2 + ) 1+ 2 + 2 2
/2+2 0
enough, i.e., c > λ , we have
0 0 1/1+0
c1/1+ c1/1+ c/2+2

Rad(abc) > 0 1/1+0 > = c1/1+ > c1/1+ ,
(λ ) λ0 λ0
hence c < Rad(abc)1+ . Since λ0 is a finite number, there are only finitely many c
for which c/2+2 ≤ λ0 . So there are only finitely many (a, b, c) triples satisfying c >
Rad(abc)1+ .
In the equation a+b = c, where a, b and c are relatively prime positive integers, the
case c Rad(abc)1+ happens most of the time, so we call it usual. Roughly speaking,
if there are lots of primes on the left side of the equation, then we get only a few
primes on the right side. Otherwise, we call it unusual. When we take the exponent
of the radical of abc to be 1 in the conjecture, there are infinitely many exceptions,
i.e., unusual cases, and we cannot find a constant that bounds c. However, when the
exponent is > 1, there exist only finitely many exceptions, i.e., unusual cases. Hence
one can list them.
We give a few examples of usual and unusual cases:
Example 3.4. Take a = 256, b = 27, so c = 283, where 283 is prime. One can write
as 28 + 33 = 283. Then Rad(abc) = 2.3.283 = 1698, and we have 283 < 1698. This is
the usual case.
Example 3.5. Take a = 625 = 54 , b = −1, so c = 624 = 24 .39. Then Rad(abc) =

2.5.39 = 390, and we have 625 > 390. This is unusual.
Example 3.6. Take a = 6561, b = 1, c = 6562, then 38 + 1 = 2.17.193. Since there are
prime numbers dividing the left hand side too many times, their presence is balanced
out by larger primes dividing the right hand side only a few times.
57
3.1. The Origin of Abc: Mason-Stothers Theorem
Mason-Stothers theorem is an analogy of abc conjecture with polynomials.
Theorem 3.7 (Mason-Stothers Theorem, [18]). Let f, g, h ∈ C[t] be nonconstant rel-

atively prime polynomials satisfying f + g = h. Then
max(deg(f ), deg(g), deg(h)) ≤ n0 (f gh) − 1
where n0 denotes the number of distinct roots.
Instead of the original proof of the theorem, we give the version of Synder’s.
Lemma 3.8. Let f be a nonzero polynomial in C[t], then deg(f ) = deg(f, f 0 ) + n0 (f ).
Proof. If f is constant, then the proof is trivial. So assume that f is not constant, and
let β be a root of f. We write f as a sum of powers of (t − β) since (t − β)|f.
f (t) = c1 (t − β)a1 + · · · + cn (t − β)an
where constants ci > 0 and ai > 0. Then order the multiplicities ai in a decreasing
manner so that an is the smallest.
Taking the derivative, we get f 0 (t) = a1 c(t − β)a1 −1 + · · · + an c(t − β)an −1 , then
(t − β)an −1 is the largest power of (t − β) dividing f 0 . Apply this to all of the roots of
f.
Let β1 , . . . , βr be distinct roots of f. Using the unique factorization in C[t], we

write f (t) = c(t − β1 )b1 . . . (t − βr )br with some 0 6= c ∈ C. Since the only factors of
degree 1 that could be shared by f and f 0 are of the form (t − βi ) for some i, we write
(f, f 0 ) = c0 (t − β1 )k1 . . . (t − βr )kr for some c0 ∈ C and ki ≥ 0, and in fact, ki = bi − 1
for each 1 ≤ i ≤ r by the reasoning above.
58
Then deg(f ) = b1 + · · · + br = (b1 − 1) + · · · + (br − 1) + r = deg(f, f 0 ) + n0 (f ),

and we are done.
Proof of Mason-Stothers Theorem. We have f +g = h, by taking the derivative of each

side we get f 0 + g 0 = h0 . Then
f 0 g − f g 0 = f 0 (f + g) − f (f 0 + g 0 ) = f 0 h − f h0 .
Now, we have three relations: (f, f 0 )|f 0 h − f h0 , (g, g 0 )|f 0 h − f h0 , and (h, h0 )|f 0 g − f g 0 .
Since f, g, h are relatively prime, we get (f, f 0 ).(g, g 0 ).(h, h0 )|f 0 g − f g 0 . Then
deg(f, f 0 ) + deg(g, g 0 ) + deg(h, h0 ) ≤ deg(f 0 g − f g 0 ) ≤ deg(f ) + deg(g) − 1,
and adding deg(h) to the leftmost and rightmost sides gives us
deg(f ) − deg(f, f 0 ) + deg(g) − deg(g, g 0 ) + deg(h) − deg(h, h0 ) − 1 ≥ deg(h)
so n0 (f ) + n0 (g) + n0 (h) − 1 ≥ deg(h). Finally, since they are relatively prime, we have
n0 (f gh) = n0 (f ) + n0 (g) + n0 (h), so the result follows.
One can also derive Fermat’s last theorem for polynomials from Mason-Stothers
theorem effortlessly.
59
4. HALL CONJECTURE
Consider the equation x3 − y 2 = k, where x, y, and k ∈ Z. This equation is known

as Mordell equation. In the case of k = 0, we have infinitely many solutions which
can be seen by parametrizing with (t2 , t3 ), where t is an integer. However, when k is
nonzero, we have only finitely many solutions by Mordell’s theorem in 1922.
Theorem 4.1 (Mordell’s theorem, [19]). For each k ∈ Z−{0}, the equation y 2 = x3 +k
has finitely many integer solutions.
Later in 1929, this finiteness was handled by a more general theorem about elliptic
curves by Siegel. Mordell curve is, in fact, an elliptic curve, hence Siegel’s theorem
contains Mordell curves. For more details, see Section 2.4.
Theorem 4.2 (Siegel’s Theorem). Every elliptic curve has only finitely many integer
points.
Before giving the statement of the main conjecture, Hall conjecture, on these
solutions, we give a few examples on their computations.
Recall that for odd primes p, −1 is equivalent to a square mod p if and only if p
is equivalent to 1 mod 4.
Example 4.3. Let E : y 2 = x3 + 511. Assume for a contradiction there exists an

integral solution (x, y) on E. If x is even, then y 2 ≡ 511 ≡ 3 mod 4, which is not
possible. Hence x is odd, and y is even. Note that x3 ≡ −511 ≡ 1 mod 4, then x ≡ 1
mod 4 since x is odd. Write the equation as:
y 2 + 1 = x3 + 512 = (x + 8)(x2 − 8x + 64).
We have x2 − 8x + 64 ≡ 59 ≡ 3 mod 4, and x2 − 8x + 64 = (x − 4)2 + 48 > 0,

then there exists a prime p ≡ 3 mod 4 such that p | x2 − 8x + 64, so p | y 2 + 1. It
60
follows that −1 is a square mod p which can happen only when p is 1 mod 4. It is
a contradiction. Hence the elliptic curve does not have any integral points other than
the point at infinity O.
Example 4.4. The integral solutions of y 2 = x3 + 4 are (0, ±2), which can be seen
easily by substituting x = 0.
Example 4.5. Let E : y 2 = x3 − 13. If x is even, then y 2 ≡ −13 ≡ 3 mod 4, which

is a contradiction. So x is odd, and y is even. Then y = 2m and x = 2n + 1 for some
m, n ∈ Z. Plug in, we get
4m2 = 8n3 + 12n2 + 6n − 12.
Looking at the equation mod 4, we get n ≡ 0 mod 4, hence n = 4k for some k ∈ Z,

and x = 8k + 1. Again plug in, we get
4m2 = 512k 3 + 192k 2 + 24k − 12,
then divide the both sides by 4 :
m2 = 128k 3 + 48k 2 + 6k − 3.
Looking at the equation mod 4, we see that m2 ≡ 2k + 1 mod 4. The system is not
solvable for m2 ≡ 0 mod 4. When m2 ≡ 1 mod 4, we have k ≡ 2 mod 4. Trying
the equations for k = 2, we get two integral points on our Mordell curve: (17, 70) and
(17, −70).
Moreover, these are the the only integral points on the curve. In Example 2.101,
we find that the torsion group of E is an abelian group of order 1, and the rank of E is
1. So E has no torsion points other than the point at infinity O and E(Q) ∼
= Z. Then
two rational points (17, 70) and (17, −70) are of infinite order. Moreover, (17, 70) is
a generator for the group E(Q). That means one can find all of the rational points
using the group structure with (17, 70). Since it is too messy to do by hand, we use the
61
code IntegralPoints(E) on Magma, and this is exactly what Magma does to search for
integral points; it adds the rational points and checks whether they are integral or not.
However, Magma does not do this process forever, it uses some bounds on its search
criteria to shorten the time.
Figure 4.1. An example of Mordell curve
In Figure 4.1 above we see our example, the curve given by y 2 = x3 − 13, with
its two and only integral points.
There are many conjectures on integral solutions of Mordell equations or more

generally elliptic curves, but the most famous one for Mordell equations is Hall con-
jecture, which was originally formulated by Marshall Hall, Jr. in 1970.
Conjecture 4.6 (Hall Conjecture, [19]). There is a constant C > 0 such that if y 2 =
x3 + k in Z with k 6= 0 then |x| ≤ C|k|2 and |y| ≤ C|k|3 .
62
Although it seems like a useful conjecture at first for making our job easier on
listing the integral solutions on a Mordell curve, integral points can be quite large with
respect to k. Very simple equations with small k values can have really large integral
solutions.
Example 4.7. 4478849284284020423079182 = 58538865167812233 − 1641843, found

by Noam Elkies in 1998.
63
4.1. Abc Implies Weaker Hall
We state a weaker version of Hall conjecture which can be derived from abc
conjecture.
Conjecture 4.8 (Weak Hall Conjecture, [19]). Pick > 0. There is C > 0 such
that for each k ∈ Z − {0}, if (x, y) is a point on the curve y 2 = x3 + k in Z then
|x| ≤ C |k|2(1+) , |y| ≤ C |k|3(1+) .
Theorem 4.9. Assume abc conjecture, then for all > 0 there is C > 0 such that
whenever y 2 = x3 + k in Z with k 6= 0,
|x| ≤ C |k|2(1+) , |y| ≤ C |k|3(1+) .
Proof. Note that x and y cannot be both zero, since k is non-zero. If one of them is
zero, then it is trivial. Hence assume x, y 6= 0. Let n = gcd(x3 , y 2 ). Divide both sides
of Mordell equation by n,
y2 x3 k
= + .
n n n
x3 k y2
Let a = , b = , and c = . Now we have c = a + b, using abc conjecture we get
n n n
two inequalities:
|x|3 |y|2
≤ λ Rad(abc)1+ , ≤ λ Rad(abc)1+ , so |x|3 , |y|2 ≤ nλ Rad(abc)1+ .
n n
Y Y |k|
For the radical, we have Rad(abc) ≤ p. p ≤ |x||y|Rad(b) ≤ |x||y| .
n
p|ac p|b
Plugging into above inequality, we get
1+
3 2 |x||y||k|
|x| , |y| ≤ nλ < λ (|x||y|)1+ |k|1+ .
n
Considering Mordell equation, we have either |y|2 ≤ |x|3 or |x|3 ≤ |y|2 .

