Number Theory III | Field (Mathematics) | Group (Mathematics)


i 1,;



,’ I”

Serge Lang (Ed.)

:, 91 s -8%-J .*’

Number Theory IIf& ”
Diophantine Geometry

Springer-Verlag Berlin Heidelberg New York London Paris Tokyo Hong Kong Barcelona

Encyclopaedia of Mathematical Sciences
Volume 60


R.V. Gamkrelidze


In 1988 Shafarevich asked me to write a volume for the Encyclopaedia of Mathematical Sciences on Diophantine Geometry. I said yes, and here is the volume. By definition, diophantine problems concern the solutions of equations in integers, or rational numbers, or various generalizations, such as finitely generated rings over Z or finitely generated fields over Q. The word Geometry is tacked on to suggest geometric methods. This means that the present volume is not elementary. For a survey of some basic problems with a much more elementary approach, see [La 90~1. The field of diophantine geometry is now moving quite rapidly. Outstanding conjectures ranging from decades back are being proved. I have tried to give the book some sort of coherence and permanence by emphasizing structural conjectures as much as results, so that one has a clear picture of the field. On the whole, I omit proofs, according to the boundary conditions of the encyclopedia. On some occasions I do give some ideas for the proofs when these are especially important. In any case, a lengthy bibliography refers to papers and books where proofs may be found. I have also followed Shafarevich’s suggestion to give examples, and I have especially chosen these examples which show how some classical problems do or do not get solved by contemporary insights. Fermat’s last theorem occupies an intermediate position. Although it is not proved, it is not an isolated problem any more. It fits in two main approaches to certain diophantine questions, which will be found in Chapter II from the point of view of diophantine inequalities, and Chapter V from the point of view of modular curves and the Taniyama-Shimura conjecture. Some people might even see a race between the two approaches: which one will prove Fermat first? It




is actually conceivable that diophantine inequalities might prove the Taniyama-Shimura conjecture, which would give a high to everybody. There are also two approaches to Mordell’s conjecture that a curve of genus 2 2 over the rationals (or over a number field) has only a finite number of rational points: via l-adic representations in Chapter IV, and via diophantine approximations in Chapter IX. But in this case, Mordell’s conjecture is now Faltings’ theorem. Parts of the subject are more accessible than others because they require less knowledge for their understanding. To increase accessibility of some parts, I have reproduced some definitions from basic algebraic geometry. This is especially true of the first chapter, dealing with qualitative questions. If substantially more knowledge was required for some results, then I did not try to reproduce such definitions, but I just used whatever language was necessary. Obviously decisions as to where to stop in the backward tree of definitions depend on personal judgments, influenced by several people who have commented on the manuscript before publication. The question also arose where to stop in the direction of diophantine approximations. I decided not to include results of the last few years centering around the explicit Hilbert Nullstellensatz, notably by Brownawell, and related bounds for the degrees of polynomials vanishing on certain subsets of group varieties, as developed by those who needed such estimates in the theory of transcendental numbers. My not including these results does not imply that I regard them as less important than some results I have included. It simply means that at the moment, I feel they would fit more appropriately in a volume devoted to diophantine approximations or computational algebraic geometry. I have included several connections of diophantine geometry with other parts of mathematics, such as PDE and Laplacians, complex analysis, and differential geometry. A grand unification is going on, with multiple connections between these fields.
New Haven Summer 1990 ticknowledgment

Serge Lang

I want to thank the numerous people who have made suggestions and corrections when I circulated the manuscript in draft, especially Chai, Coleman, Colliot-Thblbne, Gross, Parshin and Vojta. I also thank Chai and Colliot-Th&ne for their help with the proofreading. S.L.


Preface Notation CHAPTER Some

. .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I Qualitative Diophantine Statements . . .. . ...



. . . . . .

2 9 15 25 30 35 40

$1. $2. $3. $4. $5. $6. $7.

Basic Geometric Notions . .......... .. . .... The Canonical Class and the Genus ........ ... The Special Set ... .... .. . ...... .. ... Abelian Varieties .... . ..... . .... Algebraic Equivalence and the N&on-Severi Group Subvarieties of Abelian and Semiabelian Varieties . Hilbert Irreducibility . . . . . . . . .. . . . . . . . . . .
II and Rational Points .... ... . ............

.. . . ..


. ..... ...

43 43 51 58 61

$1. $2. $3. $4.

The Height The Height The Height Bound for
III Varieties

for Rational Numbers and Rational Functions in Finite Extensions ....................... on Varieties and Divisor Classes ............ the Height of Algebraic Points ..............

. ..




. .. . . . . . . . . . . .

. ...


. .. . . . ..

68 68 71 76 82 85

$0. $1. $2. $3. $4.

Basic Facts About Algebraic Families and Ntron Models . The Height as a Quadratic Function ...... . ... .. Algebraic Families of Heights ... .... .. . .. . . . . . . .. Torsion Points and the I-Adic Representations . . . . . . . . . .. Principal Homogeneous Spaces and Infinite Descents . . .



$5. The Birch-Swinnerton-Dyer Conjecture ......... $6. The Case of Elliptic Curves Over Q ............
CHAPTER Faltings’ IV Finiteness Theorems on Abelian Varieties and Curves

. .

91 96

$1. Torelli’s Theorem .............. ............ .. . . . . .. . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 92. The Shafarevich Conjecture . . . . . . . . . . . . . .. $3. The I-Adic Representations and Semisimplicity $4. The Finiteness of Certain l-Adic Representations. Finiteness I Implies Finiteness II . . . . . . . . . .. ....... $5. The Faltings Height and Isogenies: Finiteness I . . . . . . . . . . . . . . . .. ... $6. The Masser-Wustholz Approach to Finiteness I
CHAPTER Modular V Curves Over Q ...........................

. . . . . .

101 102 103 107 112 115 121

01. $2. $3. $4. 55.

............................... Basic Definitions Mazur’s Theorems .............................. Modular Elliptic Curves and Fermat’s Last Theorem ............... Application to Pythagorean Triples Modular Elliptic Curves of Rank 1 ...............
VI Case of Mordell’s Conjecture . ....

.. .. .. .. .. . .

.. . .

123 124 127 130 135 137

.. .. .



$0. $1. 52. 53. $4. $5.

Basic Geometric Facts . ........ . ... The Function Field Case and Its Canonical Sheaf Grauert’s Construction and Vojta’s Inequality ... Parshin’s Method with (O&r) ...... .. .... ... . Manin’s Method with Connections Characteristic p and Voloch’s Theorem . . . . . .
VII Theory ...............................

.. .. .. .. .. .. ..

.. . .. .. .. ..


143 143 145 147 149 153 161


CHAPTER Arakelov

$1. Admissible Metrics Over C ................... $2. Arakelov Intersections ....................... $3. Higher Dimensional Arakelov Theory ..........
CHAPTER Diophantine VIII Problems and Complex Geometry ...

.. ..


. .

163 164 166 171

. . .

176 177 184 187 189 192

$1. $2. $3. $4. $5.

...... Definitions of Hyperbolicity Chern Form and Curvature ...... ..... Parshin’s Hyperbolic Method Hyperbolic Imbeddings and Noguchi’s Nevanlinna Theory .... .. ...

.. ..... .. . ... ... . ... Theorems ........ ..

.. ..

. .

.. . ... ..... .. . . $2. . .. ... ... .... ...... ... . . $4. ...... .. $1... . . .. .. 205 207 213 216 222 225 228 233 CHAPTER Existence . ..... ... .... .. . .. ... .. $3... . ... .. The Brauer Group of a Variety and Manin’s Obstruction Local Specialization Principle ... .. . . ............. ... Weil Functions and Heights ........... .... .. . .. .. . ..... The Theorems of Roth and Schmidt Integral Points . . ... . ... .. . ... ....... From Thue-Siegel to Vojta and Faltings Diophantine Approximation on Toruses . .. ....... . . . ... Anti-Canonical Varieties and Rational Points .. .. . ... . ..... .. Connection with Hyperbolicity . . ... .. ........ $7..... $3. . . . ..... .... . . X of (Many) Rational Points .. .. .. ..... .. .. ... . ...... . ... . .... ..... ... .. 244 245 250 258 259 263 283 Bibliography Index . .. .... $5. . . . . $4.... . . . ..... Vojta’s Conjectures ... 92........ ... .. ... . . . . . . ... .. . . ........ . ... .... ... . .... .... .CONTENTS xi CHAPTER Wail IX Integral Points and Diophantine Approximations Functions....... .... ... ... ...... . ..... .... Forms in Many Variables ..... . .. ... .. $1.... . .. .. . . . . .. $6.. ... ... .... . . . . . .. . ....

for complex conjugates. Also the notation F” is in line with F” or F”’ for the separable closure. I hope. functorially with respect to linguistics? . << is used synonymously with the big Oh notation. A[m] is the kernel of multiplication cp when A is an abelian by an integer m. so why not use the expression in English. and whatnot. or less universal meaning. and there is no reason to lag behind in English. but again the French have been using faisceau vectoriel. etc. I list a few of these. since the bar is used for reduction mod a prime. Also I object to using the same expression vector bundle for the bundle and for its sheaf of sections. The French have been using the expression “faisceau en droites” for quite some time. replace locally free sheaf of finite rank. or the unramified closure. # denotes number of elements. and because the terminology becomes functorial with respect to the ideas. Line sheaf is what is sometimes called an invertible sheaf. both because it is shorter. Vector sheaf will. and have a more F” denotes the algebraic closure of a field F. I am fighting an uphill battle on this. Then f XX< g means f << g and g << f. A[rp] means the kernel of a homomorphism group. I am trying to replace the older notation F.Notation Some symbols will be used throughout systematically. so #(S) denotes the number of elements of a set S. g are two real functions with g positive. then f << g means that f(x) = O(g(x)). If f.

