Diophantine Geometry and Dynamical Systems

Antoine Chambert-Loir
Irmar, Université de Rennes 1, Campus de Beaulieu, F-35042 Rennes Cedex, France
Courriel : antoine.chambert-loir@univ-rennes1.fr

Abstract. — Solving Diophantine equations, that is finding the solutions in rational integers of polyno-
mial equations is one of the oldest task of mathematicians. During the 20th century, this subject has been
geometrized so as to become what is now known as Diophantine Geometry.
A lot of questions have been asked, and many beautiful answers have been given, by mathematicians
like Mordell, Weil, Manin, Mumford, Lang, Bogomolov, Faltings, Ullmo, Zhang... They mostly concern sub-
varieties of Abelian varieties, torsion points, or subgroups of finite rank. The notion of height often plays a
crucial rôle.
Around 1990, notably under the impulse of Silverman, some of those questions have been broadened so
as to replace Abelian varieties by Algebraic dynamical systems, namely algebraic varieties together with a
self-map turning all of the old answers into totally open new questions.
In this talk, I shall brush a picture of this remarkable story.

1. Diophantine equations
Diophantine equations are polynomial equations where the unknowns are integers.
When the degree of the equations is 1, they are basically solved using the so called Chinese
Remainder Theorem. Indeed, its first appearance can be found in an old text (III-Vth century
A . D., the precise date is unknown) by Sūnzı̌, called the Sūnzı̌ Suàngjı̄ng 孫子算經, “The mathe-
matical classic of Sūnzı̌”: (see Fig. 1):
有物不知其數，三三數之剩二，五五數之剩三，七七數之剩二。問物几何？
which translates as:
Now there are an unknown number of things. If we count by threes, there is a re-
mainder 2; if we count by fives, there is a remainder 3; if we count by sevens, there is
a remainder 2. Find the number of things. — Translation from Lam & Ang (2004)
This kind of exercise has been reprinted in later mathematical manuals, such as the 1247
Shùshū Jiǔzhāng (數書九章, Mathematical Treatise in Nine Sections), itself included in the
XIX th century’s Sìkù quánshū (四庫全書, The Complete Library of the Four Treasures, see
Fig. ??), a kind of encyclopaedia commissioned by the Qing emperors to show they could
surpass the Ming’s Encyclopaedia (from 1403 A . D.). Its modern mathematical formulation
would be (the main unknown is n)
n = 3x + 2 = 5y + 3 = 7z + 2,
whose smallest positive solution is n = 23.
In degree 2, the most prominent example is probably the Pell equation x 2 − n y 2 = 1, the
unknowns being x and y, and n being a non-zero parameter which is not a square. In fact, this
equation was first studied by the Indian mathematician Brahmagupta (628 A . D.) who already
observed that from two solutions (x, y) and (x 0 , y 0 ), one could define a third one (x 00 , y 00 ), given
by
x 00 = xx 0 + n y y 0 , y 00 = x y 0 + x 0 y.
2 ANTOINE CHAMBERT-LOIR

p p
This relation is better understood if one writes x 2 − n y 2 = (x + n y)(x − n y) as the norm of
p
x + n y, and observes that
p p p
(x + n y)(x 0 + n y 0 ) = (xx 0 + n y y 0 ) + n(x y 0 + x 0 y).
In fact, all solutions can (up to sign) be constructed by this process from a single, minimal, one.
This equation also intervenes in the solution to Archimedes’s cattle problem (IIIrd century B. C .,
Fig. 3).
The study of equations of degree 3 or more is much more recent. One must of course quote
Fermat’s “Last Theorem”, staten in 1637 and first proven by Andrew Wiles (partly with Richard
Taylor) in 1995. Fermat was in fact studying the works of Diophantus of Alexandria who solved
the quadratic equation in three integer variables a, b, c,
a2 + b2 = c 2,
whose solutions give the possible triangles having a right angle and sides with integer lengths.
In the margin of his own edition (see Fig. 4 for a later edition incorporating Fermat’s com-
ments), Fermat claimed to have proven that similar higher degrees equations,
a3 + b3 = c 3, a4 + b4 = c 4, . . .
have no solutions besides the obvious ones where one of the variable is zero. Although mathe-
maticians like Fermat himself, Euler, Sophie Germain, Kummer,... proved many cases, the full
proof of this statement had to await much more modern tools.
In his proof of the quartic case, Fermat invented the method of infinite descent. Assuming
there is a nontrivial solution (a, b, c) of equation a 4 +b 4 = c 4 and consider the smallest one (one
for which, say, c is minimal), he used the parametrization of solutions of the quadratic equation
A 2 +B 2 = C 2 to construct a smaller solution. This is a contradiction which proves that there are
no nontrivial solutions.
In fact, it is simpler to show that the equation a 4 + b 4 = c 2 has only trivial solutions. So take a, b, c
positive integers such that a 4 + b 4 = c 2 , with c minimal. Necessarily a, b, c are coprime; set A = a 2 ,
B = b 2 , C = c. One has A 2 + B 2 = C 2 hence (assuming that A is odd and B is even, which is possible,
up to exchanging a and b), there are coprime integers U and V , such that
A = U 2 − V 2, B = 2UV, C = U 2 + V 2.
In other words,
a2 + V 2 = U 2, b 2 = 2UV, c = U 2 + V 2.
Looking modulo 4, we see that V is even and U is odd. Since U and V are coprime and b 2 = 2UV ,
there are integers u and v such that U = u 2 and V = 2v 2 ; then, b = 2uv. From the equation a 2 + V 2 =
U 2 , we see that there are coprime integers X , Y such that a = X 2 − Y 2 , V = 2X Y , U = X 2 + Y 2 . Then
v 2 = X Y and, X and Y being coprime, they both must be squares: X = x 2 , Y = y 2 . Finally, we get
U = u 2 = x 4 + y 4 which is an equation of the same form as the one studied. Moreover, one checks that
c = U 2 + V 2 > U 2 > U , which contradicts the minimality of c.