64
Case 1: If |y|2 ≤ |x|3 , then using |y| ≤ |x|3/2 , we get |x|3 < λ (|x|)(5/2)(1+) |k|1+ ,
hence |x|(1−5)/2 < λ |k|1+ .
Therefore for 0 < < 1/5, we have |x| < λ 2/(1−5) |k|2(1+)/(1−5) and |y| ≤ |x|3/2 <
λ 3/(1−5) |k|3(1+)/(1−5) .
Case 2: If |x|3 ≤ |y|2 , then using |x| ≤ |y|2/3 , we get |y|2 < λ (|y|)(5/3)(1+) |k|1+ ,
hence |y|(1−5)/3 < λ |k|1+ .
Therefore for 0 < < 1/5, we have |y| < λ 3/(1−5) |k|3(1+)/(1−5) and |x| ≤ |y|2/3 <
λ 2/(1−5) |k|2(1+)/(1−5) .
In conclusion we get the same bounds for both cases:
|x| < λ 2/(1−5) |k|2(1+)/(1−5) , |y| < λ 3/(1−5) |k|3(1+)/(1−5) .
1+
Let = 1 + 0 . Then when 0 < < 1/5, we have 0 < 0 < ∞, and 0 is small if
1 − 5
and only if is small. Finally setting C0 = max(λ 2/(1−5) , λ 3/(1−5) ), we are done.
65
4.2. Abc Implies Fermat for Large Exponents
One can derive Fermat’s last theorem for large exponents using abc conjecture
easily.
Theorem 4.10 (Fermat’s Last Theorem, [19]). The equation xn + y n = z n , where

x, y, z, and n are integers, has no non-zero solutions for n > 2.
Theorem 4.11. Assume abc conjecture. Let x, y, z, n be positive integers such that
gcd(x, y, z) = 1 and xn + y n = z n . Then there is no solution for n > 6.
Proof. Suppose xn + y n = z n with x, y, z relatively prime positive integers and n ≥ 3.

Let λ1 = 1, then by abc conjecture we have, z n ≤ Rad(xn y n z n )2 where Rad(xn y n z n )2 =
Rad(xyz)2 . Then z n ≤ Rad(xyz)2 ≤ (xyz)2 ≤ z 6 , which shows that n ≤ 6. Hence we
have Fermat’s Last Theorem for n ≥ 7.
66
5. MORDELL CONJECTURE
Let f (x1 , x2 , . . . , xn ) be a polynomial with coefficients in Q. Finding rational

solutions to f = 0 is equivalent to finding rational points on V, where V is the affine
space given by f. The variety V will be (n − 1)-dimensional, if f is not identically
zero. When finitely many polynomials are taken into account, V is defined to be the
common zero locus.
A smooth curve C as defined above may be thought of as a Riemann surface

with finitely many punctures; after taking an appropriate compactification by filling
these punctures, the resulting compact Riemann surface has a topological genus, which
is essentially the number of holes. For simplicity, we will assume that our curve C is
projective nonsingular curve and it is connected.
When the curve C has genus 0 or 1, it is possible for the curve to have infinitely
many rational points. For the genus 0 case, it is easily determined by Hasse Principle:
Hasse Principle: [20] A genus 0 curve C has a rational point if and only if it
has a solution over the real numbers R and the p-adic numbers Qp for all primes p.
When the genus is 1, it is a big mystery. Genus 1 curves with a given rational
point are known as elliptic curves. For more details, see Section 2.4.
On the other hand, curves with genus ≥ 2 cannot have infinitely many rational
points.
Conjecture 5.1 (Mordell Conjecture, 1922). A curve of genus greater than 1 over the
field Q of rational numbers has only finitely many rational points.
Faltings proved Mordell conjecture and generalized it over number fields.

67
Theorem 5.2 (Faltings, 1983). If K is any number field and C is any curve of genus
> 1 defined over K, then C has only a finite number of K-rational points.
Example 5.3. The curve defined by xn + y n + 1 has finitely many rational points when
n ≥ 4 by Example 2.77.
68
6. ABC IMPLIES MORDELL
6.1. Adaptation and Examples
The aim is to prove Mordell conjecture using abc conjecture. Before giving the
proof, we construct the necessary set-up.
Theorem 6.1 (Mordell conjecture 1922, Faltings theorem). Any curve of genus at
least 2 over a number field K has only finitely many K-rational points.
First we reconcile our terminology with the paper of Elkies [1], hence restate abc
conjecture.
Conjecture 6.2 (abc conjecture). For any relatively prime nonzero A, B, C ∈ Z such
that A + B + C = 0, and for all > 0,
N (A, B, C) H(A, B, C)1− , (6.1)
with the constant implied in depending on but not on A, B, C, where
Y
N (A, B, C) = p , p prime (6.2)
p|ABC
and
H(A, B, C) = max(|A|, |B|, |C|). (6.3)
Note that here H stands for the height, and N for the conductor where both
of them are usually known with definitions on curves. This is the first hint of the
connection we make between the theory of Diophantine approximation and the theory
of points on curves of high genera.
69
Example 6.3. Let A = 128, B = −125, and C = −3. Then N (A, B, C) = 30, and
H(A, B, C) = 128. By abc conjecture, there is a constant depending on , say λ , so that
128 ≤ λ .301+ . As seen clearly in this example, the conjecture is not true in the form
of H(A, B, C) ≤ N (A, B, C), in other words when simply = 1, and λ = 1. There are
infinitely many exceptions in this case. However, once the exponent is greater than 1,
there exist only finitely many exceptions.
The statement of abc conjecture is for integers, whereas the statement of Mordell
conjecture is for number fields. So we give a more general construction of abc conjecture.
First, we state the conjecture over Q and hence remove the condition of being
relatively prime with the new definitions of the height and the conductor:
Conjecture 6.4 (abc conjecture over Q). For any nonzero A, B, C ∈ Q such that
A + B + C = 0, and for all > 0,
N (A, B, C) H(A, B, C)1− , (6.4)
with the constant implied in depending on but not on A, B, C, where
Y
H(A, B, C) = max(|A|v , |B|v , |C|v ), (6.5)
v∈MQ
and
Y
N (A, B, C) = p, (6.6)
p∈I
where I = {p prime : max(|A|p , |B|p , |C|p ) > min(|A|p , |B|p , |C|p )}.
Notation: The set of normalized valuations on Q is the set MQ consisting of the

archimedean absolute value | |∞ and the p−adic absolute values | |p for every prime p.
70
Note that Formulas 6.5 and 6.6 are finite products: Let A = a/b ∈ Q − {0} with
(a, b) = 1 and a, b ∈ Z. Then |A|p = 1 unless p | a or p | b. So there are finitely many
primes that contribute to both of the products.
Moreover, the values of N (A, B, C) and H(A, B, C) do not change when A, B,

and C are multiplied by a scalar in Q∗ and they agree with Formulas 6.2 and 6.3 when
A, B, and C are relatively prime integers. We prove these in the Proposition 6.5 below.
Proposition 6.5. (i) The height N and the conductor N are scaling invariant.
(ii) When A, B, C are relatively prime integers, Formulas 6.5 and 6.6 agree with
Formulas 6.2 and 6.3, respectively.
Proof. (i) Let λ ∈ Q∗ . Using the Product Formula 2.10, we find that
Y
H(λA, λB, λC) = max(|λA|v , |λB|v , |λC|v )
v∈MQ
  
Y Y Y
= λ  max(|A|v , |B|v , |C|v ) = max(|A|v , |B|v , |C|v ) = H(A, B, C).
v∈MQ v∈MQ v∈MQ
Y
For the conductor, we have N (λA, λB, λC) = p, where
p∈I
I = {p prime : max(|λA|p , |λB|p , |λC|p ) > min(|λA|p , |λB|p , |λC|p )}.
By the multiplicativity of valuation, we can also write the set I as
I = {p prime : (|λ|p ).max(|A|p , |B|p , |C|p ) > (|λ|p ).min(|A|p , |B|p , |C|p )}.
Hence, the primes contributing to the set I stay invariant and N (λA, λB, λC) =
N (A, B, C).
71
(ii) Let A, B, C be relatively prime nonzero integers. First, looking at the height
formula
Y
H(A, B, C) = max(|A|v , |B|v , |C|v ),
v∈MQ
the p-adic absolute values in MQ affects the formula if the primes appear in the factor-
ization of A, B, C. However, A, B, C are relatively prime, so none of them shares the
same prime factor. Moreover, they are all integers, so their p-adic values are either
less than 1 or equal to 1. In the end, we get 1 as the maximum value from each finite
prime. Therefore, the height formula becomes H(A, B, C) = max(|A|, |B|, |C|).
For the conductor formula, the only primes in the set I are the primes in the
factorization of A, B, C, in other words p ∈ I if and only if p|ABC. So the conductor
Y
formula becomes N (A, B, C) = p.
p|ABC
We give an example to show the consistency of the old formulas with the new
formulas when relatively prime integers are taken.
Example 6.6. Take A = −9, B = 5, C = 4. Applying the old Formulas 6.2 and 6.3,
we get H = 9, and N = 2.3.5 = 30. Applying Formulas 6.6 and 6.5, we get:
• For the prime p = 2, we have |A|2 = 1, |B|2 = 1, |C|2 = 1/4.