we try to characterize those situations when the set of rational points is small by algebraic geometric conditions. as well as conjectures which make the theory more coherent. admitting very simple statements. and from complex analysis or differential geometry applied to the complex points of the variety after such an imbedding.CHAPTER I Some Qualitative Diophantine Statements The basic purpose of this chapter is to list systematically fundamental theorems concerning the nature of sets of rational points. The elaborate machinery being built up strives partly to prove such results. and hence for the convenience of those readers whose background is foreign to algebraic geometry. but for which no known proofs are known today. both in the form of theorems and of conjectures. Most cases treated in this chapter are those when the set of rational points is “as small as possible”. . The qualitative statements of this first chapter will be complemented by quantitative statements in the next chapter. I have started with a section reproducing the basic definitions which we shall use. “Small” may mean finite. and many readers might want to read it first. It shows the sort of fundamental results one wishes to obtain. These conditions seem to have to do with group structures in various ways. As much as possible. We shall also try to describe conjecturally qualitative conditions under which there exist many rational points. In Chapter VIII we relate these algebraic conditions to others which arise from one imbedding of the ground field into the complex numbers. We use a fairly limited language from algebraic geometry. The first section of Chapter II is extremely elementary. it may mean thinly distributed. One of the purposes is to describe what this means. or if there is a group structure it may mean finitely generated.

) are equivalent and only if there exists c E F. c # 0 such that (Y0 9 . if where two such (n + l)-tuples (x0. . . It just happens that we need very few notions.) with xj E not all xi = 0. .T. . . .. . .z.y. reduced.z. .2 SOME QUALJTATJVE DJOPHANTJNE STATEMENTS lx §13 I.). and such that these properties are preserved under arbitrary extensions of the ground field k. .. . . . . The condition that the polynomials generate a prime ideal is to insure what is called the irreducibility of the variety.... . . . .m and zi E k’ for all i = 1. the set of zeros of I with coordinates z i . . and quasi projective.. integral. If k’ is a field containing k. .. . . and is denoted by X(k’). . E k’ is called the set of rational points of X in k’. defined below. .) and (yO.x.n. . . .gm generate a prime ideal in the ring k”[z. it is not possible to express a variety as the finite union of proper subvarieties. . But for those acquainted with the scheme foundations of algebraic geometry. Let I be an ideal in this ring.m) . Assume that g1 .. . The set of zeros of I is called an affine variety X. . A reader acquainted with these notions may skip this part.]. ... Let P” denote projective n-space. projective.) = 0 with j= 1.z.. It is equal to the set of solutions of the finite number of equations gjCzl ) . BASIC GEOMETRIC NOTIONS For the convenience of the reader we shall give definitions of the simple. Consider the polynomial ring in n variables i . . . generated by a finite number kCz of polynomials gi . . we shall limit ourselves ad hoc to the three types of varieties which we shall consider: affine. . . a variety is a scheme over a field k. By pasting together a finite number of affine varieties in a suitable way one obtains the general notion of a variety. The variety defined by the zero ideal is all of affine space A”. and so a totally uninformed reader might still benefit if provided with these basic definitions. basic notions of algebraic geometry which we need for this chapter. . . By a projective variety X over a field k we mean the set of solutions in a projective space P” of a finite number of equations fj(To. .Y”) = (c&J 3. $1. To avoid a foundational discussion here. . Under our condition. . . Thus P”(F) consists of equivalence classes of (n + I)-tuples P = (x0 ) . If F is a field. then P”(F) denotes the set of points of P” over F. .) = 0 (j= l.x. Let k be a field. z.] over the algebraic closure of k. . . . separated and of finite type. . . . .

zJ. . Consider a maximal chain of sub- . . .) = fj(l. .m) is an affine variety. J"). . .9 2. which is an open subset of X. .) such that xj # 0 is an affine open Uj. x. .f. . . . (x. zi = T/T’. . The set of points in X(k”) is called the set of algebraic points (over k).zn) = O (j = 1.. . (i = 1.CL411 BASIC GEOMETRIC NOTIONS 3 such that each fi is a homogeneous polynomial in n + 1 variables with coefficients in k. and let . . . . . . Then the polynomials gi. generate a prime ideal in k”[z. as follows. . If for instance x. . . . The projective variety X is covered by the variety X we shall always means a closed subspecified.) E X such that x0 # 0.!j) = T:/T J Thus the set of points subset of X denoted by open sets U. We can define the Zariski topology on P” by prescribing that a closed set is a finite union of varieties. . and fi. . . . . we could have picked any index instead of 0. . generate a prime ideal in the polynomial ring k”[‘I’. . .‘. . . . # 0. J. . or equivalently of some ideal. . . Zl. and the set of solutions of the equations !3jCzl 9 .n and i # j.. ..x. A closed subset is simply the set of zeros of a finite number of polynomials.. . ..U.) with xi E k’ for all i = 0. .g.n) and let gjtzl 3 . . .. . . . We denote by k(x) (the residue class field of the point) the field k(x) = k(x... .. Similarly. We say that X(k’) is the set of rational points of X in k’. then 64 = khlx. It \ consists of those points (x0. . By a subvariety of a variety unless otherwise for i = 0. x. by X(k’) we mean the set of such zeros having some projective coordinates (x0. It does not matter which such coordinate is selected. T. .]. If k’ is a field containing k. which need not be a prime ideal. . denoted by U. . By a quasi-projective variety. A Zariski open set is defined to be the complement of a closed set.x. . . . say j. say.‘. . A projective variety is covered by a finite number of affine varieties. n.) such that at least one of the projective coordinates is equal to 1. Let. ./xo)~ We shall give a more intrinsic definition of this field below. . . . .. . . we mean the open subset of a projective variety obtained by omitting a closed subset. .

. and is denoted by k(X). and defined over a field k.Y. . above are non-singular. .. in which case such conditions will be specified. . The function field is denoted by k(X). A projective variety is called non-singular if all the affine open sets U. A hypersurface is a subvariety of P” of codimension 1. .m) be a set of defining equations for Z. The function field of X over k is defined to be k(Ui) (for any i).) = 0... and its quotient field is called the function field of X over k. . defined by one equation f( T. A rational function can also be expressed as a quotient of two homoge- \ . If X is a variety defined over the complex numbers. . . . Let X be a projective variety. . where Y0 is a point and x # x+1 for all i. If k is a subfield of the complex numbers. The degree of f is called the degree of the hypersurface. then X(C) is a complex analytic space of complex analytic dimension r.z.T... Then the function fields k(U. . The ring R = k[z i.. . We say that the point P is simple if the matrix (Digi( has rank n .) be a point of Z. We have used Di for the partial derivative a/azi. and r is called the dimension of X. This ring has no divisors of zero. Let X be an affine variety. or simply the affine ring of X. .. defined by r equations fj = 0 (j = 1. with coordinates (z i. A curve is a variety of dimension 1. . . In the course of a discussion.]/I is called the affine coordinate ring of X. and are generated by the restrictions to X of the quotients TJ?. .a. c Y. . or satisfies additional conditions such as being non-singular (to be defined below). Let gj = 0 (j= 13. .). .) are all equal. An element of the function field is called a rational function on X. . A rational function on X is therefore the quotient of two polynomial functions on X. .. j such that q is not identically 0 on X). Let Z be an affine variety in affine space A”. (for all i. Let P = (a. . . U.). Y). c Y.k(U. . .z. = x.]. Suppose k algebraically closed and ai E k for all i. . such that the denominator does not vanish identically on X. . . then X is nonsingular if and only the set of complex points X(C) is a complex manifold. one may wish to assume that a curve or a surface is projective.4 varieties SOME QUALITATIVE DIOPHANTINE STATEMENTS CL 011 Y. Then all such chains have the same number of elements Y. A surface is a variety of dimension 2. A projective variety of dimension Y is sometimes called an r-fold. defined by an ideal I in k[z.. . .z. We say that Z is nonsingular if every point on Z is simple. . then we say that X is a complete intersection. .. .r. c . If X is a subvariety of P” of dimension n . where Y is the dimension of Z.