2. Birth of Diophantine geometry: curves and abelian varieties

During the XIXth century, the geometry of algebraic curves and abelian varieties has emerged
from complex analysis under the hands of Riemann, Jacobi, Abel, and others.
Concerning curves, it was understood that the degree of a plane curve does not fully reflect
its geometry, unless the curve is without singularities. The genus is a better invariant. In that
respect, equations of degree 1 or 2 furnished curves of genus 0; equations of degree 3 or more
tend to give curves of higher genus.
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 3

Elliptic curves are curves of genus one, together with a point o on it. Over the complex num-
bers, such a curve (X , o) has a cubic plane equation of Weierstrass’s form, namely
y 2 = x 3 = ax + b
(in the affine plane; homogeneize to get the projective equation), the point o being the point
at infinity. It can be parametrized via Weierstrass’s ℘-function: there exists a lattice Λ ∈ C such
that the map
µ ¶
0 1 1 1
z 7→ (℘(z), ℘ (z)), ℘(z) = 2 +
X
− 2
ω∈Λ\{0} (z − ω) ω
z 2

∼
is Λ-periodic and induces an isomorphism C/Λ − → X . Observe that this parametrization en-
dows X with a structure of an Abelian group, the point o being the origin. In fact, this group
structure has a well-known geometric construction: take two points (x, y) and (x 0 , y 0 ); the line
through them intersects the curve in a third point (x 00 , y 00 ), the symmetric of which with respect
to the x-axis is the sum of the two points (x, y) and (x 0 , y 0 ).
The question of finding rational (resp. integer) solutions to such cubic Weierstrass’s equa-
tions is therefore expressed as finding rational points on elliptic curves (resp. integral points
on X \ {o}).
The results are as follows:

Theorem 2.1. — Let X be an elliptic curve given by a Weierstrass equation with coefficients in a
number field F .
The set of rational points of X is an Abelian group of finite type (Mordell (1922); Weil (1929)).
The set of integral points of X \ o is finite (Siegel (1929))

In higher genus, the following result had been conjectured by Mordell.

Theorem 2.2. — Let X be an algebraic curve of genus g > 2 defined over a number field. Then
the set of rational points of X is finite (Faltings (1983)).

Vojta (1991) gave another proof which was subsequently simplified by Bombieri (1990). That
proof inspired Faltings (1991) to give the proof of a more general conjecture of Serge Lang.
2 2
p with equation x + y = 1, the indefinite integral of the differential form
On the circle
dy/x = dy/ 1 − y 2 leads to trigonometrical functions and their functional equations. The
XIX th -century geometers understood how this phenomenon generalizes in higher genus. First
of all, on a compact Riemann surface X of genus g , there are exactly g independent differential
RP
forms, say ω1 , . . . , ωg . Fixing a base point o ∈ X , they lead to indefinite integrals, P 7→ o ω j ,
which however are only well-defined up to the integrals over closed loops. R This furnishes a map
from X to the quotient of the space Cg by a lattice Λ, consisting of all ( γ ω1 , . . . , γ ωg ), where γ
R

ranges over the first homology group of X . This variety J X = Cg /Λ is called the Jacobian variety
of X , and the map X → J X is an embedding (if g > 2), an isomorphism if g = 1.
Weil gave an algebraic description of this Jacobian variety. When X is defined over a number
field, so is J X , hence one can talk about the rational points of J X . The following result of Weil
applies in fact to a broader class of varieties, called abelian varieties: these are projective vari-
eties endowed with a structure of group whose law is given by regular algebraic functions. In
dimension 1, we find elliptic curves again; in higher dimension, we have Jacobian varieties but
they don’t exhaust the multiplicity of abelian varieties.