• For p = 3 : |A|3 = 1/9, |B|3 = 1, and |C|3 = 1.
• For p = 5 : |A|5 = 1, |B|5 = 1/5, |C|5 = 1.
• For all primes other than 2, 3 or 5, we have |A|p = |B|p = |C|p = 1.
• For the usual real absolute value, we have |A|∞ = 9, |B|∞ = 5, and |C|∞ = 4.
To evaluate the height, we multiply the maximum values from each valuation: H = 9;
and to evaluate the conductor, we multiply the primes at which the maximum value is
strictly larger than the minimum value: N = 2.3.5 = 30.
72
Example 6.7. We also give an example to show that the products are scaling invariant:
(i) If λ has a new prime factor:

Let λ = 7 and take the set-up in Example 6.6, then A = −63, B = 35, C = 28.
It is enough to check for the new prime and the usual absolute value since there
is no change in the computations related to old primes.
We get |A|7 = |B|7 = |C|7 = 1/7 and |A|∞ = 63, |B|∞ = 35, and |C|∞ = 28,
1
then we have H = 63. = 9 and N = 2.3.5 = 30. For the new prime, whether
7
it is in the nominator or in the denominator, the value | |λ is the same for all
of them because we multiply A, B, and C with the same multiplicity of λ. Also
| |∞ increases or decreases by the same multiplicity, hence ends up canceling any
change in the height. Since they are all multiplied with the same factor, the
conductor stays the same.
(ii) If λ is a prime which already exists in the prime factorization of one of A, B,
or C : Let λ = 3, then for A = −27, B = 15, and C = 12, we get |A|3 = 1/27,
|B|3 = 1/3, |C|3 = 1/3 and |A|∞ = 27, |B|∞ = 15, and |C|∞ = 12. So H =
1
27. = 9 and N = 2.3.5 = 30. The effect of the p−adic valuation of the new
3
prime λ and the effect of the usual absolute value cancel each other out again.
3 −1 −13
Example 6.8. Take A = , B = , and C = .
5 6 30
• For p = 2 : |A|2 = 1, |B|2 = 2, |C|2 = 2.

• For p = 3 : |A|3 = 1/3, |B|3 = 3, and |C|3 = 3.
• For p = 5 : |A|5 = 5, |B|5 = 1, |C|5 = 5.
• For p = 13 : |A|13 = 1, |B|13 = 1, |C|13 = 1/13.
• |A|∞ = 3/5, |B|∞ = 1/6, and |C|∞ = 13/30.
3
Then H = 2.3.5. = 18, N = 2.3.5.13 = 390 and we have N = 390 > 18 as usual.
5
73
Conjecture 6.9 (abc conjecture over K). Let K be a number field. For all > 0, and
for all nonzero A, B, C ∈ K with A + B + C = 0,
N (A, B, C) H(A, B, C)1− (6.7)
with the constant implied in depending on K and , but not on A, B, C, where
Y
H(A, B, C) = max(|A|v , |B|v , |C|v ), (6.8)
v∈MK
and N (A, B, C) is the product of the absolute norms of all the finite primes P of K at
which max(|A|P , |B|P , |C|P ) strictly exceeds min(|A|P , |B|P , |C|P ).
Notation: The set of normalized valuations on a number field K is the set MK

consisting of all absolute values on K whose restriction to Q is one of the normalized
valuations on Q. We write MK ∞ for the set of archimedean absolute values in MK , and
MK 0 for the set of nonarchimedean absolute values in MK .
Note that archimedean absolute value is also denoted as infinite absolute value,
and nonarchimedean as finite.
H and N are finite products: there are finitely many nonarchimedean valuations,
since there are finitely many prime ideals in the factorization of an ideal in a Dedekind
domain; and finitely many archimedean valuations, because K is a finite extension of
Q, so the number of embeddings of K is finite.
√ √ √ 1
Example 6.10. Set K = Q( 2) with OK = Z[ 2]. Let A = 2, B = , and
2
√ 1
C =− 2− .
2
√ √
Archimedean: K has two real embeddings, say φ1 : a + b 2 7→ a + b 2 and
√ √
φ2 : a + b 2 7→ a − b 2, hence two archimedean absolute values. First, we look at the
image under the embedding, then take the usual real absolute value of the outcome.
74
√ 1 √ 1
|A|φ1 = 2, |B|φ1 = , |C|φ1 = 2 +
2 2
√ 1 √ 1
|A|φ2 = 2, |B|φ2 = , |C|φ2 = 2 −
2 2
Nonarchimedean: We decompose principal fractional ideals into prime ideals in

OK , then consider valuations up to these resulting primes.
Note that the discriminant of K is 8, so 2 is the only prime that ramifies in K.
√ √ √
• A= 2 creates the principal fractional ideal h 2i = 2OK . We find its prime
√ √
factorization via norm: N( 2OK ) = |NK/Q ( 2)| = 2. We look at how 2 decom-
√ √
poses. Using the method in Example 2.34, we find that 2Z[ 2] = h2, 2i2 , where
√ √ √ √ √
N(2Z[ 2]) = 4 = N(h2, 2i)2 , so N(h2, 2i) = 2, and 2OK = h2, 2i. Let
√ 1
P = h2, 2i. |A|P = and |A|Q = 1 for all other primes Q.
2

1 1 1 1 1
• B = creates = OK . We compute its norm: N OK = NK/Q =

2 2 2 2 2
1 1 √ 1 −2
. Then OK = h2, 2i−2 , so |B|P = = 4 and |B|Q = 1 for all other
4 2 2
primes Q.
√ √ √

1 1 1
• C = − 2 − creates − 2 − = − 2− OK . We compute its norm:
2 2 2
√ 1

√

1 7
N − 2− OK = NK/Q − 2 − = . Using the method in Example
2 2 4
2.34 with p = 7, we see that 7O
K is prime. Let 7OK = P1 , then N(P1 ) = 49.
√

7 1 1
Then the ideal of norm is − 2 − = P1 .P −2 . Hence, |C|P = 4, |C|P1 = ,
4 2 7
and |C|Q = 1 otherwise.
√
In the end, multiplying these maximum values, we obtain H = 8 + 2 2; and
choosing the valuations where maximum value strictly exceeds minimum value out of
finite primes, which is all of them in this case, we get N = 2.7 = 14. We have N =
√
14 > 8 + 2 2 + H as usual.
75
Example 6.11. Set K = Q(i) with OK = Z[i]. Let A = 3+2i, B = −5, and C = 2−2i.
Archimedean: K has two embeddings which are complex-conjuagates, σ1 : a+bi 7→

a + bi and σ2 : a + bi 7→ a − bi, hence one archimedean absolute value. Let σ = (σ1 , σ2 ).
|A|σ = |σ1 (A).σ2 (A)|∞ = |(3 + 2i)(3 − 2i)|∞ = 13, |B|σ = 25, |C|σ = 8.
Nonarchimedean: Note that the discriminant of K is −4, so 2 is the only prime

that ramifies in K.
• A = 3 + 2i generates the principal fractional ideal (3 + 2i)OK with norm N((3 +

2i)OK ) = |NK/Q (3 + 2i)| = 13. Using the method in Example 2.34 with p = 13, we
get the factorization 13OK = h13, i + 8ih13, i − 8i. Let P1 = h13, i + 8i and P2 =
h13, i − 8i, where N(P1 ) = N(P2 ) = 13. Then we have either (3 + 2i)OK = P1
1
or (3 + 2i)OK = P2 . In both cases, the same result follows: |A|P1 = |A|P2 = .
13
1
So we can take P1 into account without loss of generality. Then |A|P1 = and
13
|A|L = 1 for all other primes L.
• B = −5 generates (−5)OK with norm N(−5OK ) = |NK/Q (−5)| = 25 = 52 . Us-

ing the method in Example 2.34 with p = 5, we get the decomposition 5OK =
(−5)OK = h5, i + 2ih5, i − 2i. Let Q1 = h5, i + 2i and Q2 = h5, i − 2i, where
1
N(Q1 ) = N(Q2 ) = 5. We get |B|Q1 = |B|Q2 = . So take Q1 into account without
5
loss of generality, and |B|L = 1 for all other primes L.
• C = 2 − 2i generates (2 − 2i)OK with norm N((2 − 2i)OK ) = |NK/Q (2 − 2i)| =

8 = 23 . The prime 2 is ramified in K, and its factorization is 2OK = h2, 1 + ii2
with N(h2, 1 + ii) = 2. Then (2 − 2i)OK = h2, 1 + ii3 . Let P denote the prime
1 3 1
ideal h2, 1 + ii, so |C|P = = , and |C|L = 1 for all other primes L.
2 8
In the end, the only contribution to H comes from the archimedean absolute value
for this case: H = 25, and N = 2.5.13 = 130. So we have N > H as usual.
76
√ √
√ 2 + 3i −5i −3 2 + i
Example 6.12. Let K = Q( 2, i). Let A = ,B= , and C = .
2 3 6
Archimedean: K has four complex embeddings, which are
√ √ √ √




 φ1 : a + b 2 + ci + d 2i →
7 a + b 2 + ci + d 2i,
√ √ √ √



φ2 : a + b 2 + ci + d 2i 7→ a + b 2 − ci − d 2i,

√ √ √ √ (6.9)
φ3 : a + b 2 + ci + d 2i 7→ a − b 2 + ci − d 2i,





φ : a + b√2 + ci + d√2i 7→ a − b√2 − ci + d√2i.



4
As seen clearly there are two pairs of complex-conjugate embeddings, say (φ1 , φ2 ) := σ1
and (φ3 , φ4 ) := σ2 , hence two archimedean absolute values.
√2 + 3i √2 − 3i 11
|A|σ1 = |φ1 (A).φ2 (A)|∞ = . = ,

2 2 ∞ 4
−5i 5i 25
|B|σ1 = |φ1 (B).φ2 (B)|∞ = = ,

3 3 ∞ 9
−3√2 + i −3√2 − i 19
|C|σ1 = |φ1 (C).φ2 (C)|∞ = = .

6 6 ∞ 36
−√2 + 3i −√2 − 3i 11
|A|σ2 = |φ3 (A).φ4 (A)|∞ = . = ,

2 2 ∞ 4
−5i 5i 25
|B|σ2 = |φ3 (B).φ4 (B)|∞ = = ,

3 3 ∞ 9
3√2 + i 3√2 − i 19
|C|σ2 = |φ3 (C).φ4 (C)|∞ = = .

6 6 ∞ 36
Nonarchimedean: Again we decompose principal fractional ideals into prime ide-

als in OK using norm, then we consider valuations up to primes that appear in the
factorization.
The discriminant of K is equal to 147456 = 214 .32 . Therefore the only primes
ramifying in K are 2 and 3.
77
√
Note that we can write K as Q( 2 + i) by Primitive Element Theorem, then the
√
minimal polynomial m(x) of 2 + i over Q is m(x) = x4 − 2x2 + 9.
√ *√ +
√2 + 3i
2 + 3i 2 + 3i
• A= creates the principal fractional ideal = OK .
2 2 2
√2 + 3i 112
We compute the norm as NK/Q = 4 .