A variety is said to be normal if the local ring of every point is integrally closed. f induces an injection of function fields k(Y) 4 k(X). By a morphism f:X+Y defined over k we mean a map which is given locally in the neighborhood of each point by polynomial functions./cAfp. . We say that f is an imbedding if f induces an isomorphism of X with a subvariety of Y. A non-singular variety is normal.e. there may be a variety over a field k which is isomorphic or birationally equivalent to projective space P” over an extension of k. and gof =id.) where fi. T. In this case.CL811 BASIC GEOMETRIC NOTIONS 5 neous polynomial functions fi(T. map. defined over a field k. + x: = 0. A birational map is a rational map which has a rational inverse. and g: V + Y is a morphism which is equal to f on U n V. but not over k itself. Y be varieties. Let f: X -+ Y be a rational f is generically surjective if the image of a non-empty .. If I/ is a Zariski open subset of X.)/fi( To. We say that Zariski open subset of X under f contains a Zariski open subset of Y. If f is a birational map.. expressible as a quotient cp = f/g. . Two varieties X. Let P be a point of X. a morphism g: Y + X such that fog=id. We then have the local ring of regular functions 0. f2 have the same degree. where f. Example: the curve defined by the equation in P2 xi + x. A rational map f: X + Y is a morphism on a non-empty Zariski open subset U of X. i. . Thus we think of a rational map as being extended to a morphism on a maximal Zariski open subset of X. . defined over the field k. This local ring has a unique maximal ideal J?%!~. we specify the field over which rational maps or birational maps are defined. Let X. The residue class field at P is defined to be k(P) = O. . T. For instance. then g is uniquely determined. If needed. then f induces an isomorphism of the function fields. An isomorphism is a morphism which has an inverse. . . g are polynomial functions on X and g(P) # 0. at P. . consisting of quotients as above such that f(P) = 0. Y are said to be birationally equivalent if there exists a birational map between them. which is defined to be the set of all rational functions cp.

We consider pairs (U. A Weil divisor can therefore be written as a linear combination D = c niDi where Di is a subvariety of codimension 1. The Cartier divisor is said to be effective if for all representing pairs (U. The order function extends to a homomorphism of the group of non-zero rational functions on X into Z. In other words. To each rational function we can associate its divisor (f) = 1 orddf HW. cp) and (U. VP’). Such Cartier divisors are said to be rationally or linearly equivalent to 0.-module O. Chapter II. We then view the Cartier divisor as a hypersurface on X. .6 SOME QUALITATIVE DIOPHANTINE STATEMENTS I?.Z. if Cartier divisors are represented locally by (U. Let cp be a non-zero rational function. Let 0. A Cartier divisor is defined as follows. We say that two such pairs are equivalent. A maximal family of equivalent pairs whose open sets cover X is defined to be a Cartier divisor. 911 A variety is said to be rational (resp. and ni E. A Weil divisor is an element of the free abelian group generated by the subvarieties of codimension 1. Next we describe divisors on a variety.. then their sum is represented by VJ. The factor group of all Cartier divisors modulo the group of divisors of functions is called the Cartier divisor class group or the Picard group Pit(X). Indeed. cp).) One can also define the notion of linearly equivalent to 0 for Weil divisors. If f is a rational function on X which lies in O. cp’) respectively.(V. then we define the order off at W to be ord&f) = length of the O. Let W be a subvariety of X of codimension 1. The Cartier divisors form a group. Proposition 6. cp) consisting of a Zariski open set U and a rational function cp on X. (See [Ha 771. If all n. or on the open set U./fO. There are two kinds. f # 0. 2 0 then D is called effective. represented by the pairs (U. for every P E U n V. A pair (U. It is a basic fact that if X is non-singular then the groups of Weil divisors and Cartier divisors are isomorphic in a natural way.15. a rational image of) projective space. cp) is said to represent the divisor locally. $) if the rational function cp+V’ is a unit in the local ring 0. and write (U. be the local ring of rational functions on X which are defined at W. both VI+-’ and cp-‘$ are regular functions at all points of U n I/. Q) for all open sets U. unirational) if it is birationally equivalent to (resp. defined locally on U by the equation cp = 0. Then cp defines a Cartier divisor denoted by (cp). . 40) the rational function cp is regular at all points of U.

Equivalently. we shall usually assume that the variety is complete and non-singular in codimension 1. and similarly for the definition of an ample divisor class. every divisor D is linearly equivalent D N E. we shall now state some properties of divisor classes for the Cartier divisor class group. See Fulton’s book Intersection Theory. where E. or even very ample. We have already defined effective divisors. a divisor class c is ample if and only if there exists a positive integer q such that qc is very ample. A divisor class c is called effective if it contains an effective divisor. Analogous properties also apply to the Chow group. For simplicity. which is injective if X is normal. It is a pain to have to deal with both groups. Divisors also satisfy certain positivity properties. Proposition 1. and in Chapter II we shall see how it gives rise to positivity properties of heights. In particular. Let X be a projective variety. One reason why the Chow group is important for its own sake is that one can form similar groups with subvarieties of higher codimension. There is a natural homomorphism from Cartier divisors to Weil divisors. By the support of a Cartier divisor D we mean the set of points P . We call a divisor D ample if there exists a positive integer 4 such that qD is very ample. there exists a positive integer n such that D + nE is ample.. We shall see in Chapter VIII that this property has an equivalent formulation in terms of differential geometry. are very ample. . and the factor group is called the Chow group CH’(X). and these are interesting for their own sake. But there is a stronger property which is relevant.1.K 911 BASIC GEOMETRIC NOTIONS 7 The subgroup of the Weil divisor group consisting of the divisors of rational functions defines the group of divisors rationally equivalent to 0. Given a divisor D and an ample divisor E. A divisor D on X is called very ample if there exists a projective imbedding f:X+P” such that D is linearly equivalent to f-‘(H) for some hyperplane H of A divisor class c is called very ample if it contains a very ample divisor. We view ampleness as a property of “positivity”. We have a basic property: P”. When dealing with the Chow group. inducing a homomorphism Pit(X) -+ CH’(X). E. and an isomorphism if the variety X is non-singular.E.

Example. and suppose f(X) is not contained in the support of D..‘. A morphism f: X -+ Y induces an inverse mapping f *: Pic( Y) + Pit(X).T.. be elements of k not all 0. This inverse image defines the inverse image of the divisor class. *. of course. More generally. ./lJ + . .(T. The equation f(T) defines a hyperplane. and the complement of this hyperplane is the affine open set which we denoted by U. and thus defines our mapping f *.8 SOME QUALITATIVE DIOPHANTINE STATEMENTS CI. A(~)) Zariski open subset U of X. . The equation To = 0 defines a hyperplane in P”. + a. which is a Cartier divisor on X.T. + a... . Let D be a divisor on X. . ./ZJ. PN(k).. and we view the is defined on a non-empty .. II/). On Ui with i # 0. we could have selected any other index. tional function f(T) ~ T = aoTo + . the hyperplane is represented by the rational function To/T. Suppose D is represented by (V. for all j. . 011 such that if D is represented by (U. .(T. be the projective variables. Let (pN} be a basis of H”(X. . and let T. and for some j we have cpj(P) # 0 then (cpO(P)~ . .. Let X be a projective non-singular variety defined over an algebraically closed field k. Let X = P” be projective space.. Instead of the index 0. ‘pj E 0. The support of D is denoted by supp(D).D where E is an effective divisor. Indeed.. (cp) = E . then cp is not a unit in the local ring 0. Then (fV)> $of) re P resents the inverse image f -‘D. which space . If P E X(k) is a point such that {cpo. D).a. We let H”(X. . let a. this hyperplane = a. let D be a Cartier divisor on Y. cp) in a neighborhood of P. = 0 is represented by the ra- On Ui. &?Nm) is viewed association as a point in projective f: p H (cpom as a map. D) = k-vector space of rational functions cp E k(X) such that In other words.

and a subvariety of codimension 1 is a point. and Pi E X(ka). Hence a divisor D can be expressed as a linear combination of points D = i i=l mi(Pi) with miE Z. Curves We define a curve to be a projective variety of dimension 1. a divisor D. for each positive multiple we obtain a morphism mD of D. It is a result of Kodaira that: On a non-singular projective variety. Similarly. and then we deal with the general case. Thus the property of being pseudo ample is a property of divisor classes. Let X be a non-singular curve over k. We define the degree of the divisor D to be degD=zmi. If there exists some positive integer m such that f. is an imbedding of some non-empty Zariski open subset of X into a locally closed subset of PN@)3 then we say that D is pseudo ample. mD). More generally. is defined to be pseudo ample if and only if D. $2. using a basis for H’(X. is linear equivalent to a divisor D which is pseudo ample in the above sense.CL@I THE CANONICAL CLASS AND THE GENUS 9 Thus we obtain a morphism f: U+PN. We first deal with varieties of dimension 1. I. Then divisors can be viewed as Weil divisors. THE CANONICAL CLASS AND THE GENUS We shall discuss a divisor class which plays a particularly important role. a divisor D is pseudo ample $ and only if there exists some positive integer m such that mD-E+Z where E is ample and Z is effective. f.: a non-empty Zariski open subset of X + PNcm). .

Let 0. function y as above we can associate a Weil divisor (Y)= c ordp(yW). y E k(X).2. Let P E X(k). Since every rational differential form is of type uy dx for some rational function u. a.(y) = r. i. It is a fact that deg(y) = 0. The completion of 0. we write y dx = y(t)2 dt.(y) = ord. which is called a local parameter at P. it follows that the degrees of non-zero differential forms are all equal. Then JltP is a principal ideal. We can associate a divisor to the differential form y dx by letting (Y4 = c orMyW(P). E k. where y. We define ord. be the local ring with maximal ideal P. Hence the degree is actually a function on divisor classes. To each rational with a. Let F = k(X). Let y E k(X) be a rational function.(y dx) = order of the power series y(t) 2. can be identified with the power series ring k[[t]]. and y # 0. One possible definition of the genus of X is by the formula deg(y dx) = 2g . Let P be a point in X(k). Let x. on CH’(X). generated by one element t. # 0. Then we define ord. The function y H v. d enoted by k((t)). Every element y # 0 of k(X) has an expression y = t’u where u is a unit in OP.10 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL 621 Suppose for simplicity that k is algebraically closed. In terms of a local parameter t.(y) defines an absolute value on k(X). . In this quotient field. We may view F as a subfield of the quotient field of k[[t]]. x are expressible as power series in t. y has a power series expansion y = art’ + higher terms. A differential form y dx will be called a rational differential form.e.