Theorem 2.3 (Mordell-Weil Theorem). — Let A be an abelian variety defined over a number
field. Then its set of rational points is an Abelian group of finite type (Weil (1929)).
4 ANTOINE CHAMBERT-LOIR

This is a generalization of the previous theorem for elliptic curves. Many questions remain
on that topic, the most obvious being the conjecture of Birch and Swinnerton-Dyer, which gives
an analytic formula for the rank of this Abelian group. Except in some important cases (elliptic
curves over Q,... according to the fundamental results of Wiles, Taylor and others) this formula
expresses the rank as the order of the zero of an analytic function at s = 1, while this function is
defined by a Dirichlet series which is only known to converge for ℜ(s) > 3/2!
Faltings’s generalization of Mordell’s conjecture alluded to above is the following:
Theorem 2.4. — Let X be a subvariety of an abelian variety defined over a number field. The
set of translates of abelian subvarieties of positive dimension contained in X has finitely many
maximal elements. Let X ∗ be X deprived of these translates; unless X is itself a translate of
an abelian variety, X ∗ is non-empty. Then, X ∗ has only finitely many rational points (Faltings
(1991)).
If X is a curve of genus g > 2 embedded in its Jacobian variety J X , one has X ∗ = X , so that
that X has only finitely many rational points. On the opposite, the hypothesis of the Theorem
is necessary: if B is a subabelian variety of positive dimension of A, then up to enlarging the
ground field, B has potentially infinitely many rational points.
In fact, the more general “Mordell–Lang conjecture” is true:
Theorem 2.5. — Let X be a subvariety of an abelian variety A. Let X ∗ be X deprived of these
translates; unless X is itself a translate of an abelian variety, X ∗ is non-empty. Any subgroup Γ
of A which is finitely generated intersects X ∗ in a finite set.
(This follows from an extension of Faltings’s theorem where one allows finitely generated,
possibly nonalgebraic, extensions of Q.)

3. Geometrizing the infinite descent: heights

All proofs of the theorems quoted above make a crucial use of the notion of height.
The height of a rational number α = m/n is defined as h(α) = log max(|m| , |n|), provided the
fraction m/n is in lowest terms. There are finitely many rational numbers of height bounded
from above by any prescribed constant.
Let us consider points of the projective space Pd of dimension d ; such a point P has ho-
mogeneous coordinates (x 0 : . . . : x d ), not all zero and well defined up to multiplication by a
common scalar. If F is a field (containing Q), one says that P is F -rational if it admits a system
of homogeneous coordinates in F ; equivalently, if the ratii x j /x i (for x i 6= 0) all belong to F . One
writes Pd (F ) for the set of F -rational points of Pd ; one writes also Q(P ) for the field generated
by the ratioi x j /x i over Q and call it the field of definition of P .
To begin with, let P ∈ Pd (Q); let us consider a system of homogeneous coordinates (x 0 : . . . :
x d ) for P , consisting of rational numbers. We can first chase the denominators and assume that
all the x j are integers; then we can divide by their gcd and assume moreover that these integers
are coprime. Only one more possibility is left, namely multiplying the system by −1. In any
case the real number
h(P ) = log max(|x 0 | , . . . , |x d |), (x 0 , . . . , x d ) coprime integers
is well-defined; we call it the height of the point P . Again, there are finitely many Q-rational
points of Pd whose height are bounded from above by any prescribed constant.
The definition can be generalised to all points in Pd (Q), where Q is the algebraic closure of Q
in C. There are many ways (all equivalent) to define h(P ); for example, one may choose a system
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 5

of homogeneous coordinates (x 0 : . . . : x d ) consisting of algebraic integers which generate, as an

ideal, the ring of all algebraic integers. Then,
1 X X
h(P ) = log max(|σ(x 0 )| , . . . , |σ(x d )|), if x j Z = Z,
[Q(x 0 , . . . , x d ) : Q] σ:Q(P ),→C
where the sum runs over the field embeddings σ of Q(P ) into the field C of complex numbers.
When d = 1, P = (1 : α), then h(P ) is related to the Mahler measure M(Πα ) of the minimal
polynomial Πα of α:
Z 1
1 ¯ ¯
h(P ) = M(Πα ) = log(¯Πα (e 2i πt )¯) dt .
¯ ¯
[Q(α) : Q)] 0
One of the geometric contents of this height function lies in its functorial properties:

Proposition 3.1. — Let X be a closed subvariety of a projective space Pn , defined over the field
of algebraic number and let f : X → Pm be a morphism from X to Pm given by forms of degree d
which have no common zero on X . Then there exists a real number c such that, for any point
P ∈ X (Q),
d h(P ) − c 6 h( f (P )) 6 d h(P ) + c.