2 2
In order to find the factorization into prime ideals, one can look at the minimal
polynomial to use Proposition 2.38, since OK is monogenic by Example 2.37. The
minimal polynomial m(x) = x4 −2x2 +9 has no solution mod 11, hence 11OK = P
with residual degree 4 and N(P) = 112 . On the √ other hand, m(x) ≡ (x − 1)4
2 + 3i
mod 2, hence 2OK = Q4 with N(Q) = 2. Then OK = P.Q−4 . So, we
2
1
get |A|P = , |A|Q = 16 and 1 for all the other primes.
121
−5i

−5i

5i 5i 54
• B= creates = OK with the norm N OK = 4 .
3 3 3 3 3
4 2
The minimal polynomial m(x) = x − 2x + 9 factorizes mod 5 as m(x) ≡
(x2 − x + 2)(x2 + x + 2), hence 5OK = IJ with N(I) = 52 and N(J ) = 52 , where
I = hx2 − x + 2i and J = hx2 + x + 2i . However, m(x) stays irreducible mod 3,
5i
4
so 3 stays prime above: 3OK = H with N(H) = 3 . Then OK = I.J .H−1 ,
1 1 1 −1 3
and |B|I = , |B|J = , |B|H = = 81, and for all the other
25 25 81
primes we get 1.
√ −3√2 + i −3√2 + i
−3 2 + i
• C = creates OK with the norm N OK =
6 6 6
2
19
.
24 .34
The minimal polynomial m(x) = x4 −2x2 +9 has no√solution mod 19, so 19OK = L
−3 2 + i
2
with residual degree 4 and N(L) = 19 . Then OK = L.Q−4 .H−1 . We
6
1
obtain |C|L = , |C|Q = 16, |C|H = 81, and 1 for all the other primes.
271
In the end, we have N > H with H = 3600 and N = 121.2.25.25.81.271 = 3320088750.

78
Example 6.13. An example

√ of√a Kummer but √ not a Galois extension: we take K =
√ 1− 23 3
2 3
2−3
Q( 3 2) with A = ,B= , and C = .
2 3 6
√
Note that the minimal polynomial of 3
2 over Q is x3 − 2, and the conjugates of
√ √ √
3
2 are 3 2w and 3 2w2 with w = e2πi/3 .
√ √
An element in K can be expressed as a + b 3 2 + c 3 4 where a, b, c ∈ Q.
Archimedean: K has three embeddings, which are

√ √ √ √
φ1 : a + b 3 2 + c 3 4 7→ a + b 3 2 + c 3 4,





√ √ √ √

φ2 : a + b 3
2 + c 3
4 →
7 a + b 3
2w + c 3
4w2 , (6.10)

√ √ √ √



φ3 : a + b 3 2 + c 3 4 7→ a + b 3 2w2 + c 3 4w.

We get two archimedean absolute values, φ1 and (φ2 , φ3 ) := σ1 . Then
√
3
2−1
|A|φ1 = |φ1 (A)|∞ = ,
2
√
3
2
|B|φ1 = |φ1 (B)|∞ = ,
3
√
3− 32
|C|φ1 = |φ1 (C)|∞ = .
6
1 − √ 3
√
2w 1 − 3 2w2
√ √
1+ 32+ 34
|A|σ1 = |φ2 (A).φ2 (A)|∞ = . = ,

2 2 ∞ 4
√
3
√
2w 3 2w2
√3
4
|B|σ1 = |φ2 (B).φ2 (B)|∞ = . = ,

3 3 ∞ 9
√3
√
2w − 3 3 2w2 − 3
√
9+332+ 34
√
|C|σ1 = |φ2 (C).φ2 (C)|∞ = . = .

6 6 ∞ 36
Nonarchimedean: The discriminant of K is −108, so only primes that ramify in

K are 2 and 3.
79
√ √
Using the basis {1, 3 2, 3 4}, the matrix representation for multiplication by α =
√ √ √ √
a + b 3 2 + c 3 4 where a, b, c ∈ Q is obtained by multiplying α by 1, 3 2, and 3 4 :
√ √ √ √ √ √ √ √
α.1 = a + b 3 2 + c 3 4, α. 3 2 = 2c + a 3 2 + b 3 4, and α. 3 4 = 2b + 2c 3 2 + a 3 4, then
using these calculations we get [mα ] =
 
a 2c 2b
 
 
 b a 2c
 
c b a
√ √
Therefore the norm of an element is NK/Q (a + b 3 2 + c 3 4) = a3 + 2b3 + 4c3 − 6abc.
√ 1 − √
1− 32 3
2 1
• A = creates the principal deal OK with the norm . Since
2
√ 2 8
OK = Z[ 3 2], we can factorize into prime ideals by looking at the factorization of
the minimal polynomial√ mod p. As x3 − 2 ≡ x3 mod 2, 2 is totally ramified in
1 − 3 2
K, and we have OK = P −3 with N(P) = 2. So, |A|P = 8 and for all
2
the other primes we get 1.
√
3 √ 3
2 2 2
• B= creates OK with the norm 3 . Since x3 −2 = (x−2)(x2 +2x+1) =
3 3 3
√3
2
2 3
(x + 1)(x + 1) = (x + 1) ∈ F3 [x], we have OK = PQ−3 with N(Q) = 3.
3
1
So, |B|P = , |B|Q = 27, and for all the other primes we get 1.
2
√ √
3
2−3 3
2 − 3 52
• C= creates OK with the norm 3 3 . Since 5 does not ramify,
6 √ 6 2 .3
3 2 − 3
we have OK = P −3 Q−3 R with N(R) = 25. So, |C|P = 8, |C|Q = 27,
6
1
|C|R = , and for all the other primes we get 1.
25
In the end, we have N > H with H = 27 and N = 30.

80
Example 6.14. An example of a Kummer, Galois extension: we take K = Q(α) with

√ 1 − 2α 2α −1
α := 7 6. Let A = ,B= , and C = .
3 3 3
A basis for K is {1, α, α2 , α3 , α4 , α5 , α6 }. Hence an element in K can be shown as

a + bα + cα2 + dα3 + eα4 + f α5 + gα6 with all coefficients in Q. Moreover the minimal
polynomial of α is x7 − 6, so its conjugates are α, αw, αw2 , αw3 , αw4 , αw5 , and αw6
2π 2π
where w = e2πi/7 = cos + i.sin .
7 7
Archimedean: K has seven embeddings:
φ1 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bα + cα2 + dα3 + eα4 + f α5 + gα6 ,
φ2 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw + cα2 w2 + dα3 w3 + eα4 w4
+ f α5 w5 + gα6 w6 ,
φ3 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw2 + cα2 w4 + dα3 w6 + eα4 w
+ f α5 w3 + gα6 w5 ,
φ4 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw3 + cα2 w6 + dα3 w2 + eα4 w5
+ f α5 w + gα6 w4 ,
φ5 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw4 + cα2 w + dα3 w5 + eα4 w2
+ f α5 w6 + gα6 w3 ,
φ6 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw5 + cα2 w3 + dα3 w + eα4 w6
+ f α5 w4 + gα6 w2 ,
φ7 : a + bα + cα2 + dα3 + eα4 + f α5 + gα6 7→ a + bαw6 + cα2 w5 + dα3 w4 + eα4 w3
+ f α5 w2 + gα6 w,
(6.11)
We get four archimedean absolute values coming from one real embedding φ1
and three complex-conjugate pairs, which we denote as (φ2 , φ7 ) := σ1 , (φ3 , φ6 ) := σ2 ,
(φ4 , φ5 ) := σ3 . Then
81
2α − 1 2α 1
|A|φ1 = |φ1 (A)|∞ = , |B|φ1 = |φ1 (B)|∞ = , |C|φ1 = |φ1 (C)|∞ = .
3 3 3
The maximum value for φ1 is B.
1 − 2αw 1 − 2αw6 1 − 4α cos(2π/7) + 4α2

|A|σ1 = |φ2 (A).φ7 (A)|∞ = . = ,

3 3 ∞ 9
2αw 2αw6 4α2
|B|σ1 = |φ2 (B).φ7 (B)|∞ = . = ,

3 3 ∞ 9
1
|C|σ1 = |φ2 (C).φ7 (C)|∞ = .
9
Since cos(2π/7) is approximately equal to 0.64, the maximum value for σ1 is B.
1 − 2αw2 1 − 2αw5 1 − 4α cos(4π/7) + 4α2

|A|σ2 = . = ,

3 3 ∞ 9
4α2 1
|B|σ2 = , |C|σ2 = .
9 9
Since cos(4π/7) gives us a negative value, the maximum value for σ2 is A.
1 − 2αw3 1 − 2αw4 1 − 4α cos(6π/7) + 4α2

|A|σ3 = . = ,

3 3 ∞ 9
4α2 1
|B|σ3 = , |C|σ3 = .
9 9
Since cos(6π/7) gives us a negative value, the maximum value for σ2 is A.
Nonarchimedean: Using the tools from Proposition 2.39, we find that OK =

√ √
Z[α] = Z[ 7 6]. Moreover, the only primes of Z that ramify in Z[ 7 6] are 2, 3 and 7.
√ √
Explicitly, 2Z[ 7 6] = hα, 2i7 , and 3Z[ 7 6] = hα, 3i7 . Let P := hα, 2i and Q := hα, 3i.
1 − 2α 1 − 2α
• A = creates the principal fractional ideal OK with the norm
3 3
−13.59 1 − 2α
7
. Then OK = IJ Q−7 with N(I) = 13, N(J ) = 59, and N(Q) =
3 3
1 1
3. So, |A|I = , |A|J = , |A|Q = 37 = 2187, and for other primes we get 1.
13 59
82
2α 2α 28 2α
• B = creates OK with the norm 6 . Then OK = P 8 Q−6 with
3 3 3 3
1
N(P) = 2. So, |B|P = , |B|Q = 729, and for all the other primes we get 1.
256
−1 −1 −1 −1
• C = creates OK with norm 7 . Then OK = Q−7 . So, |C|Q =
3 3 3 3
2187, and for all the other primes we get 1.
In the end, we have N > H with H = 8α3 (1−4α cos(4π/7)+4α2 )(1−4α cos(6π/7)+
4α2 ) and N = 13.59.2.3 = 4602.
Remark: In fact, both of the formulas, H and N, depend on a ratio:
Divide the equation A + B + C = 0 by −B, then the values of H and N do not

change, since H and N are scaling invariant. We get:
−A/B − 1 + (A + B)/B = 0
where r := (−A/B), with r ∈ P1 K − {0, 1, ∞}, since A, C, and B cannot be zero,

respectively. Hence,
H(A, B, C) = H(−A/B, −1, A/B + 1) = H(r, −1, 1 − r),
N (A, B, C) = N (r, −1, 1 − r).