Suppose that X is non-singular.(X(C). Then ~ genusofX=(n-1)Z(n-2). called the canonical class. on a non-singular only if curve X. = 0 over a field of characteristic genus is 2 2 when n 2 4. Finally. Z) is the first topological homology group. the divisors of rational differential forms constitute in CH’(X). also called a compact Riemann surface. For instance. We shall give the value only in the nonsingular case. the Fermat curve x. . the canonical class is ample if and only if g 2 2. The differential forms of first kind form a vector space over k. denoted by Q’(X). A differential form o is said to be of the first kind if ord. Z). we want to be able to compute the genus when the curve is given by an equation. a divisor class c is ample if and deg(c) > 0. This In general. Suppose k = C. which is a free abelian group over Z. over the algebraically k. Thus we may say that: The degree of the canonical class is 2g . The dimension of Cl’(X) is equal to g. It also satisfies the formula 2g = rank H. Then X(C) is a compact complex manifold of dimension 1.L-11 921 THE CANONICAL CLASS AND THE GENUS 11 a class Furthermore. + x. where H.(X(C). closed field p with p j n has genus (n . Therefore. The genus g is equal to the number of holes in the surface.l)(n .2)/2.2. The following property also characterizes the genus. The genus can also be characterized topologically over the complex numbers. + x. Let the curve X be defined by the homogeneous polynomial equation of degree n in the projective plane P’.(w) 2 0 for all P.

Then X has only a jinite number of rational points in F.12 SOME QUALITATIVE DIOPHANTINE STATEMENTS [IT VI class has For g = 1. Theorem 2. defined over k. one then obtains the following corollary. that is. A curve of genus 1 with a rational point is called an elliptic curve over k. . . t.g2x . Mordell’s conjecture made in 1922 [Mord 221 became Fakings theorem in 1983 [Fa 831. the canonical Examples. Suppose X has genus 2 2. .. whose constant field k is the set of elements in F which are \ algebraic over the prime field. and there exist algebraically independent elements t i. For g = 0. X(F) is finite. we call k the constant field. We refer to this situation as the split case of a variety over F. applica- We now have defined enough notions to pass to diophantine tions. Then X has a rational point in k if and only if X is isomorphic to a curve defined by an equation y2 = 4x3 . A function field. the canonical degree -2. Then X(F) is jinite. over a field k. Suppose F is the function field of a variety W over k. Under these circumstances. which by definition is a finite extension of Q. Corollary 2. not necessarily algebraically closed.(F) and the rational maps W + X. in F over k such that F is a finite separable extension of k(t.t. Let X be a curve dejined over a field erated over the rational numbers. If F is a finitely generated field over the prime field. Then there is a natural bijection directly from the definitions between the set of rational points X. An extension F of k is a function field over k if and only if F is finitely generated over k. Let X be a curve dejned over a number jield F. every element of F algebraic over k lies in k.. Using specialization techniques dating back to the earlier diophantine geometry. . . then F is a function field. Let X0 be a variety defined over k. Let X be a projective non-singular curve of genus 0 over a field k. which is defined as the function field of a variety.). Let X be a projective non-singular curve of genus 1 over a field k of characteristic # 2 or 3. We shall deal with the following kinds of fields: A number field. .27gj # 0. Then X is isomorphic to P’ over k if and only if X has a rational point in k. . g: . class is 0.1.g3 with g2.2. g3 E k. Such fields can be characterized as follows. days of gen- F finitely Aside from this formulation in what we may call the absolute case. .

Then the hypothesis that X is non-singular implies . If k is a finite field with q elements and X0 is defined over k. One can then apply induction to the case of dimension 1 to handle the general case. case.. 29.-. Then there exists a curve X.3. Note as in [La 60aJ that the essential difficulty occurs when F has transcendence degree 1 over k. be the local ring of P on X. Let X be a curve dejined over the function 0. be a curve in characteristic 0.. that is. and X. 021 THE CANONICAL CLASS AND THE GENUS 13 there is a relative formulation. c F. when F is a finite extension of a rational field k(t) with a variable t. with maximal ideal AP. and all but a jinite number of points in X(F) are images under this isomorphism of points in X. and [La 83a]. Let X. Suppose that X has inJinitely many rational points in X(F). variety of dimension n. For a proof see [La 60a].[I. c . 223. Remark.3 when X is isomorphic to some curve X0 over may be reformulated in the form.. The case of Theorem following geometric p. of genus 2 2. as well as several other proofs given since. then iterations of Frobenius on this point yield other points. c F. W is a divisor on X. Let W be an arbitrary variety. in what is usually called the function field analogue of Mordell’s jield F over k In [La 60a] I conjectured the following conjecture for a curve (assumed non-singular). k is the split case in which the conclusion 2. Higher dimensions Let X be a projective non-singular an algebraically closed jield k. Indeed. This formulation was proved by Manin [Man 631. such that X0 is isomorphic to X over F. of dimension 1. and of genus >= 2. and let 0. In particular. has a point in an extension F of k. Theorem of de Franchis. in Chapter VI.. We shall describe certain features of Manin’s proof. so exceptional cases have to be excluded in characteristic p. $5. of characteristic Theorem 2. there exists a tower k = F. Elementary reduction steps reduce the theorem to this case. = F such that each Fi is a function field over Fi-l. Let P E X(k). p. Then there is only a finite number of generically surjective rational maps of W onto X. defined over Let W be a subvariety of codimension 1.. defined over k. See Chapter 6.(k).

.t. then cp is an irreducible (or prime) element in OP. If W passes through P. for some rational function $. A canonical divisor is sometimes denoted by K. Let t 1. if its associated divisor is effective. where f is a homogeneous polynomial of degree d.. which is called the divisor associated with w. and for any point P on P’ the canonical class on P’ is given by Kp.such that W is defined in a Zariski open neighborhood U of P by the equation cp = 0. or by K. Let o be such a form. . In particular. .-2(P). The collection of pairs (U. then cp is a unit. . II/) where 1+9 OP. generators for the maximal ideal in the local ring at P). and again this class is called the canonical class of X. Suppose that X is a non-singular defined by the equation fG. so in k(X). Thus the canonical class is ample if and only if d 2 m + 2. Let y. xi. whose dimension is called the geometric genus. and U n W is empty. as well as its class. Let Hx be the restriction to X of a hyperplane which does not vanish identically on X.... or in other words. A a*. The regular differential forms of top degree form a vector space over k. be rational functions on X. II/) defines a Cartier divisor.(m + l))H. .. and is denoted by (0). cp) as above define the Cartier divisor associated with W. . Examples. A dt. is a unique factorization ring. The E differential form is called regular if it is regular at every point.. if we wish to emphasize the dependence on X. A differential form of top degree is called regular at P if its divisor is represented in a neighborhood of P by a pair (U. The canonical class of P” itself is given by K rrn N -(m + 1)H for any hyperplane H on P”. and is classically denoted by pn. Then the canonical class on X is given by K. . .L) hypersurface in projective = 0 space Pm. A form of type ydx.14 SOME QUALITATIVE DIOPHANTINE STATEMENTS [I> 921 that 0. All such divisors are in the same linear equivalence class. There exists an element cpE O. The collection of pairs (U.(d . be local parameters at P (that is. any two points on P’ are linearly equivalent. In a neighborhood U of P we may write co = t. A . . . is called a rational differential form of top degree on X. A dx.hdt.. . . If W does not pass through P. for m = 1. well defined up to multiplication by a unit in 6JPp.x.

resolution of singularities is known.) Finally. there exists a birational morphism f: Xl-+X such that X’ is projective and non-singular and f is an isomorphism over the Zariski open subset subset of X consisting of the simple points. $3. Instead of pseudo canonical. Let X be a variety defined over an algebraically closed field of characteristic 0. We ask under what conditions can there be infinitely many rational points of X in some such field F? One can a priori describe such a situation. A non-singular THE SPECIAL SET 15 class. First. I am trying to make the terminology functorial with respect to the ideas. In characteristic 0. but possibly singular. is called the co-canonical or anti-canonical projective variety is defined to be: canonical if the canonical class K.CL 031 The class -K. f*: H’(Y. suppose that X is a projective variety. containing F. We say that X is Mordellic if X(F) is finite for every finitely generated field F over Q. and due to Hironaka. Then X can also be defined over a finitely generated field F. We say that X is pseudo canonical if X is birationally equivalent to a projective non-singular pseudo canonical variety. It is an elementary fact of algebraic geometry that if f: X + Y is a birational map between non-singular projective varieties. but with the support of Grijtiths. is very ample pseudo canonical if K.) -+ H’(X. We shall relate these conditions with diophantine conditions in the next section. is pseudo ample anti-canonical if -K.(X) = dim H’(X. This means that given X a projective variety. I. or pseudo ample. is pseudo ample. For simplicity let us now assume that our jields have characteristic I shall give a list of conjectures stemming from [La 743 and [La 863. is ample and so on.) is an isomorphism. a variety has been called of general type. In particular. (I know I am fighting an uphill battle on this. THE SPECIAL SET 0. nK. then for every positive integer n. one defines the geometric genus p. In analogy with the case of curves. is ample very canonical if K.. if . K. It is a basic problem of algebraic geometry to determine under which conditions the canonical class is ample. nK. is pseudo ample if and only if K.). K. over the rational numbers.