(The upper bound is relatively obvious, the lower bound depends on Hilbert’s Nullstellen-
satz.)
This has application to abelian varieties: since the addition law can be written with quadratic
equations, one can prove that for any abelian variety X , there exist a quadratic form q and a
linear form ` on the group X (Q), and a real number c, such that for any point P ∈ X (Q),
q(P ) + `(P ) − c 6 h(P ) 6 q(P ) + `(P ) + c.
For “symmetric embeddings”, the linear form ` vanishes and the quadratic form q is called the
Néron-Tate height of the point P . It has been constructed by Néron (1965) via a difficult and
profound analysis; a straightforward proof has soonafter been given by John Tate—we shall
return on this.
Moreover, the height satisfies an easy, but crucial, boundedness property:

Theorem 3.2. — For any real number B and any integer d , there are only finitely many points
P ∈ Pn (Q) such that [Q(P ) : Q] 6 d and h(P ) 6 B (Northcott (1950)).

If X is an abelian variety, one can use this boundedness property to deduce the corollary,
which characterizes torsion points in X (Q) in terms of their heights:

Corollary 3.3. — Let X be an abelian variety defined over the field of algebraic numbers. A point
P ∈ X (Q) is a torsion point if and only if its Néron-Tate height vanishes.

Let us say explan how heights are used to prove the Mordell-Weil Theorem (Theorem 2.3).
The points P in A(Q) such that 2P = 0 form a finite subgroup A 2 . One may assume that F is
large enough so that all of these points are defined over F .
First of all, one proves that the group A(F )/2A(F ) is finite. This is done nowadays by ob-
serving that for any point P ∈ A(F ), the points Q ∈ A(Q) such that 2Q = P are defined over
an extension of F of degree 6 4g which is unramified outside a fixed set of places of F . By
an important theorem of Hermite, the set of such fields is finite, so that there is a finite Ga-
lois extension F 0 of F , independent of P , such that Q ∈ A(F 0 ). Now, given P ∈ A(F ), define
c(P ) to be the map from Gal(F 0 /F ) to A(F 0 ) given by c(P )σ = σ(Q) − Q, where Q ∈ A(F 0 ) is
any chosen point such that 2Q = P (this does not depend on the choice of Q). Observe that
6 ANTOINE CHAMBERT-LOIR

2c(P )σ = σ(2Q) = 2Q = σ(P ) − P = 0, so that the image of c(P ) lies within the finite set of
points of A(Q) whose order divides 2. Moreover, if P ∈ 2A(F ), then c(P ) = 0 (choose Q such that
P = 2Q); conversely, if c(P ) = 0, then the chosen point Q satisfies σ(Q) = Q for all σ ∈ Gal(F 0 /F ),
hence Q ∈ A(F ). The map c is linear in P ; it induces an injective morphism of Abelian groups
from A(F )/2A(F ) to the finite group Hom(Gal(F 0 /F ), A 2 ). So A(F )/2A(F ) is finite.
Then, one uses heights and descent. Take a finite system of representatives (P 1 , . . . , P m ) for
P
A(F )/2A(F ) and consider Q ∈ A(F ). One may write it Q = a i P i + 2Q 1 , with a i ∈ {0, 1} and
Q 1 ∈ A(F ). Using the fact that the Néron-Tate height ĥ is a quadratic form, one gets
1 X 1 X 1 X 1 1 X 1
ĥ(Q 1 ) = ĥ(Q − a i P i ) 6 ĥ(Q − a i P i )+ ĥ(Q + a i P i ) = ĥ(Q)+ ĥ( a i P i ) 6 ĥ(Q)+c,
4 4 4 2 2 2
P
where c is the maximum of all heights ĥ( a i P i ), when (a 1 , . . . , a m ) varies among {0, 1}. if ĥ(Q) >
c, then one obtains ĥ(Q 1 ) < ĥ(Q).
Go on from Q 1 with the same process, defining Q 2 ,..., as long as the height decreases.
The sequence obtained must stop eventually, because of Northcott’s finiteness theorem, and
reach a point Q n such that ĥ(Q n ) 6 c, and the point Q can be written as a linear combination
of P 1 , . . . , P m and Q n . By Northcott’s theorem again, there are finitely many points in A(F )
whose heights are less than c, so that Q n belongs to a fixed finite set. In other words, A(F ) is
generated by the union of the set {P 1 , . . . , P m } and of the set of points with heights 6 c.