83
We simplify our notation:
Proposition 6.15. Let r1 be the number of real embeddings of K and r2 be the number
of the complex embeddings of K. Then
Y
H(A, B, C) = max(1, |r|v ) × κ, (6.12)
v∈MK
where 1 ≤ κ ≤ 2r1 +r2 .
Proof. Since | − 1|v = |1|v = 1, we have
Y
H(A, B, C) = H(r, −1, 1 − r) = max(1, |r|v , |1 − r|v ).
v∈MK
Now we divide the product into two parts: over all nonarchimedean absolute values
and over all archimedean absolute values.
Y Y Y
max(1, |r|v , |1 − r|v ) = max(1, |r|v , |1 − r|v ) × max(1, |r|v , |1 − r|v ).
v∈MK v∈MK 0 v∈MK ∞
By definition, |1 − r|v ≤ max(1, |r|v ) for all nonarchimedean valuations, then we have
Y Y
max(1, |r|v , |1 − r|v ) = max(1, |r|v ).
v∈MK 0 v∈MK 0
For the archimedean part of the product:
Case 1: If |1 − r|v , |r|v ≤ 1, then we take 1, and in this case κ = 1.
Case 2: If |1 − r|v ≤ |r|v , then we take |r|v , and in this case κ = 1.
Y |1 − r|v
Case 3: If 1 < |r|v < |1 − r|v , then we multiply max(1, |r|v , |1 − r|v ) by
v∈MK ∞
|r|v
m
Y |1 − r|v
to make it equal with max(1, |r|v ), hence κ = where m is the num-
v∈MK ∞
|r|v
ber of the archimedean absolute values satisfying the inequality.
84
|1 − r|v 1 + |r|v 1
Moreover, ≤ = 1+ < 2. Since there are r1 + r2 many archimedean
|r|v |r|v |r|v
absolute values in total, we have 20 = 1 ≤ κ ≤ 2r1 +r2 .
Case 4: If |r|v ≤ 1 < |1 − r|v , then κ = (|1 − r|v )m where m is the number of the
archimedean absolute values satisfying the inequality. Since |1 − r|v ≤ 1 + |r|v ≤ 2, we
are done.
Further setting M ≥ (r1 + r2 ).ln2 with |O(1)| ≤ M, we shortly write
H(A, B, C) = H(r) × exp(O(1)), (6.13)
Y
where H(r) denotes the naive height max(1, |r|v ) by abuse notation.
v∈MK
Proposition 6.16. N (A, B, C) is the product of the absolute norms of all the finite
primes of K at which r, 1/r or r − 1 has a positive valuation.
Proof. By definition N (r, −1, 1−r) is the product of the absolute norms of all the finite
primes of K at which max(1, |r|P , |1 − r|P ) strictly exceeds min(1, |r|P , |1 − r|P )}.
First, we check if the finite primes in the product N are the primes such that
r, 1/r or r − 1 has a positive valuation at, i.e., ordP > 0.
Since finite primes are nonarchimedean absolute values, we have |1 − r|P ≤

max(1, |r|P ) for all primes P, which gives us the cases below to consider:
(i) When |1 − r|P < max(1, |r|P ) : those primes contribute to the product because
of the strict inequality.
• If max(1, |r|P ) = 1, then |1 − r|P < 1 and |r|P < 1. So both r and 1 − r have
positive valuation at P.
• If max(1, |r|P ) = |r|P and |1 − r|P < 1 < |r|P , then |r|P > 1. So both 1/r
and 1 − r have positive valuation at P.
85
• If max(1, |r|P ) = |r|P and 1 < |1 − r|P < |r|P , then 1/r has a positive
valuation at P.
(ii) When |1 − r|P = max(1, |r|P ) :
• If |1 − r|P = 1 > |r|P , then P ∈ I, and r has a positive valuation at P.
• If |1 − r|P = 1 = |r|P , then P ∈
/ I.
• If |1 − r|P = |r|P < 1, then P ∈ I, and 1/r has a positive valuation at P.
For the other way around:
• If r has a positive valuation at P, then |r|P < 1, hence P ∈ I.
• If r − 1 has a positive valuation at P, then |r − 1|P = |1 − r|P < 1, hence
P ∈ I.
• If 1/r has a positive valuation at P, then |1/r|P < 1 or equivalently |r|P > 1,
hence P ∈ I.
We shortly denote as N (r) := N (r, −1, 1−r), and factor N (r) = N0 (r).N1 (r).N∞ (r)
as the product of absolute norms of the prime ideals containing r, r−1, 1/r respectively.
86
6.2. Motivation for the Proof
The motivation for the proof of Elkies is the Fermat bound obtained by abc
conjecture. As in the set-up of abc conjecture, we start with integers.
Proposition 6.17. Assume abc conjecture holds, then there are only finitely many
integer solutions to xn + y n + z n = 0 with gcd(x, y, z) = 1 and n ≥ 3.
Proof. Let n ≥ 3 and x, y, z be relatively prime, nonzero integers such that xn + y n +

z n = 0. Taking (A, B, C) = (xn , y n , z n ), we have
N (A, B, C) = N (xn , y n , z n ) = N (x, y, z) ≤ |xyz| < max{|x|, |y|, |z|}3 = H(xn , y n , z n )3/n ,
(6.14)
or N (A, B, C) < H(A, B, C)3/n , hence we get a contradiction with abc conjecture for all
< 1 − (3/n) once H(A, B, C) is large enough. Therefore there are only finitely many
(A, B, C) satisfying the equation. As another way to see this, we plug the inequality we
obtanied N (A, B, C) < H(A, B, C)3/n into abc conjecture, then we get H(A, B, C)1− <
C H(A, B, C)3/n or H(A, B, C) < C 1−−3/n , for some constant C , which gives us an
effective height bound, limiting the possibilities to a finite number.
Next, we carry this set-up to rationals by considering the ratios of variables:
Proposition 6.18. Assume abc conjecture holds, then there are only finitely many
rational solutions to xn + y n + z n = 0 with n ≥ 3.
Proof. Let (x, y, z) ∈ P2 (Q), then we can take x, y, z to be relatively prime integers. We
divide the equation xn +y n +z n = 0 by −y n , then we get −(x/y)n −1+(xn + y n /y n ) = 0.
Let r := −(x/y)n ∈ Q, then 1 − r = 1 + (x/y)n = −(z/y)n and 1/r = −(y/x)n . Note
that all fractions are in lowest terms because x, y, z are relatively prime. Then we are
looking for points (r, −1, 1 − r) with r ∈ Q.
87
Since N is scaling invariant by Proposition 6.5, N (A, B, C) = N (r, −1, 1 − r) =

N0 (r).N1 (r).N∞ (r) using Proposition 6.16. Explicitly,
N0 (r) : product of primes p such that r ≡ 0 mod p or − (x/y)n ≡ 0 mod p

(6.15)
N1 (r) : product of primes p such that 1 − r ≡ 0 mod p or r ≡ 1 mod p (6.16)
N∞ (r) : product of primes p such that 1/r ≡ 0 mod p or “r ≡ ∞ mod p”.

(6.17)
Hence N (r, −1, 1 − r) becomes the product of primes p such that r mod p is one of
0, 1, ∞, in other words, primes dividing x, y, z, respectively for N0 (r), N1 (r), N∞ (r).
Without loss of generality, assume x is positive and max{|x|∞ , |y|∞ , |z|∞ } = x.

For the height, we have
Y
H(r, −1, 1 − r) = H(−xn /y n , −1, −y n /xn ) = max||r||v , 1, ||1 − r||v .
v∈MQ
We get | − xn /y n |∞ for archimedean valuation from our assumption. For the fi-
nite primes, we consider the primes dividing denominators; otherwise we choose 1
as maximum. Hence multiplying for all nonarchimedean valuations, we get xn y n . Then
H(r, −1, 1 − r) = x2n .
We obtain the inequality N (r, −1, 1 − r) = N0 (r).N1 (r).N∞ (r) ≤ |xyz| < x3 =
H(r, −1, 1 − r)3/2n , where 3/2n < 1 since n ≥ 3. Then N < H 3/2n , so (r, −1, 1 − r)
gives a counter example to abc conjecture over Q for < 1 − (3/2n) once H is large
enough, i.e., for all but finitely many (r, −1, 1 − r).
88
In the final step of set-up, we take the problem into consideration in a geometric
manner. We take (x, y, z) as a rational point on the Fermat curve and we build the
rational function f = (−x/y)n on the Fermat curve, then one can interpret as that N
reflects the ramification of the map above 0, 1, ∞ and H represents the degree of the
map. Hence, we first compute the degree of the map f and the number of preimages
of f corresponding the points in the set {0, 1, ∞}, which will be useful in the proof.
Proposition 6.19. Let Fn be the zero locus of xn + y n + z n = 0 and let f : Fn (Q) →

P1 (C) with (x, y, z) 7→ (−xn , y n ). Then the degree of f is n2 .
Proof. On the affine part of the curve Fn given by y = 1 and xn + z n + 1 = 0, we have

f : (x, 1, z) 7→ (−xn , 1), which induces the map
f ∗ : C[x] → CFn := C[x, z]/hz n + xn + 1i with g 7→ g ◦ f + hz n + xn + 1i.
Then deg f = [CFn : f ∗ (C[x])] = [CFn : C[x]].[C[x] : f ∗ (C[x])].
Since z n + xn + 1 is irreducible and monic in z,
[CFn : C[x]] = [C[x, z]/hz n + xn + 1i : C[x]] = n.
Moreover, as a C[xn ]−vector space, C[x] has dimension n since f ∗ : x 7→ xn , and

f ∗ : a 7→ a for all a ∈ C. So
[C[x] : f ∗ (C[x])] = [C[x] : C[xn ]] = n.
Therefore, we get deg f = n2 .