or the multiplicative group G. Let X be a projective variety. Roughly speaking.2. are birationally equivalent to P’.. if and only if every rational map of a group variety into X is constant. or it may be the whole of X. Note that the maps f may be defined over finitely generated extensions. it may be part of X. we are dealing with the “geometric” situation. Note that the affine line A’. I conjectured: 3. We define a projective variety to be algebraically hyperbolic if and only if the special set is empty.4.e. then from two random rational points in some field F we can construct lots of other points by using the law of composition. i. Hence: 3. is the Zariski closure of the union of all rational maps of P’ and abelian varieties Since P’ itself is a rational we may also state: image of an abelian variety of dimension 1. i. Note that we are dealing here with curves (and later subvarieties) defined over some finitely generated extension. 3.e. The complement of the special set is Mordellic. The special set Sp(X) images of non-constant into X. i. so the presence of rational curves in X can be viewed from the point of view that these lines are rational images of group varieties.e. Let us define the algebraic special set Sp(X) to be the Zariski closure of the union of all images of nonconstant rational maps f: G + X of group varieties into X. This special set may be empty. i. The special set Sp(X) is the Zariski closure of the union of the images of all non-constant rational maps of abelian varieties into X. a curve birationally equivalent to P’. 3. We shall study their diophantine properties more closely later.e. We make this definition to fit conjec- .e. and the function field of a linear group is unirational but rational over an algebraically closed field. A more general example is given by a group variety. A group variety which is projective is called an abelian variety. the conjecture is that these are the only examples.1. If G is a group variety. then all the rational points of this curve give rise to rational points on X. subgroups of the general linear group which are Zariski closed subsets. Other examples of group varieties are given by linear group varieties. A projective variety is Mordellic if and only if the special set is empty. that is a variety which is a group such that the composition law and the inverse map are morphisms. the special set is defined geometrically. i. Let us make this conjecture more precise.3.16 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL §31 there is a rational curve in X. A general structure theorem due to Chevalley states that the only group varieties are group extensions of an abelian variety by a linear group.

e. the variety is algebraically hyperbolic.e. In light of this conjecture. See Chapter V. i.5 below.e. Examples of hyperbolic projective varieties will be given in Chapter V. We have X = Sp(X). A variety which is not special can be called general. we note that for any finitely generated field F over Q. The converse is not true. Every subvariety hyperbolic. there are no other examples of Mordellic projective varieties besides hyperbolic ones. As a result. There are the two extreme cases: when the special set is empty. that is. This fits older terminology (general type) in light of Conjecture 3. In this context. The Zariski closed subset Y can be taken to be the special set Sp(X). although already a diophantine flavor intervenes because one is led to consider generic fiber spaces. then X(F) is not Zariski dense in X. and when the special set is the whole variety. however. One basic conjecture states: 3. We shall mention below some specific examples.CL (131 THE SPECIAL SET 17 turally with the theory of hyperbolicity over the complex numbers. and X is fibered by projective lines. As a consequence of the conjecture. A projective variety X is pseudo Mordellic if and only if X is pseudo canonical. i. The following conditions are equivalent for a projective variety X: X is algebraically X is Mordellic. if the special set is the whole variety. let C be a curve of genus 2 2 and let X=CxP’. of X is pseudo canonical.7. one gets the conjecture: 3. where there may not exist a rational section.5. i. The special set Sp(X) is a proper subset if and only if X is pseudo canonical. the variety is special.Y is Mordellic. which amounts to problems principally in algebraic geometry.6. By Faltings’ . it is natural to define a variety X to be pseudo Mordellic if there exists a proper Zariski closed subset Y of X such that X . Then I conjectured: 3. One wants a classification of both types of varieties. so these examples are also examples of algebraically hyperbolic varieties. For instance. if X is pseudo Mordellic. and hence conjecturally of Mordellic varieties. We define a projective variety to be special if Sp(X) = X. Analytic hyperbolicity implies algebraic hyperbolicity. the special set is empty.

When is there a rational section of f. In the opposite direction. However. Suppose X = Sp(X) is special. do we still get the same set? Sp 2. concerning the structure of the special set. one can also ask for those conditions under which X’ would be generically fibered by a group variety rather than a rational image. Then one asks the general question: Sp 4. Notably. In connection with Sp 2. but X is not pseudo canonical. one is interested in those cases when X is special. conjecturally: 3. so that the generic fiber has a rational point over the function field k(y)? If there is no rational section. one must of course allow for blow ups and the like. independently of diophantine applications. over a number field k. namely: Sp 3. for every finitely generated field F over Q the set X(F) is not Zariski dense in X. One can also formulate an alternative for the second question.18 SOME QUALITATIVE DIOPHANTINE STATEMENTS CA §31 theorem. The condition Sp 2 means that there exists a generically surjective rational map f: X -+ Y such that the inverse image f-'(y) of a generic point of Y is a subvariety WY for which there exist a group variety G and a generically surjective rational map g: G + WY defined over some finite extension of the field k(y). 52. suppose f: X + Y is a rational map whose generic fiber is a rational image of a group variety. To get “fibrations” (in the strict sense. . Note that we are dealing here with rational fibrations. If we omit taking the Zariski closure. Sp 1. with disjoint fibers). Is there a generically finite rational map from a variety X’ onto X such that x’ is generically fibered by a rational image of a group variety? It would still follow under this property that it is not necessary to take the Zariski closure in defining the special set. Are the irreducible components of the special set generically fibered by rational images of group varieties? By this we mean the following. Let X be a special projective variety.8. The above discussion and conjectures give criteria for the special set to be empty or unequal to the whole projective variety. As a refinement of Sp 3. how many fibers have one or more rational points? We shall discuss several examples below and in Chapter X. to turn rational maps into morphisms. then the canonical class is ample. Zf the special set is empty. we have the following problems. This leads at first into problems of pure algebraic geometry. so only up to birational equivalence.

These considerations fit well with those of Chapter X. and he told me that Mori’s proof shows that in this case. which implies Theorem 3.9b as a corollary. is pseudo ample. 3. whose fibers may be homogeneous spaces for linear groups.5 implies that X is special. Theorem 1. or ample. In particular. under the hypothesis of Theorem 3. I would expect that there are more and more rational curves on the variety.e. Such algebraic geometers work over an algebraically closed field. I would raise the possibility of finding “good” models. Then through every point of X there passes a rational curve in X. without generic sections.9h raise the question whether a generically finite covering of X is generically fibered by unirational varieties.8. But still working geometrically we have the following theorem of [Mori 821. Then X is special.9. Let X be a projective non-singular variety and assume that -K. In addition. The question arises when the canonical class is 0 whether there exists a generic fibration as in Sp 2 and Sp 3 by rational images of abelian varieties. More generally I conjectured: Theorem 3. and there is of course the diophantine question whether rational curves are defined over a given field of definition for the variety or hypersurface. Let X be a projective non-singular variety and assume that -K.9a and 3. see also [ClKM 881. Interesting cases when the canonical class K. As we shall see later. is assumed more and more ample. As -K. such models can be found for abelian varieties (N&on models). or abelian varieties themselves. for the complete X. They intervene in the context of Chapter X. is not pseudo ample arise not only when K. i. X is special. is not an imbedding. where ultimately the special set is covered by rational curves and no abelian varieties are needed. I asked Todorov if Mori’s theorem would still be valid under the weaker hypothesis that the anti-canonical class is pseudo ample. and is equal to the union qf rational curves in X. Conjecture 3. contains an effective divisor. The existence of rational curves has long been a subject of interest to algebraic geometers. but we note that only rational images of P’ are needed to fill out the special set. possibly not complete. and the question is to what extent there exists an analogous theory for non-complete special group varieties. = 0 but also when -K. is ample.CL 031 THE SPECIAL SET 19 As refinements of Sp 2 and Sp 3. 91. besides intervening in the N&on . In these cases.9h.9a. Here the role of non-complete linear groups is not entirely clear. or is pseudo ample. and has received significant impetus through the work of Mori [Mori 821. there exists a rational curve passing through every point except possibly where the rational map defined by a large multiple of -K. Theorem 3.