4. The conjectures of Manin–Mumford of Bogomolov

Motivated by Mordell’s conjecture and its generalization to subvarieties of abelian varieties,
Manin and Mumford conjectured the following statement, now a theorem of Raynaud.
Theorem 4.1 (Raynaud (1983a,b)). — Let X be a subvariety of an Abelian variety A. Let X ◦
be the complement in X of all translates of positive dimensional abelian subvarieties of A by a
torsion point which are contained in X . Then, X ◦ is an open subset of X for the Zariski topology
and contains only finitely many torsion points of A.
That X ◦ is open means that there is a finite set Y1 , . . . , Ym of translates of positive dimensional
abelian subvarieties of A by torsion points which are contained in X and whose union contains
any such variety. If X itself is not a translate of an abelian subvariety of A by a torsion point,
then X ◦ is non-empty.
Here is an alternate formulation: if X contains a dense set of torsion points, then X is the
translate of an abelian subvariety by a torsion point.
Recall that torsion points are points of Néron-Tate height 0. Still motivated by Mordell’s con-
jecture, Bogomolov put forward the following statement, now a theorem of Ullmo and Zhang:
Theorem 4.2 (Ullmo (1998); Zhang (1998)). — Let X be a subvariety of an Abelian variety A
over a number field. Let X ◦ be X deprived of all translates of positive dimensional abelian sub-
varieties of A by a torsion point which are contained in X . Then there exists a positive real num-
ber δ X such that for any point P ∈ X ◦ (Q) which is not a torsion point, ĥ(P ) > δ X .
(In fact, Ullmo and Zhang’s proof provides a new proof of Raynaud’s theorem, which is not
apparent in the statement I just gave.)
The proof of this theorem relies on an equidistribution property. Let F be a number field
over which A is defined. Let P be any point P ∈ A(Q). Since its field of definition Q(P ) is a
number field, the point P has finitely many conjugates; equivalently, the orbit of P under the
action of Gal(F /F ) is finite; let P 1 , . . . , P m be those conjugates (with m = [F (P ) : F ]) and define
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 7

the probability measure δ(P ) as the discrete measure on A(C) which gives every point P j the
mass 1/m.
For any subvariety X of A, the space X (C) carries a natural positive probability measure µ X .
When X = A, µ X is the normalized Haar measure on A(C). The measures µ X are defined by
complex geometry: if A(C) = Cg /Λ (where Λ is some lattice), let ω be a Riemann form of Λ;
this is an Hermitian form on Cg (associated to some polarization) which we view as differential
form
ω = ωi , j dz i ∧ dz j
X

on A(C). Then, if d = dim(X ), the restriction of ωd to the smooth locus of X (C) is a positive
volume form of finite total mass; µ X is the unique probability measure which is proportional
to ω2 | X .

Theorem 4.3 (Szpiro et al (1997)). — Let X be a subvariety of A. Let (P j ) be a sequence of points

in X (Q) which satisfies the following properties:
(1) when j → ∞, ĥ(P j ) → 0;
(2) for any subvariety Y ( X , the set of indices j such that P j ∈ Y is finite.
Then, the measures µ(P j ) converge to µ X .

Given the equidistribution theorem, the proofs of Ullmo and Zhang are truly marvelous.
Let’s assume, for simplicity, that X is a curve of genus g in its Jacobian embedding and that g >
2, so that X ◦ = X . From the sequence (P j ), Ullmo defines a sequence (Q k ) of points in X g (Q)
which satisfy the hypotheses of the equidistribution theorem, X g being seen as a subvariety
of A g . (Each point Q k is of the form (P j 1 , . . . , P j g ).) Therefore, the measures µ(Q j ) converge to
g
the measure µ X on X g (C).
Moreover, Ullmo considers the addition map σ : A g → A. It is well known that the restric-
tion of the addition map to X g is generically finite-to-one and that σ(X g ) = A. Moreover, the
quadratic property of the Néron-Tate height imply that ĥ(σ(Q j )) converge to 0 when k → ∞.
Consequently, the sequence (σ(Q k )) also satisfies the equidistribution theorem and the mea-
sures µ(σ(Q k )) converge to µσ(X g ) = µ A .
Now comes the contradiction: for any point Q ∈ X g (C), the measure µ(σ(Q)) is equal to the
g
measure σ∗ µ(Q) obtained by pushing forward the measure µ(Q) by σ. By continuity, σ∗ (µ X ) =
µ A . Looking at this equality of measures at points where σ is an étale map, we deduce an equal-
g
ity of differential forms σ∗ µ A = g !µ X , where g ! is the degree of σ. But this equality is absurd.
g
Indeed, µ X is a volume form which is positive everywhere on X (C), and so is µ X . However,
σ∗ µ A vanishes where σ is not étale, for example along the diagonal of X g where its tangent
map has rank 1.