P1 (C) with (x, y, z) 7→ (−xn , y n ). Then f is ramified over 0, 1, ∞ and |P ∈ C(Q) :
f (P ) ∈ {0, 1, ∞}| = 3n.
89
Proof. • The value 0 ∈ C is represented by a pair (−xn , y n ) where −xn = 0 with

y 6= 0. Looking at the affine plane with y = 1, we have x = 0 and since xn + z n =
−1, there are n distinct solutions for z, namely nth roots of −1. Hence there are
n preimages for the value 0.
• The value 1 ∈ C is represented by a pair (−xn , y n ) where −xn = y n with y 6= 0.
In the affine plane with y = 1, when −(x/y)n = 1, we have z = 0 and hence
there are n distinct solutions for x, which are nth roots of −1. Hence there are n
preimages for the value 1.
• The value ∞ ∈ P1 (C) is represented by a pair (−xn , y n ) where y = 0, while
−xn 6= 0. Then looking at the affine plane with x = 1, we have z n = −1, hence
there are n distinct solutions for z, namely nth roots of −1. So there are n
preimages for the value ∞.
We see that f is ramified above 0, 1, ∞ since the number of the preimages for
these points is less than the degree of the map.
Definition 6.21 ( [11], Section 3.1). The field of maps f : C(C) → P1 (C), has the
valuations vx (f ) = −ordx (f ) for each point of x ∈ C(C). For a nonconstant map
f : C(C) → P0 (C), we define the height and the radical of P = (f, 1 − f, 1) ∈ P2 (C(C))
by h(P ) = deg f, and log N (P ) = |f −1 ({0, 1, ∞})|.

P1 (C) with (x, y, z) 7→ (−xn , y n ). Then Fn has finitely many rational points.
Proof. Using Proposition 6.19 and Proposition 6.20, we have
3n = |P ∈ C(Q) : f (P ) ∈ {0, 1, ∞}| < deg f = n2
since n > 3. So abc conjecture is regarded as a bound on the ”ramification” of the

rational number r above 0, 1, and ∞. Since we have log N < log H, we conclude that
the Fermat curve has only finitely many rational points.
90
In general, it is rare for there to be a rational function f on a curve C whose

degree exceeds the size of f −1 ({0, 1, ∞}) by a large factor. As a result, we expect that
any curve C over an arbitrary number field K, provided there is a rational function
f ∈ K(C) such that
|P ∈ C(Q) : f (P ) ∈ {0, 1, ∞}| < deg f,
has only finitely many K−rational points.
Remark 6.23. When the curve has genus 0, by Hurwitz Formula 2.74, we have
X
−2 = −2deg f + (deg f − |f −1 (Q)|).
Q
By separating the ramification points, we get
X X
−2 + 2deg f = (deg f − |f −1 (Q)|) + (deg f − |f −1 (Q)|)
Q∈{0,1,∞} Q∈{0,1,∞}
/
X
= 3deg f − |f −1 ({0, 1, ∞})| + (deg f − |f −1 (Q)|),
Q∈{0,1,∞}
/
since there may be ramification points other than 0, 1, and ∞. Then,
X
|f −1 ({0, 1, ∞})| = deg f + 2 + (deg f − |f −1 (Q)|) > deg f.
Q∈{0,1,∞}
/
Again by Hurwitz Formula 2.74, when the curve has genus 1, we have
X X
0 = −2deg f + (deg f − |f −1 (Q)|) + (deg f − |f −1 (Q)|),
Q∈{0,1,∞} Q∈{0,1,∞}
/
91
then
X
|f −1 ({0, 1, ∞})| = deg f + (deg f − |f −1 (Q)|) ≥ deg f.
Q∈{0,1,∞}
/
Therefore, we cannot find such f ∈ K(C) when C has genus < 2.
Proposition 6.24. Once the curve has genus g ≥ 2, we can find such a rational
function f satisfying deg f > |f −1 ({0, 1, ∞})|.
Proof. The proof of Belyi’s theorem 2.121 gives an effective procedure for obtaining a
rational function f ∈ K(C) ramified only above 0, 1, ∞. For a map that is only ramified
over 0, 1 and ∞, using Hurwitz Formula 2.74, we have
X
2g − 2 = −2deg f + (deg f − |f −1 (Q)|) = deg f − |f −1 ({0, 1, ∞})|,
Q∈{0,1,∞}
hence |f −1 ({0, 1, ∞})| < deg f, since g ≥ 2.

92
6.3. The Proof: Abc Implies Mordell
We prove Mordell conjecture using abc conjecture as in the paper of Elkies [1],
which gives an effective height bound for the points on the curve.
Theorem 6.25. Assume abc conjecture, then any curve of genus at least 2 over a
number field K has only finitely many K-rational points.
First, we prove a technical proposition which is one of the key points of the proof.
Notation: We write f (x) = O(g(x)) for functions f (x) and g(x) if and only if
there exist constants C and k such that |f (x)| ≤ C|g(x)| for all x > k.
Proposition 6.26 ( [1]). Let C be any curve over K and f ∈ K(C) be a rational
function of degree d. Then for any K-rational point P ∈ C(K) − f −1 (0) we have

bf (0) p
log N0 (f (P )) < 1− log HP + O( log HP + 1)
d
with the implied constant effective and depending on K, C, f but not on P and HP :=
H(f (P )).
P
Proof. We use logarithmic height functions relative to divisors. Let D = mk Dk be
k
the zero divisor of f, where Dk are prime rational divisors of degrees dk . Then we have
deg(D) = mk dk = degf = d and bf (0) = d − dk . Let D0 = Dk , then deg(D0 ) =
P P P
k k k
dk and we can express the ramification above 0 as bf (0) = deg(D) − deg(D0 ). By
P
k
Proposition 2.115 and by additivity of Theorem 2.113, we have
X
log HP = log H(f (P )) = hD (P ) + O(1) = mk hDk (P ) + O(1).
k
93
Eliminating the finitely many bad primes as in the Subsection 2.3.5, a prime v
contributes to log N0 (f (P )) only if f (P ) ≡ 0 mod v, i.e., ordv (f (P )) > 0, which holds
also for the reduction f (P ) ≡ 0 mod v. In this case, P is in the support of D. Since
the decomposition of D remain the same when reducing mod v, i.e., no coalescence or
no vanishing, P is in the support of some Dk . Moreover, P is in the support of some
Dk , hence the prime contributes to hDk (P ) for some k. By the same reasoning applied
backwards, we see that a prime contributes to logN0 (f (P )) if and only if it contributes
to hDk (P ) for some k.
Now since the contribution of any prime, finite or infinite, to the height associated
to an effective divisor is bounded below, we have hDk (P ) ≥ O(1), by positivity of
Theorem 2.113. Moreover by definition, log |f (P )|v is a multiple of log(N(v)), i.e.,
|hDk (P )| ≥ log(N(v)). Then summing over all primes, we obtain
X
log N0 (f (P )) < hDk (P ) + O(1) = hD0 (P ) + O(1),
k
with the second equation following from additivity of Theorem 2.113 and < accounting
for finite primes at which P and D0 may meet nontransversally and infinite places at
which P may come close the support of D0 .
Lastly, we construct a degree zero divisor to make use of the bound in the Theorem
2.116. So let E = (deg(D))D0 − (deg(D0 ))D. Then using the additivity of Theorem
2.113,
hE (P ) = (deg(D))hD0 (P ) − (deg(D0 ))hD (P ) + O(1)
and by Proposition 2.115 and Theorem 2.116,
p
= (deg(D))hD0 (P ) − (deg(D0 ))h(f (P )) + O(1) ≤ O( log HP + 1),
94
so
deg(D0 ) p
hD0 (P ) ≤ h(f (P )) + O( log HP + 1).
deg(D)
Combining with the above inequality, we get

bf (0) p
log N0 (f (P )) < 1− log HP + O( log HP + 1)
d
since bf (0) = deg(D) − deg(D0 ).
Proof of Theorem 6.25. Let C be an arbitrary curve of genus g ≥ 2. We fix a function

f ∈ K(C), satisfying |P ∈ C(Q) : f (P ) ∈ {0, 1, ∞}| < deg f, which exists by Belyi’s
theorem 2.121 and the inequality is satisfied once g ≥ 2 by Proposition 6.24. Let
m := |P ∈ C(Q) : f (P ) ∈ {0, 1, ∞}| and d := deg f. Using Proposition 6.26, for any
K-rational point P ∈ C(K) − f −1 (0), we have

bf (0) p
log N0 (f (P )) < 1 − log HP + O( log HP + 1).
d
Replacing f by f − 1 in the Proposition 6.26, we get that for all P ∈ C(K) −

(f − 1)−1 (0), i.e., P ∈ C(K) − f −1 (1),

bf (1) p
log N1 (f (P )) < 1 − log HP + O( log HP + 1),
d
since N0 ((f − 1)(P )) = N0 (f (P ) − 1) = N1 (f (P )) by definition and log H(f (P )) =

log H(f (P ) − 1) + O(1) by Remark 2.105.
Similarly, replacing f by 1/f in the Proposition 6.26, we get that for all P ∈
C(K) − (1/f )−1 (0), i.e., P ∈ C(K) − f −1 (∞),

bf (∞) p
log N∞ (f (P )) < 1 − log HP + O( log HP + 1),
d
95
since N0 ((1/f )(P )) = N0 (1/f (P )) = N∞ (f (P )) by definition and log H(f (P )) =

log H(1/f (P )) due to H being scaling invariant.
Then for any K-rational point P ∈ C(K) not in the finite set f −1 ({0, 1, ∞}), we
have:
m p
log NP < log HP + O( log HP + 1),
d
with the implied O-constant effective and depending on K, C, f but not on P, since
log N0 (f (P )) + log N1 (f (P )) + log N∞ (f (P )) = log N (f (P )) and
P

bf (0)

bf (1)

bf (∞)
degf − f (P )=0 (e f (P ) − 1)
1− + 1− + 1− = +
d d d d
P P
degf − f (P )=1 (e f (P ) − 1) degf − f (P )=∞ (ef (P ) − 1)
+ =
d d
|f −1 (0)| + |f −1 (1)| + |f −1 (∞)| |f −1 ({0, 1, ∞})| m