In both cases. For important results when threefolds are uniruled. I propose to define a variety X to be unigrouped if there exists a variety X’ as in condition Sp 3. by the Kodaira criterion for pseudo ampleness. Extending the classical terminology of uniruled. (c) We have H”(X. Note that -K. is pseudo ample.11. 332.10. When is X unirational (resp. In particular. rational) over k? In this direction. We shall now consider several significant examples illustrating the Sp conditions. mK. there is a condition which is much stronger than having -K. One of Mori’s theorems proved a conjecture of Hartshorne: Theorem 3. projective variety and assume Then X is isomorphic to projective This result is valid over the algebraic closure of a field of definition of X. (b) X is uniruled. mK.) = 0 for all m > 0. they reflect the existence of degenerate fibers. and for a general exposition of Mori’s program. This suggests the existence of a theory of non-complete special models of special varieties which remains to be elaborated. We shall discuss other examples in Chapter X. Let X be a non-singular projective threefold (characteristic 0). One possible characterization of d being ample is that A? is ample. Indeed one can define the notion of ampleness for a vector sheaf & as follows. Let PB be the associated projective variety and let 9 be the corresponding line sheaf of hyperplanes. pseudo ample implies H’(X. Over a given field of definition. the following questions also arise: 3. ample. For further results see also Batyrev [Bat 901. Postscript Theorem. 3. see Kollar [Koll 891. a variety X of dimension r over an algebraically closed field k is said to be uniruled if there exists an (r . The following three conditions are equivalent. (a) Through every point of X there passes a rational curve in X. (a) Is Sp 3 then necessarily satisfied with a linear group variety? (b) Suppose that X is defined over a field k and has a k-rational point. Suppose that -K. that its tangent space.) = 0 for all m > 0. Theorem . Let X be a non-singular sheaf is ample. a variety may not have any rational point. There is another notion which has currently been used by algebraic geometers to describe when a variety is generically fibered by rational curves. among other things. having to do with negativity properties of the canonical class. in addition to Mori’s paper already cited see Miyaoka-Mori [MiyM 861 and Miyaoka [Miy 881.20 SOME QUALITATIVE DIOPHANTINE STATEMENTS CA 631 models of abelian varieties which will be defined later. p. or may be only unirational.1)-dimensional variety W over k and a generically surjective rational map f: P’ x W -+ X. Indeed. which yield the following result among others.12.

Sometimes there exists a rational section and sometimes not. 92. Example 2. These are algebraic formulations of a conjecture of Kobayashi in the complex analytic case. then by a result of Green-Griffiths [GrG SO] and BogomolovMumford completed by Mori-Mukai [MOM 831. Let X be a projective non-singular surface. over number fields. or a point of infinite order on the fibral elliptic curve. Let X be the generic hypersurface of degree d in P”. In such a case. See Example 1. Hence in this case we have Sp(X) = X. but which are linearly independent modulo algebraic equivalence. By generic we mean that the polynomial defining X has algebraically independent coefficients over Q. One says that X is the generic quintic threefold in P4. The existence of such fibrations in the case of sub- . We shall see an example with the Fermat surface below. Let X be the generic hypersurface of degree 5 in P4. Appendix. so that X is unigrouped. In this case the structure of the special set is known. [Cl 843) proved the existence of infinitely many rational curves which are Zariski dense. a generically finite covering X’ of X has a generic fibration by curves of genus 1. One says that X is a K3 surface if K. The next two cases deal with generic hypersurfaces. See also Chapter VIII. When such a section does not exist. The analogous conjecture goes for the generic complete intersection. Note that in light of Conjecture 3. This example will be discussed at length in $6. Example 4. so that the canonical class is ample. $1. and in particular X is special. Clemens ([Cl 831. Following a construction of Griffiths. If A is an abelian variety of dimension 2 and 2 is the quotient of A by the group { f l}.5 of Chapter VIII. Then conjecturally the special set is empty. called a Kummer surface. especially whether a generically finite covering of X is generically fibered by elliptic curves. = 0. or in the case of the Chatelet surface of Chapter X.6 these generic complete intersections would be Mordellic. thus explaining their independence in more geometric terms. and similar questions arise in the higher dimensional case of Fermat hypersurfaces. or K3 surfaces. and suppose d 2 n + 2. the Clemens curves might then be interpreted as sections over rational curves. then a minimal desingularization of Z is a K3 surface. If X is a K3 surface. if not by abelian surfaces themselves. in other words X is special. homologically equivalent.Ck 931 THE SPECIAL SET 21 Example 1 (Subvarieties of abelian varieties). or by rational images of abelian surfaces. Then K. Example 3. It is then a problem to determine whether X satisfies conditions Sp 2 or Sp 3 above. = 0 and if every rational map of X into an abelian variety is constant. and the answers to Sp 1 and Sp 2 are yes in both cases. the problem arises how many fibers have rational points.

. For more on the quintic threefold. The less the canonical class is ample.5. $2 for the Chltelet surface. the variety becomes less and less rational. . and the Fermat hypersurface is unirational over Q. I conjectured: 3. + Tnd= Tt Td’ + . see [La 861. For instance: Let X be a hypersurface of degree d in P”. For the argument. . . Since this hypersurface Tp + . the more a variety has a tendency to contain rational curves. The above examples conjecturally illustrate some general principles on some generic hypersurfaces.. and fibrations of the special set if they exist involve abelian varieties. Changes of behavior occur especially for d = n + 2. Fermat hypersurface. it is a open subsets rather than texts. + TS5 = 0. If d 5 n . then X This result is classical and In all the above. as the degree increases (so the canonical class becomes more ample). but I see no prospect of making further progress by the methods of this paper. + Tnd= 0 or contains lines. and n. see Chapter VIII. on To5 + . For the Fermat hypersurface if d = n. Here hypersurface). §22 and Remark 5.” In general. or at least that there are an infinity of solutions in three parameters.. but the evidence at the moment is still too scarce to convince everyone that the answer will always be positive.13. these fibrations may involve only linear groups and rational or unirational fibers. class Euler was already concerned with the problem of finding rational curves. and so the anti-canonical class is very ample. that is. For some examples in other con$1 for the Brody-Green perturbation of the Chapter X. 196. Roughly speaking. Swinnerton-Dyer [SWD 521 gives explicit examples of rational curves over the rationals. problem to determine what happens on Zariski generically.22 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL §31 varieties of abelian varieties can be taken as an indication that a positive answer may exist in general. n + 1. Here X has degree d = n = 5. solving the Fermat equation with polynomials. then the rational curves are Zariski dense. Swinnerton-Dyer says: “It is very likely that there is a solution in four parameters. p.1. contains a line through every point. we see that the condition that X has ample canonical does not imply that X is Mordellic or that the special set is empty. whereas for lower degrees. and we now consider: Example 5 (The Fermat easy. see [ClKM 881. .

Elkies found theoretically that in many cases there could not be a rational point. $4. Furthermore. p. by using Theorem 12. He was led to this solution by a mixture of theory and computer search. and the other one has infinite order on the fibral elliptic curve. Then Elkies including [El 881 found infinitely many solutions in degree 4. a fibration comes from Demjanenko [Dem 741. and the question is when a jiber has a rational point. one of them can be taken as an origin. the Fermat hypersurface has Ramanujan’s taxicab rational point (1729 is the sum of two cubes in two different ways: 9. For various points of view. it is jibered by curues of genus 1.e. and for the possibility of their being rational images of projective space for low degrees compared to n.Ck 031 THE SPECIAL SET 23 Of course one must take either d odd or express the Fermat equation in an indefinite form. as far as I know. perhaps dating back to Gauss.:+ x’:= x’:+ x:. . he proved that infinitely many fibers have at least two rational points. and it is the one used by Elkies. Rather. 55 and Shioda [Shio 731. there are other fibrations. The fibration of the Fermat surface by elliptic curves over C is classical. he knew how to make the computer deliver. the Fermat surface is a rational image of P2 over Q. i. Euler had a false intuition when he guessed that there would be no non-trivial rational solutions. see also Mumford [Mu 831. 10 and 12. 26824404 + 153656394 + 187967604 = 206156734. The Fermat surface can also be viewed as an example of a K3 surface. the conjecture is true in this case. This leaves open the problem of giving an asymptotic estimate for the number of rational points on the base curve of height bounded by B --+ co. 1). both geometrically and over Q. The point is that for the degree d = n + 1 there is no expectation that the Fermat hypersurface is unirational. But so far there are no systematic results known for the general Fermat hypersurfaces from the present point of view of algebraic geometry. The Fermat equation is even more subtle for d = n + 1. related to modular curves. Over Q. Furthermore. Example: When d = II = 3. First. for d = n = 3. when one expects fewer solutions. and in one remaining case. When written in the form x. Lander and Parkin [LandP 661 found the solution in degree 5: 275 + 845 + 1105 + 1335 = 1445. such that the fiber above those points is an elliptic curve with a rational point of infinite order. for the existence of rational curves.11 of Manin’s book [Man 743.