5. Dynamical systems
We have seen that the self-map of an abelian variety given by multiplication by 2 has been
playing an important rôle for its arithmetic. This brought J. Silverman to suggest studying the
arithmetic properties of algebraic varieties endowed with self-maps. This theory is now a full
mathematical subject, whose basics (and beyond) are exposed in the book (Silverman, 2007).
In fact, the question had been posed as early as 1950, for this was the main topic of the
mentioned paper of Northcott!
So let X be a projective variety and let f : X → X be a self-map, given in some projective em-
bedding by polynomials of degree d , without common zeroes on X . Anyway, we now view
8 ANTOINE CHAMBERT-LOIR

our (X , f ) as a dynamical system. Given a point P on X , one can look at its forward-orbit
(P, f (P ), f 2 (P ), . . . ). We can also consider backward-orbits, namely sequences (. . . , P 1 , P 0 ) of
points in X such that f (P j ) = P j −1 for any j > 1. In this setting, one says that a point P is
periodic if there exists a positive integer n such that f n (P ) = P ; and that it is preperiodic if there
exist two integers m and n, with n > m, such that f n (P ) = f m (P ) — this is equivalent to asking
that the forward-orbit be finite. The integer d is called the dynamical degree of (X , f ).
For example, one could take for (X , f ) an abelian variety and its multiplication-by-2 map. In
that case, I let as an exercise to prove that preperiodic points are precisely the torsion points!
Another example is given by taking X = P1 (the projective line) and for f any rational function
f ∈ Q(T ). It is useful to develop this theory in a more general setting, to include into the picture
some automorphisms of certain K3-surfaces, as in (Silverman, 1991), but this would bring us
too far.
Let us assume that X and f are defined over the field of algebraic numbers. Then we can
look at heights and we see that there exists a real number c such that
d h(P ) − c 6 h( f (P )) 6 d h(P ) + c
for any point P ∈ X (Q). From that property, Northcott derived the following consequence:

Theorem 5.1 (Northcott (1950)). — Assume that d > 2. Then, for any integer e, there are only
finitely many points P ∈ X (Q) which are preperiodic for f and are defined over a field of degree
less than e.

Before we explain the proof, it is interesting to develop the heights-argument further.

Proposition 5.2 (Call & Silverman (1993)). — Let us assume that X and f are defined over the
field of algebraic numbers. There exists a unique function ĥ on X (Q) satisfying:
(1) the function ĥ − h is uniformly bounded on X (Q);
(2) for any P ∈ X (Q), ĥ( f (P )) = d ĥ(P ).

This function ĥ is called the canonical height.

The proof of the proposition follows closely Tate’s proof of the existence of the Néron-Tate
height: for any P ∈ X (Q), prove that when n → ∞, h( f n (P ))/d n converges to a limit, call it
ĥ(P ), and show that the function fˆ so-defined is the unique function which satisfies the two
requirements of the proposition.
Now, it is a simple matter to prove that the canonical height is nonnegative (just use the
limit formula, using the fact that by definition the height is bounded from below), and that
preperiodic points are exactly points of canonical height zero. Indeed, if f n (P ) = f m (P ) with
n > m, then
d n ĥ(P ) = ĥ( f n (P )) = ĥ( f m (P )) = d m ĥ(P ),
so ĥ(P ) = 0 since d n 6= d m (we use d 6= 1). Conversely, if ĥ(P ) = 0, then all points in the forward-
orbit P, f (P ), . . . have canonical height zero, so their usual heights are bounded. Since they are
all defined over a common number field, the finiteness theorem implies that this forward-orbit
is finite.
Theorem 5.1 also follows from the finiteness theorem, applied to the set of preperiodic
points.

But apart from that, many questions and conjectures remain.