= = .
d d d
Taking logarithms of both sides in abc conjecture, it becomes (1 − ) log H ≤ log N + C

for some constant C . Then f (P ) gives a counter example to abc conjecture over K for
< 1 − (m/d) once HP is large enough, i.e., for all but finitely many P. In particular,
combining with the inequality, we get the effective height bound
√
O( log HP + 1)
log HP <
1 − − m/d
for the points P ∈ C(K) not in the finite set f −1 ({0, 1, ∞}). Since f −1 ({0, 1, ∞}) is
finite, by Theorem 2.107, we are done.
Remark 6.27. Using Theorem 2.117 instead of Theorem 2.116 in the last paragraph
above, we can get a different height bound on the points.
96
Proposition 6.28. Let C be any curve over K and f ∈ K(C) be a rational function
of degree d. Then for all > 0, there is a constant C such that for any K-rational
point P ∈ C(K) − f −1 (0) we have

bf (0)
log N0 (f (P )) < 1− (1 + ) log HP + C .
d
Proof. With the same set-up constructed above, instead of defining a degree zero di-
visor, we use the inequality from Theorem 2.117 for the divisors D and D0 . Then
for all > 0, there is a constant C such that for all P ∈ C(K) − f −1 (0), we have
degD0
hD0 (P ) ≤ (1 + ) hD (P ) + C . Then using Theorem 2.115, we get
degD

bf (0)
log N0 (f (P )) < 1− (1 + ) log HP + C ,
d
since bf (0) = degD − degD0 .
Proof of Theorem 6.25. Replacing f by f − 1 and 1/f in Proposition 6.28 and then
adding all inequalities we obtained, we get for all points P ∈ C(K) not in the finite
set f −1 ({0, 1, ∞}),
m
log NP < (1 + ) log HP + 3C .
d
Combining with abc conjecture, which says that (1 − 0 ) log HP ≤ log NP + C0 , we
obtain

0 (1 + )m
1− − log HP < C + C0 .
d
(1 + )m
Choose , 0 > 0 so that 1 − 0 − > 0. Then we have the height bound
d
log HP < C with constant C depending on , 0 . By Theorem 2.107, we are done.
97
6.4. Examples
We give various examples using two different height bounds we obtained in the
last section.
Example 6.29. Let F5 be the Fermat curve of genus 6, computed by Example 2.77,
defined over Q by the equation x5 + y 5 + z 5 = 0. Let f : F5 (Q) → P0 (C) with (x, y, z) 7→
(−x5 , y 5 ) be a rational function. By Proposition 6.19 and Proposition 6.20, we know
that deg f = 25 and f is ramified over {0, 1, ∞} with ramification degrees 5. Using
Hurwitz Formula 2.74, we see that f is only ramified over {0, 1, ∞} and nowhere else.
P
Moreover, using deg(f ) = ef (P ), we have |{P ∈ F5 (Q) : f (P ) ∈ {0, 1, ∞}}| = 15.
P
Then f is a Belyi function satisfying |{P ∈ F5 (Q) : f (P ) ∈ {0, 1, ∞}}| < deg f.
Hence letting = 1/5 , for all P ∈ F5 (Q) such that f (P ) 6= 0, 1, ∞, we get log HP <
√ √
5O( log HP + 1) or log HP < O( log HP + 1), since Big-Oh function is a set.
Now let x, y, z be relatively prime integers, then f (P ) = (−x5 , y 5 ) or we could

write as f (P ) = (−x5 /y 5 : 1) since f (P ) 6= ∞. Then log HP = log H(f (P )) =
log max(| − x5 |∞ , |y 5 |∞ ) because for finite primes p, we have | − x5 |p ≤ 1, |y 5 |p ≤ 1, but
gcd(x, y) = 1 and we choose 1 for any finite prime. Therefore log HP is equal to either
log(±x5 ) or log(±y 5 ). Without loss of generality, let log HP = log(x5 ) where x ∈ Z>0
p
and x > |y|∞ . Plugging in the height bound, we get log(x5 ) < O( log(x5 ) + 1).
For any candidates of constants C > 0 and k, we can take x > k and we would
p
have to satisfy | log(x5 )| < C| log(x5 ) + 1|, which gives us the interval 1 < x <
p
2.71828h(C) , where h(C) = 0.1 C 3 (C + 4) + 0.1C 2 + 0.2C, for a fixed C. Since there
exist finitely many integers in an interval, we have finitely many options for x. There
exist also finitely many options for y, since x > |y|∞ . Then there are finitely many
points P ∈ F5 (Q) such that f (P ) 6= 0, 1, ∞, since z is dependent on x and y with the
equation x5 + y 5 + z 5 = 0. In the end, the Fermat curve F5 has only finitely many
rational points.
98
Example 6.30. Let C : y 2 = x7 − 1 be defined over Q, of genus 3. We fix the rational

map
f : C → P0
(x, y) 7→ (x6 + x5 + x4 + x3 + x2 + x + 1, 7)
which is a Belyi function by Example 2.123.
We are interested in points P ∈ C(Q) with f (P ) 6= 0, 1, ∞, so we are in the affine

plane with z = 1. Choose x, y relatively prime integers, then log HP = log max(|x6 +
x5 + x4 + x3 + x2 + x + 1|∞ , 7).
(i) When |x6 + x5 + x4 + x3 + x2 + x + 1|∞ ≤ 7, we have log HP = log 7 and using

√
the height bound, we obtain the inequality log 7 < O( log 7 + 1), which is a true
statement. Therefore, the integers satisfying |x6 + x5 + x4 + x3 + x2 + x + 1|∞ ≤ 7
are options for the variable x, which are −1, 0, and 1.
(ii) When |x6 + x5 + x4 + x3 + x2 + x + 1|∞ ≥ 7, we have log HP = log(|x6 + x5 +
x4 + x3 + x2 + x + 1|∞ ). Without loss of generality, choose x > 1, then using
the height bound, we obtain the inequality log(x6 + x5 + x4 + x3 + x2 + x + 1) <
p
O( log(x6 + x5 + x4 + x3 + x2 + x + 1) + 1), which gives us an interval. Then
there exist finitely many options for x, since there are finitely many integers in
an interval.
Since y depends on x, we say that the curve C has finitely many rational points.
Note that the functions in the inequality of Example 6.30 grows more slowly with
respect to the ones of Example 6.29. For example, when C = 600 in Example 6.30, we
have −8.59523 < x < 8.25943 as an interval; whereas, when C = 50 in Example 6.29,
we have 1 ≤ x < 5.618 × 10225 as an interval. Hence it is easier to find the rational
solutions of y 2 = x7 − 1 by checking out of the points satisfying the height bound.
99
√ 2 √
Example 6.31. Let C : (y − 2) = x5 − 1 be defined over Q( 2), of genus 2. We fix
the rational map
f : C → P0
(x, y) 7→ (x4 + x3 + x2 + x + 1, 5)
which is a Belyi function by Example 2.124.
√
Take arbitrary P = (x, y) ∈ C(Q( 2)) with f (P ) 6= 0, 1, ∞, then
x4 + x3 + x2 + x + 1 √
f (P ) = ∈ Q( 2), and
5
Y
HP = max{|x4 + x3 + x2 + x + 1|v , |5|v },
v∈MQ(√2)
√
where MQ(√2) 0 = | |P determined by finite primes P of Q( 2) and MQ(√2) ∞ = | |φi
√ √ √ √ √
determined by embeddings φ1 : a + b 2 7→ a + b 2, φ2 : a + b 2 7→ a − b 2 of Q( 2)
into C.
√
Note that for 5, we have |5|φ1 = |5|φ2 = |5|∞ = 5. Moreover, 5Z[ 2] = P with
norm N(P) = 25 by Example 2.34. Then we have |5|P = 1/25 and |5|Q = 1 for all the
other primes Q.
We separate into cases:
(i) Let x4 + x3 + x2 + x + 1 =: a ∈ Z,
• If (a, 5) = 1, then log HP = log max{|a|φ1 , 5} + log max{|a|φ2 , 5} =

2 log max{|a|∞ , 5}. Therefore, when |a|∞ ≤ 5, we have log HP = 2 log 5 =
√
log 25. Using the height bound, we get log 25 < O( log 25+1), which is a true
statement. So the integers satisfying |x4 + x3 + x2 + x + 1|∞ ≤ 5 are options
for x. On the other hand, when |a|∞ > 5, we have log HP = 2 log |a|∞ . Using
100
p
the height bound, we get log a2 < O( log a2 + 1), which gives us an interval.
Since there are finitely many integers in an interval, there are finitely many
options for x.
x4 + x3 + x2 + x + 1
• If (a, 5) > 1, then =: b ∈ Z and H(a, 5) = H(b, 1). We
5
get log HP = log max{|b|φ1 , 1} + log max{|b|φ2 , 1} = 2 log max{|b|∞ , 1}, but
f (P ) 6= 1, so we have one case, which is |x4 + x3 + x2 + x + 1|∞ > 5.
In that case, we get log HP = 2 log |b|∞ . Using the height bound, we obtain
p
log b2 < O( log b2 + 1), which gives us an interval. Since there are finitely
many integers in an interval, there are finitely many options for x.
√ √
(ii) Let x4 + x3 + x2 + x + 1 =: a + b 2 ∈ Q( 2) with b 6= 0.
√ √
• For archimedean valuations, we have |a + b 2|φ1 = |a + b 2|∞ and |a +
√ √ √ √
b 2|φ2 = |a − b 2|∞ . Then if |a + b 2|∞ > 5, we choose |a + b 2|∞ as
maximum in H for valuation | |φ1 ; otherwise, we choose 5. For the other
√ √
archimedean valuation, if |a − b 2|∞ > 5, then we choose |a − b 2|∞ for
| |φ2 ; otherwise, we choose 5.
√
• We look into nonarchimedean valuations. The element a + b 2 forms the
√ √ √ √
ideal (a + b 2)Z[ 2] with norm N((a + b 2)Z[ 2]) = a2 − 2b2 .
√
– If 2k | a2 − 2b2 for some k ∈ Z, then |a + b 2|h2,√2i = (1/2)k , since
√ √ 2 √
2Z[ 2] = h2, 2i with N(h2, 2i) = 2. Moreover, we have |5|h2,√2i = 1.
Therefore, when k ≥ 1, we choose 1 for | |h2,√2i ; otherwise we choose
(1/2)k .
√
– If 5k | a2 − 2b2 for some k ∈ Z, then |a + b 2|P = (1/25)k/2 = (1/5)k .
Moreover, we have |5|P = 1/25 = (1/5)2 . Therefore, when k ≥ 2, we
choose (1/5)2 for | |P ; otherwise we choose (1/5)k .
– When a prime q ≡ ±3 mod 8 with q 6= 5, it stays prime above, i.e., Q ∈
√
qZ[ 2], by Quadratic Reciprocity and Example 2.34. Say q k | a2 − 2b2
√
for some k ∈ Z, then |a + b 2|Q = (1/q)k . Moreover, we have |5|Q = 1.
Therefore, when k ≥ 1, we choose 1 for | |Q ; otherwise we choose (1/q)k .
– When a prime q ≡ ±1 mod 8, it splits into two prime ideals Q1 , Q2
√
above with Q1 k1 Q2 k2 ∈ qZ[ 2], by Quadratic Reciprocity and Example
√
2.34. Say q k | a2 − 2b2 for some k = k1 + k2 ∈ Z, then |a + b 2|Q1 =
101
√
(1/q)k1 and |a + b 2|Q2 = (1/q)k2 . Moreover, we have |5|Q1 = |5|Q2 = 1.
Therefore, if k1 ≥ 1, we choose 1 for | |Q1 ; otherwise we choose (1/q)k1 .
Also, if k2 ≥ 1, we choose 1 for | |Q2 ; otherwise we choose (1/q)k2 .
In the end, we get 27 cases for the height log HP . Note that there is no possibility of
having a value less than 1 inside log, hence a negative value for log HP . Using Theorem
6.25 with the height bound in Proposition 2.117, for each case, we get log HP < C
for some constant C. Hence, by Theorem 2.107, there are only finitely many points
√ √
P ∈ C(Q( 2)) with f (P ) 6= 0, 1, ∞. Since |f −1 ({0, 1, ∞})| is finite, C(Q( 2)) is
finite.
102
7. CONCLUSION
In this thesis, we started by giving a long Preliminaries section providing the