Assume that there exists a variety T and a generically surjective rational map f: T x W+X. It then follows that there exists a parameter variety T and a generically surjective rational map f:T x W+X. motivated by my conjecture that a hyperbolic projective variety is Mordellic. x W. From this one wants to split X birationally. there is only a Jinite number of isomorphism classes of pseudo-canonical varieties X0 for which there exists a generically surjective rational map of W onto X. following self-contained version as a conjecture. Then X is birationally equivalent to a product X.14 is the split case. and let X. Theorem 3.17. Let W be a oariety. Suppose that the set of rational points X(F) is Zariski dense in X. For a partial result in this direction.(k) under the birational isomorphism. we only have a conjecture.14 (Kobayashi-Ochiai). so that they lie in a finite number of families. The following analogue of de Franchis’ theorem was proved by Kobayashi-Ochiai [KoO 751. $4. be a pseudo-canonical projective variety. Let zn: X -+ W be a generically surjective rational map. Let X be a pseudo-canonical projective variety defined over a function field F with constant field k. Then there is only a finite number of generically surjective rational maps of W onto X. In [La 861 I stated the Theorem 3. 3. . it has been conjectured that: 781. analogous to Theorem 2. Theorem 3. All the rational points in X(F) outside some proper Zariski closed subset of X are images of points in X. Given a variety W. over F.. whose generic fiber is geometrically irreducible and pseudo canonical.24 SOME QUALITATIVE DIOPHANTINE STATEMENTS Ck §31 Let us now discuss the function field case..16. In the non-split case. On this subject see also [DesM More generally. see Chapter VIII. see Maehara [Mae 831.15.3 for curves. the problem is to bound the degrees of sections. Again see [Mae 831. As in the case of curves. Then there exists a variety X0 defined over k such that X is birationally equivalent to X. 3. who investigates algebraic families of pseudo-canonical varieties and rational maps.

although the fact that abelian varieties are commutative in all characteristics is due to Chevalley. see Vojta’s conjectures in Chapter II. Weil originally developed the theory algebraically.4(k) consist of the solutions of the affine equation with x. there exists a lattice.g2x . In other words. Let A be an abelian variety of dimension 1.CL §41 ABELIAN VARIETIES 25 Viehweg pointed out to me that this statement is essentially proved in [Mae 831 over an algebraically closed field of characteristic 0. Then A can be defined by an affine equation in Weierstrass form y2 = 4x3 . thus coming to the conjecture that a variety is Mordellic (or hyperbolic) if and only if every subvariety is pseudo canonical. abelian varieties are thus compact complex Lie groups. The function function.279: # 0. such that the map is an isomorphism of C/A with A(C). Example. . Over the complex numbers. The interplay between the diophantine problems and algebraic geometry is reflected in the history surrounding Theorem 3. whose points . 3. readers might look up my survey [La 863. then . g3Ek and A = g. and suppose A is defined over a field k of characteristic # 2.4(C) can be parametrized by the Weierstrass functions with respect to some lattice A.g3 with g2. and this theorem in turn made me conjecture that pseudo-canonical varieties are pseudo Mordellic. If k = C is the complex numbers. and coming to the definition of the special set.14. ABELIAN VARIETIES An abelian variety is a projective non-singular variety which is at the same time a group such that the law of composition and inverse are morphisms. $4. I. For quantitative formulations. My conjecture that hyperbolic projective varieties are Mordellic led Kobayashi-Ochiai to their theorem about pseudo-canonical varieties in the split case. $4. with basis ol. y E k together with the point at infinity in P2. For more information on the topics of this section. The corresponding projective curve is (isomorphic to) A. and are thus commutative groups. defined by @ is the Weierstrass . o2 over Z.

Let A be an abelian variety dejined over a number jield F. See Mumford [Mu 663 and Manin-Zarhin [MaZ 721. for themselves. and are defined over k. N&on [Ne 523 extended the theorem to finitely generated fields over Q. z) satisfies the universal mapping property . B) we mean the homomorphisms of A into B which are algebraic. it is much more difficult to write down equations for abelian varieties. and a homomorphism defined over F. By an F/k-trace of A we mean a pair (B. B) is a free abelian group. Abelian varieties of dimension 1 over a field k are precisely the curves of genus 1 together with a rational point. In the above parametrization. which is taken as the origin on the abelian variety for the group law. By Horn&l. wfo The sums over o # 0 are taken over all elements of the lattice # 0. Next we handle the function field case.26 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL $41 The coefficients of the equation g2 = 60 1 U4 ozo are given by and g3 = 140 1 (0.e. As usual. so when A is a curve of genus 1. finitely generated. we consider the two important cases when F is a number field and F is a function field. B be abelian varieties over a field k.1 (Mordell-Weil theorem). i. such that F is the function field of a variety defined over k. z) consisting of an abelian variety B defined over k. One of these ways is described in $5. the finite generation of A(Q) was conjectured by Poincarl and proved by Mordell in 1921 [MO 213. Let A be an abelian variety defined over F. Let A. the lattice points map to the point at infinity on P*. We sometimes omit the reference to k in the notation. Abelian varieties are of interest intrinsically. Weil extended Mordell’s theorem to number fields and arbitrary dimension [We 281. A basic theorem states that Horn. such that (B. In higher dimension. and also because they affect the theory of other varieties in various ways. the morphisms of A into B which are also group homomorphisms. especially if k is algebraically closed. Let k be a field and let F be a finitely generated extension of k. We are interested in the structure of the group of rational points of an abelian variety over various fields. When F = Q and dim A = 1. Then A(F) is a finitely generated abelian group. Theorem 4.@.

1. Example. then there exists a unique homomorphism c1*: C -+ B over k such that the following diagram commutes. when j E k or when j is transcendental over k. the situation in- .2 (Lang-N&on theorem). Suppose A is defined by the Weierstrass equation as recalled at the beginning of this section. Over the rational numbers. and follows since the set of points of a variety in a finite field (the constant field) is finite. Suppose A is defined over the function field F with constant field k./A. Let F be jinitely generated over the prime field. Chow defined and proved the existence of the that the homomorphism r is injective. Shioda in dimension 1 (for elliptic curves) [Shio 721 and Silverberg in higher dimension [Slbg 851 have shown that if A is the generic member of such families. See [La The analogue of the Mordell-Weil theorem proved in the function field case as follows [LN F/k-trace. of which the generic members are defined over a function field F over the complex numbers. This corollary is the absolute version of Theorem 4. and let (B. 3. We assume the characteristic is # 2. if (C. Then A(F)/z(B(k)) is jnitely generated.3. or in characteristic 0. Let A be an abelian variety defined over F. or over number fields. First consider the generic case. In other words. Let A be an abelian variety dejned over F. Let F be the function jield of a variety over k. Theorem 4. Torsion for elliptic curves over a base of dimension 1 has been studied extensively. was then formulated and 593. Corollary 4. t) be its F/k-trace.Ck §41 ABELIAN VARIETIES for such pairs. There may be two cases. then A(F) is finite. Consider for example the case when dim A = 1. Then A(F) is finitely generated. Questions arise as to the rank and torsion of the group A(F). and a homomorphism ~1:C + A over F. by using Theorem 4. and I mention only the latest paper known to me giving fairly general results by Miranda-Persson [MirP 891. the Lang-N&on theorem guarantees that A(F) is finitely generated. Abelian varieties distribute themselves in algebraic families. Assume that the F/k-trace is 0. E) consists of an abelian variety C over k. no matter what the variety. Let as usual j = 1728g. He also proved 591.2. In both cases.

d). [Ku 791. Mazur has conjectured. Here.. Question 4.. some results have been obtained by Kubert [Ku 761. Mazur has proved that the order of the torsion group is bounded by 16.5. Furthermore. For elliptic curves over the rationals. see Goldfeld [Go 791. see Zagier-Kramarz [ZaK 871 and Brumer-McGuinness [BruM 901. of the torsion group is bounded by C(F. One problem is to give a quantitative measure. or more cautiously. and also uniformly for families. . we mention only two qualitative conjectures. Given a number jield F. . The problem is also connected with the BirchhSwinnerton-Dyer conjecture. the following Kamienny [Kam 901 has shown for d = 1 and n = 2 that stronger uniformity is true: There exists a constant C(n. there exists a constant C(F.. for individual abelian varieties. Aside from the rank. and which we shall discuss in Chapter III. raised the question whether the following is true. . and C is stable under the Galois group Gr.. .jccFJ such that if A is an elliptic curve over F with a cyclic subgroup C of order N 2 N. and for computations giving relatively high frequency of rank > 1. or probabilistic description. some of which will be mentioned other chapters. For some current partial results on the rank. .6. . which relates the rank to certain aspects of a zeta function associated with the curve. the order #A(F). is bounded by C(n. For rank 10 see [Ne 521 and rank 14 see [Me 861. there is an integer N. one also wants to describe the torsion group. Over the rational numbers Q.jc(r. and a positive integer d. Over number fields. . d) such that for all abelian varieties A of dimension d defined over F.. No example of such elliptic curves is known today. . d) such that for all abelian varieties A of dimension d over a number jield F of degree n the order of the torsion group #A(F).28 SOME QUALITATIVE DIOPHANTINE STATEMENTS UT 041 in in volves a great deal of arithmetic. 4. . .(F) and a jinite number of values jI. Given a number field F.4. developing in the process an extensive theory on modular curves [Maz 771 and [Maz 781 which we shall mention in Chapter V. The general expectation lies in: Conjecture 4. then j(A) is equal to one of the j. and an asymptotic estimate of how many have a given positive integer as rank.(F). Shafarevich-Tate have given such examples for elliptic curves defined over function fields over a finite field [ShT 671. line with the general topics which have been discussed. Conjecture 4. of those which have one rank or the other. there exist elliptic curves A (abelian varieties of dimension 1) such that A(Q) has arbitrarily high rank.