What is the analogue of Manin-Mumford conjecture? Namely, let X ◦ be X deprived of all
subvarieties of positive dimension Y which are preperiodic (that is, for which there exists m
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 9

and n such that n > m and f n (Y ) = f m (Y )). Is it true that X ◦ is Zariski open in X and contains
only finitely many preperiodic points?
This would be an analogue of Manin-Mumford’s conjecture because in the case of Abelian
varieties, the preperiodic subvarieties are exactly the translates of abelian subvarieties by tor-
sion points.
Although Zhang (1995) had conjectured that the answer the preceding question is positive,
Ghioca and Tucker discovered a simple counterexample in 2009. It is given as follows. Let
E = C/Z[i ] be “the” elliptic curve with complex multiplication by Z[i ] and let X = E 2 , with the
self map given by f (x, y) = (5x, (3 + 4i )y). It has dynamical degree 52 = 32 + 42 .
One can check that preperiodic points of this dynamical system are pairs (x, y) of torsion
points.
The tangent map to f has a diagonal matrix, with eigenvalues 5 and 3 + 4i whose quotient is
not a root of unity. Except horizontal or vertical lines, no line is invariant under any non-trivial
power of f . This implies that the preperiodic subvarieties are the vertical or horizontal curves,
of the form {x} × E and E × {y}, where x and y are preperiodic points. However, their union is
not a Zariski closed subset of E × E .
Alternatively, the diagonal ∆ in E × E carries infinitely many preperiodic points (so a dense
set of this since it has dimension 1) but is not itself preperiodic.
Recent work of Ghioca, Tucker and Zhang aims proposed a (possibly correct) version of a
dynamical Manin-Mumford conjecture.

Anyway, Xiniyi Yuan has recently shown an analogue of the equidistribution theorem. To
state it, I need to recall one fact from complex dynamical systems.

Theorem 5.3 (Yuan (2008)). — Let (X , f ) be an algebraic dynamical system of dynamical de-
gree > 2, defined over the field of algebraic numbers, and let ĥ be the canonical height. Let (P j )
be a sequence of points in X (Q) satisfying:
(1) when j → ∞, ĥ(P j ) → 0;
(2) for any subvariety Y ( X , the set of indices j such that P j ∈ Y is finite.
Then, the measures δ(P j ) converge to the equilibrium probability measure µ X , f on X (C).

This equilibrium measure has been introduced by mathematicians in complex dynamics

(Brolin, Lyubich, Hubbard-Papadopol, Dinh-Sibony) and can be defined as follows: take
any smooth volume form ν on X (C) and consider the sequence (νn ) of forms given by
νn = d −n dim(X ) ( f n )∗ ν. Then, νn → µ X , f when n → ∞.
This measure is very complicated in general. In fact, generalizing earlier work of Berteloot,
Dupont, and Loeb, Cantat (2008) showed that µ X , f is orthogonal to the Lebesgue measure,
unless (X , f ) is “built” from abelian varieties.
Backward orbits furnish a natural way to define sequences of points (P j ) as in the Theorem.
Indeed, one has ĥ(P j ) = d − j ĥ(P ). So Yuan’s theorem implies that, provided no subsequence
of (P j ) is contained in some strict subvariety of X , the measures δ(P j ) converge to µ X .
However, nobody found yet how to use this equidistribution theorem to derive geometric
results in the spirit of the Manin-Mumford conjecture.

There are other questions worth to be described, like the dynamical Mordell-Lang conjecture
(for which many interesting papers were written in the recent years).
10 ANTOINE CHAMBERT-LOIR

In recent years, there has been much activity about refinements of the Mordell-Lang conjec-
ture. The keywords are “anomalous intersections” and new developments concern dynamical
systems. (See Chambert-Loir (2011) for a survey on that topic.)
But also about the generalization to abelian varieties and dynamical systems of the theorem
of Merel (1996) according to which, for any elliptic curve A defined over a number field, the
cardinality of the torsion subgroup of A(F ) is bounded in terms solely of the degree of F .
However, it is time to stop. I refer the interested reader to a long survey of mine on this theme,
namely interaction of Diophantine geometry, heights and dynamical systems, see Chambert-
Loir (2010).