background necessary for further proofs and examples. We introduced many concepts
and tools such as absolute values, height functions, Belyi maps and so on. Next, we
worked on abc conjecture, which has been a hot topic in mathematics lately, and its
relation between other conjectures, and theorems such as Mordell, Hall, and so on.
Apart from its bridge role between the operations of addition and multiplication, abc
conjecture provides such simple and elegant proofs for strong theorems. For exam-
ple, we showed that one can prove Fermat’s last theorem for large exponents using
abc conjecture in one elementary paragraph, whereas Wiles had to prove Modularity
conjecture for semi stable elliptic curves. We also gave the proof of Mason-Stothers
theorem, which is the analogue of abc conjecture for polynomials.
Later, in Chapter 4, we introduced Hall conjecture, which puts a bound on the

size of the integral points on a Mordell curve. We gave an example on how to put a
bound on the size of those points using the properties of elliptic curves, by showing
that a Mordell curve is in fact an elliptic curve. Then we gave the proof of Weak Hall
conjecture using abc conjecture.
Finally, after giving the statement of Mordell conjecture, now Faltings’ theorem,
we moved on to the proof of Mordell conjecture using abc conjecture, which is a great
result from the article “Abc Implies Mordell” by Noam Elkies, [1]. The proof represents
a great connection between the theory of Diophantine approximation and the theory
of points on curves of high genera. Before giving the proof, we gave the ideas which
created the motivation for the proof. The effective proof of Belyi and the proof showing
that the Fermat curve has no solutions, by using a rational map on the curve, were two
of them. Then, we presented the proof using abc conjecture, given by Noam Elkies [1],
which is of great importance because it is effective. The proof of Faltings is not effective;
it proves that there are finitely many rational points on a curve of genus > 1, but does
not give a method for finding them or give an upper bound on the size of the points. On
103
the other hand, the proof by Elkies provides an effective upper bound on the size of the
points, hence limits the search criteria for finding the points. We also acquired another
height bound by using a different theorem, bounding the height functions associated
to divisors, with the technical proposition by Elkies [1]. In the end, we gave examples
using both of the height bounds we obtained.
104
REFERENCES
1. Elkies, N. D., “ABC implies Mordell”, International Mathematics Research No-

tices, Vol. 1991, No. 7, pp. 99–109, 1991.
2. Goldfeld, D., “Beyond the last theorem”, Math Horizons, Vol. 4, No. 1, pp. 26–34,
1996.
3. O’Neil, C., “Notes on the Oxford IUT workshop

by Brian Conrad”, https://mathbabe.org/2015/12/15/
notes-on-the-oxford-iut-workshop-by-brian-conrad/, accessed in February
2018.
4. Milne, J. S., “Algebraic Number Theory (v3.07)”, www.jmilne.org/math/, ac-

cessed in May 2018.
5. Conrad, K., “Ostrowski for number fields”, https://kconrad.math.uconn.edu/

blurbs/gradnumthy/ostrowskinumbfield.pdf, accessed in February 2018.
6. Lorenzini, D., An invitation to arithmetic geometry, Vol. 9, American Mathemat-

ical Soc., 1996.
7. Conrad, K., “Discriminants and ramified primes”, https://kconrad.math.

uconn.edu/blurbs/gradnumthy/disc.pdf, accessed in March 2018.
8. Silverman, J. H., The arithmetic of elliptic curves, Vol. 106, Springer Science &
Business Media, 2009.
9. Bruin, N., “Generalization of the ABC-conjecture”, Citeseer , 1995.
10. Pazuki, F., “Heights”, https://www.math.u-bordeaux.fr/~abesheno/heights.

pdf, accessed in May 2018.
105
11. Van Frankenhuysen, M., “The ABC conjecture implies Roth’s theorem and
Mordell’s conjecture”, Math. Contemp, Vol. 16, pp. 45–72, 1999.
12. Lozano-Robledo, Á., Elliptic curves, modular forms, and their L-functions, Amer-
ican Mathematical Society Providence, RI, 2011.
13. Hindry, M. and J. H. Silverman, Diophantine geometry: an introduction, Vol. 201,

Springer Science & Business Media, 2013.
14. Brown, M., J. P. Serre and M. Waldschmidt, Lectures on the Mordell-Weil theorem,
Springer, 1989.
15. Lang, S., Fundamentals of Diophantine geometry, Springer Science & Business
Media, 2013.
16. Ines, M., “Belyi’s theorem”, https://alexjbest.github.io/buntes/

sec-belyi-thm.html, accessed in May 2018.
17. Waldschmidt, M., “On the abc Conjecture and some of its consequences”, Math
Horizons, February 2018.
18. Lang, S., Math talks for undergraduates, Springer Science & Business Media, 2012.
19. Conrad, K., “Examples of Mordell’s equation”, https://kconrad.math.uconn.

edu/blurbs/gradnumthy/mordelleqn1.pdf, accessed in April 2018.
20. Ho, W., “How many rational points does a random curve have?”, Bulletin of the
American Mathematical Society, Vol. 51, No. 1, pp. 27–52, 2014.

603727

Uploaded by

Copyright:

Available Formats

603727

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

603727

Uploaded by

Copyright:

Available Formats

ABC CONJECTURE AND ITS IMPLICATIONS

Submitted to the Institute for Graduate Studies in

Graduate Program in Mathematics

ABC CONJECTURE AND ITS IMPLICATIONS

Assoc. Prof. Ekin Özman ...................

Prof. K. Ilhan Ikeda ...................

Assist. Prof. Ayberk Zeytin ...................

DATE OF APPROVAL: 21.06.2019

I thankfully acknowledge the financial support given by TÜBİTAK PIA Bospho-

ABC CONJECTURE AND ITS IMPLICATIONS

ABC VARSAYIMI VE ONUN ÇIKARIMLARI

Bu tezde, amacımız abc varsayımının önemini anlatmak ve onu kullanarak birçok

Figure 4.1. An example of Mordell curve . . . . . . . . . . . . . . . . . . . . . 61

Gal(K/K) the Galois group of K/K

OK the ring of integers of K

∆E discriminant of an elliptic curve E

φD rational map associated to the divisor D

⊕ group law on an elliptic curve

Conjecture. But if correct it would nonetheless be a tremendous achievement, setting

Later, as another important result of abc conjecture, we introduce Hall conjecture

2.1. Absolute Values

Definition 2.1. An absolute value or multiplicative valuation on a field K is a real-

(i) |x|v ≥ 0 with equality if and only if x = 0

Note that Cv ≥ 1, which can be easily seen by taking y = 0 in (iii). If we can

By the first two conditions, we see that | |v : K ∗ → R>0 is a homomorphism

A discrete (additive) valuation ord on K determines an absolute value by |x|v =

Proposition 2.3. Let | |v be a nontrivial nonarchimedean absolute value, and put

• v(xy) = v(x) + v(y),

If v(K ∗ ) is discrete in R, then v is a multiple of a discrete valuation ord : K ∗ ,→ Z ⊂ R.

Definition 2.4. Let K be a field. A discrete valuation on K is a nonzero homomor-

For a real embedding φ : K ,→ R, |a|φ := |φ(a)|∞ , with | |∞ usual real absolute

For a pair of complex-conjugate embeddings (φ1 , φ2 ) : K ,→ C, |a|(φ1 ,φ2 ) =

3  1 ord3 (3/5) 1 3  1 ord5 (3/5)

Let P be a nonzero prime ideal in OK . If α ∈ K ∗ and ordP (α) ≥ 0, then α = x/y

Proposition 2.8 ( [4], Chapter 7, Proposition 7.8). Let | |v , | |w be nontrivial absolute

(i) | |v and | |w define the same topology on K;

In this case | |v and | |w are said to be equivalent.

There is a classification theorem on valuations on the field of rational numbers:

Theorem 2.9 (Ostrowski Theorem, [5]). Let | |v be a nontrivial absolute value on Q.

• If | |v is archimedean, then | |v is equivalent to | |∞ .

Theorem 2.10 (Product Formula for Q). For p = 2, 3, 5, 7, . . . , ∞, let | |p be the

This is a simple reflection of the fact that Z has unique factorization.

There is also a classification theorem of valuations for number fields:

Theorem 2.12 (Product Formula for K). Let a ∈ K ∗ . Then

2.2. Prime Ideals and Factorization

Let K be a field and K be an algebraic closure of K. In the first subsection

Let f (x, y) ∈ K[x, y] be a polynomial in two variables with coefficients in K and

Zf (K) := {(a, b) ∈ K × K|f (a, b) = 0}.

Definition 2.16. Let f ∈ K[x, y] be an irreducible polynomial, so that the ring Cf :=

Let f (x, y) = an (x)y n + · · · + a0 (x) be an irreducible polynomial in K[x, y]. The

φ : K[x] → Cf := K[x, y]/hf i

is injective if f (x, y) 6= cx + d, for all c, d ∈ K. In this case, it can be extended to an

Fact 2.18. Let A be a domain of dimension 1. Let B be a domain containing A and

Fact 2.19. Let A ⊆ B be two rings. If A is noetherian and B is a finitely generated

When an (x) = 1, every element of Cf is integral over K[x], and hence Cf is an

2.2.2. Points and Maximal Ideals

Corollary 2.21. Let f ∈ K[x, y] be an irreducible polynomial. An ideal M of the ring

2.2.3. Morphisms of Curves

Any such continuous map φ : Zf (K) → Zg (K) defines a map:

φC ∗ : C(Zg (K), K) → C(Zf (K), K)

3 1 ord3 (3/5) 1 3 1 ord5 (3/5)