References
E. B OMBIERI (1990), “The Mordell conjecture revisited”. Ann. Scuola Norm. Sup. Pisa Cl. Sci.
(4), 17 (4), p. 615–640.
G. C ALL & J. S ILVERMAN (1993), “Canonical heights on varieties with morphisms”. Compositio
Math., 89, p. 163–205.
S. C ANTAT (2008), “Caractérisation des exemples de Lattès et de Kummer”. Compos. Math.,
144 (5), p. 1235–1270.
A. C HAMBERT-L OIR (2010), Théorèmes d’équidistribution pour les systèmes dynamiques
d’origine arithmétique, Panoramas et synthèses 30, p. 97–189. ArXiv:0812.0944.
A. C HAMBERT-L OIR (2011), “Relations de dépendance et intersections exceptionnelles”.
Astérisque. Séminaire Bourbaki, Vol. 2010/2011, Exposé 1032, to appear. ArXiv:1011.4738.
G. FALTINGS (1983), “Endlichkeitsätze für abelsche Varietäten über Zahlkörpern”. Invent.
Math., 73 (3), p. 349–366.
G. FALTINGS (1991), “Diophantine approximation on abelian varieties”. Ann. of Math., 133, p.
549–576.
L. Y. L AM & T. S. A NG (2004), Fleeting footsteps. Tracing the conception of arithmetic and algebra
in ancient China, World Scientific Publishing Co. Inc., River Edge, NJ, revised edition. With
a foreword by Joseph W. Dauben.
L. M EREL (1996), “Bornes pour la torsion des courbes elliptiques sur les corps de nombres”.
Invent. Math., 124 (1-3), p. 437–449.
L. J. M ORDELL (1922), “On the rational solutions of the indeterminate equations of the third
and fourth degrees.” Cambr. Phil. Soc. Proc., 21, p. 179–192.
A. N ÉRON (1965), “Quasi-fonctions et hauteurs sur les variétés abéliennes”. Ann. of Math. (2),
82, p. 249–331.
D. G. N ORTHCOTT (1950), “Periodic points on an algebraic variety”. Ann. of Math., 51, p. 167–
177.
M. R AYNAUD (1983a), “Courbes sur une variété abélienne et points de torsion”. Invent. Math.,
71 (1), p. 207–233.
M. R AYNAUD (1983b), “Sous-variétés d’une variété abélienne et points de torsion”. Arithmetic
and Geometry. Papers dedicated to I.R. Shafarevich, edited by M. A RTIN & J. TATE, Progr.
Math. 35, p. 327–352, Birkhäuser.
C. L. S IEGEL (1929), “Über einige Anwendungen diophantischer Approximationen.” Abh.
Preuss. Akad. Wiss. Phys.-Math., 1, p. 209–266.
J. H. S ILVERMAN (1991), “Rational points on K 3 surfaces: a new canonical height”. Invent.
Math., 105 (2), p. 347–373.
J. H. S ILVERMAN (2007), The arithmetic of dynamical systems, Graduate Texts in Mathemat-
ics 241, Springer, New York.
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 11

L. S ZPIRO, E. U LLMO & S.-W. Z HANG (1997), “Équidistribution des petits points”. Invent. Math.,
127, p. 337–348.
E. U LLMO (1998), “Positivité et discrétion des points algébriques des courbes”. Ann. of Math.,
147 (1), p. 167–179.
P. V OJTA (1991), “Siegel’s theorem in the compact case”. Ann. of Math. (2), 133 (3), p. 509–548.
A. W EIL (1929), “L’arithmétique sur les courbes algébriques”. Acta Math., 52 (1), p. 281–315.
X. Y UAN (2008), “Big line bundles on arithmetic varieties”. Invent. Math., 173, p. 603–649.
arXiv:math.NT/0612424.
S.-W. Z HANG (1995), “Positive line bundles on arithmetic varieties”. J. Amer. Math. Soc., 8, p.
187–221.
S.-W. Z HANG (1998), “Equidistribution of small points on abelian varieties”. Ann. of Math.,
147 (1), p. 159–165.
12 ANTOINE CHAMBERT-LOIR

F IGURE 1. The Sūnzı̌ Suàngjı̄ng (孫子算經, “The mathematical classic of Sūnzı̌”). —

A print page from a Qing dynasty edition, reproduced from Wikipedia.
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 13

F IGURE 2. Mathematical Treatise in Nine Sections (數書九章, Shùshū Jiǔzhāng) from

the 1847 Sìkù Quánshū 四庫全書. Reproduced from Wikipedia
14 ANTOINE CHAMBERT-LOIR

F IGURE 3. Archimedes’s Cattle Problem, Archimedis Opera omnia, cum commentariis

Eutocii. Edited by J. L. Heiberg, B. G. Teubner, Leibzig, Volume 2 (1881), pages 448–450
 DIOPHANTINE GEOMETRY AND DYNAMICAL SYSTEMS 15

F IGURE 4. Diophantus’s Arithmetica with the comments of Fermat, stating his “Last
Theorem” (1670 edition; reproduced from Wikipedia)