5 by an argument due to Frey (see [Fr 87a.) of F. b]. and A is simple over L. is Jinite if and only if A does not have complex multiplication over F. Let A be a simple abelian variety over a number field F.7. for which I refer to his paper. is called the cyclotomic Z. $1) implies Conjecture 4. Zarhin also has results for the finiteness of torsion points in non-abelian extensions. if and only if End. Let A be an abelian variety over F. Theorem 4. Let F. Questions also arise as to the behavior of the rank and torsion in infinite extensions. Let p be an odd prime (for simplicity). Let A be an abelian variety dejined over a number Jield F. Let F be a number field. We also have: Theorem 4. = u F”. Let F.Ck 941 ABELIAN VARIETIES 29 We mention here that a certain diophantine conjecture. This was proved by Ribet (see the appendix of [KaL 821). Then F. Then the group of torsion points is jinite. Mazur raises the is jinitely generated. Then A(F. Zf A[p] n A(L) is # 0 for infinitely many primes p.9 ([Zar 891). we recall the definition that A has or has CM type over a field F. . with the above question of Mazur.) of p-power degree. be the cyclic extension of F consisting of the largest subfield of F(pr.-extension possibility that the following statement is true: Let A be an abelian variety over F. and also [His 881). and we shall state his main conjecture. The above theorem of Zarhin comes from other theorems concerned with non-abelian representations of the Galois group. Let p denote the group of all roots of unity in the algebraic numbers.. It also has something to do.8 (Zarhin [Zar 871). For the convenience Mazur [Maz 721 has related the question of points of abelian varieties in certain cyclotomic extensions with the arithmetic of such extensions. but less clearly. then A has CM type over L.(A) contains a semisimple commutative Q-algebra of dimension 2 dim A. Let 1 be a prime number and let L be an infinite Galois extension of F such that Gal(L/F) is a compact I-adic Lie group. Then A(Fab)to. complex of the reader. for instance: multiplication. For instance: Theorem 4. the abc conjecture (stated in Chapter II. [La 901.

for reasons which we shall explain later in this section. The factor group NS(X) = CH’(X)/CH. The superscript t indicates a transpose.. defined say over a field k. This subgroup of CH’(X) is denoted by CHh(X).. and with . 55. Let X be any non-singular variety.‘c(y2) for all pairs of points y. group. and let Y be a non-singular variety. and is also called the connected component of CH’(X). namely ‘c is the transpose of c on Y x X. The N&on-Severi group is Jinitely The history of this theorem is interesting. [Ne 521). For other results concerning points in extensions whose Galois group is isomorphic to the p-adic integers Z. Let X be a projective variety. see Mazur [Maz 831. non-singular in codimension 1. Let c be a divisor class on X x Y If x is a simple point of X we write c(x) = restriction Similarly. Severi had the intuition that there was some similarity between his conjecture that NS(X) is finitely generated. y. of c to {x} x Y identified with Y for a point y of Y we write ‘c(y) = restriction of c to X x (y} identified with X.(X) is called the N&on-Severi Theorem 5.1 (N&-on generated. The group generated by all classes of the form ‘c(yi) . and all classes c on products X x Y. and an even clearer connection was established by Lang-N&on. who showed how to inject the N&on-Severi group in a group of rational points of an abelian variety over a function field. I. will be said to be the group of classes algebraically equivalent to 0. E Y. To do this we have to give some definitions. see for instance Wingberg [Win 871. N&on made this similarity more precise when he proved the theorem.30 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL 951 For a survey of results and conjectures connecting the rank behavior of the Mordell-Weil group in towers of number fields with Iwasawa type theory and modular curves. and the Mordell-Weil theorem that A(F) is finitely generated for number field 6’. ALGEBRAIC EQUIVALENCE AND THE NCRON-SEVERI GROUP There is still another important possible relation between divisor classes.

: T(X)+A is a homomorphism. satisfying the universal mapping property phisms of X into abelian varieties. For any field k’ containing k. one can show that even if X does not admit a rational point over k. The morphism f: X -+ A can be extended. then A is also the Albanese variety of X over k’. that is those cycles such that cni = 0. if cp(P) = 0 then b = 0. X-A f for mor-+ B is a a unique following Of course. defined over a field k and having no rational point. then there exists homomorphism f.(a) is independent of the map f. Then the image S. and is called the Albanese variety of X. As a result. if cp: X morphism of X into an abelian variety B. we may identify the curve with its Albanese variety. there may be a projective curve of genus 1. Over any extension of k where this curve acquires a rational point.I% 951 a rational morphism ALGEBRAIC EQUIVALENCE AND THE NkRON-SEVER1 GROUP 31 point P E X(k). which was determined only up to a translation. In other words. The abelian variety A is uniquely determined up to an isomorphism. the curve does not admit an isomorphism with an abelian variety. There exists an abelian variety A over k and a j-:X-A such that f(P) = 0. Note that the existence of a simple rational point. Then with PieX(k’) and ni E Z. S. For instance. define the group of O-cycles b(X(k’)) to be the free abelian group generated by the points in X(k’). there still exists an abelian variety A . If k’ is an extension of k. but over the field k itself.: A + B and a point b E B such that the diagram commutes. Let Z?‘c(X) be the subgroup of O-cycles of degree 0. whenever a is of degree 0. A zero cycle a can then be expressed as a formal linear combination a = C ni(Pi) We define where the sum on the right-hand side is taken on A. so one usually does not need to mention a field of definition for the Albanese variety. or some sort of condition is needed on the variety X.

Observe how a geometric object. z) be the k(u)/k-trace of J. defined over k(u). deJined over the function jield k(u) (purely transcendental over k). and let S: 2Z’o(Cu) + J.2 (Lang-N&on [LN 591). If X and J are defined over a field k. k) 4 J&(u))/W4. k) consisting of those divisor classes c whose restrictions c. the N&on-Severi group. defined over k.(X. Let X be a projective uariety. we shall see instances when geometric objects are associated to rational points. Conversely. have degree 0. It will now be important to deal with fields of rationality... and S = S. be the Albanese homomorphism. Let %7 be the subgroup of CH’(X. is reduced to a diophantine object. be a linear variety defined by linear polynomials with algebraically independent coefticients u. The projective k) and CH’(X.2. That NS(X. k)/%’ % Z. A canonical map of X into its Jacobian is an imbedding. and we call S the Albanese bomomorpbism on the O-cycles of degree 0.. in the case of curves. be the Jacobian of C. Theorem 5. If X is a curve. and of dimension such that the intersection X.. such that over any field k’ where X has a rational point. so suppose the projective variety X is defined over a field k.... is the Albanese variety of X. k) is finitely generated is then a consequence of the LangN&on Theorem 4. Let (B. non-singular in codimension 1. is a non-singular curve C. and also an injective @/CK%X. k) we mean the group of divisor classes on X. then its Albanese variety is called the Jacobian. By CH’(X. and defined over an algebraically closed Jield k.32 SOME QUALITATIVE DIOPHANTINE STATEMENTS CI. Let J. the rational points of an abelian variety in some function field. so that the map imbedding of X can be chosen originally ct-+S(c.(k(u)).C. L. C.. We again call A the Albanese variety of X. . to C. 051 over k and a homomorphism (the sum) S: Z&(X(k’)) + A(k’) for every field k’ containing k. Let L. then one way to approach the study of the rational points X(k) is via its imbedding in J(k). A.) for cE% induces an injective homomorphism homomorphism 97 4 J. Then g XI CH.

we get an injection of NS(X. 6) is the dual of A then (A. given a projective imbedding. non-singular in codimension 1. It is a theorem that if (A’. we next describe the subgroup CHh(X) by showing how it can be given the structure of an algebraic group.(k(u)).) is also the Picard variety of X over k’. dejined over a field F jinitely generated over the prime jield. We call the pair (A’.. F) is jinitely generated. according algebraic equivalence [We 541. tually. This property is called the hiduality of abelian varieties.4. we find: Theorem 5. If (A’. the base change (A. k) Having described the N&on-Severi group. This explains why we call it a connected component. determined up to a translation. ‘6) is the dual of A’.3. Let A be the Albanese variety of X and let f: X -+ A be a canonical map. Then CH’(X. In the statement of the theorem. non-singular in codimension 1. which proceeds in the same way. then we can form its pull back to get the Picard variety of X. It is a theorem that for every extension k’ of k.. and in fact an abelian variety. and the pull back (f x id)*(d) is the Poincare class making A’ also the Picard variety of X. and for any jield k’ containing a’ H ?(a’) c(P) = 0. Let X be a projective variety.) is always finitely generated. after projective imbedding. the kernel of CH S(c. Then there exists an abelian variety A’ = A’(X).CL P51 ALGEBRAIC EQUIVALENCE AND THE NkRON-SEVER1 GROUP 33 Remark. Indeed. This suffices for the is finitely generated. a suitable choice of k) into J. Let X be a projective variety. k’). . The Picard variety of an abelian variety is also called the dual variety. Acthe homomorphism to basic criteria of proof that NS(X. and de$ned over a jield k. Let P E X(k) be a simple rational point. 6) is the Picard variety of A. Theorem 5. Putting together the finite generation of the N&on-Severi group and the finite generation of the Mordell-Weil group. k the map for a’ E A’(k’) gives an isomorphism A’@‘) 2 CHA(X. ck. The class c is called the PoincarC class. and a class c E CH’(X x A’) such that ‘c(0) = 0. The abelian variety and c are uniquely determined up to an isomorphism over k. we have a morphism fx id:X x A’+A x A’. c) the Picard variety of X over k. C.

Such a class is called a polarization. k) for higher codimension m.c. One can define Chow groups CHm(X.c equi- is a homomorphism of A into A’. B have the same dimension.34 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL VI Remark 5. A homomorphism of polarized abelian varieties f:(A. By abuse of notation. F) is finitely generated if F is finitely generated over Q. The map cpc: A -+ A’ satisfying cp. Proposition 5.c. A’). See for instance Beilinson [Be 851.(u) = element a’ such that ‘@a’) = c. let c. . = c. so we recall them there. depending only on the algebraic valence class of c. Thus one obtains factor groups analogous to the N&on-Severi group. In fact there are even much deeper conjectures of Beilinson and Bloch connecting the rank with orders of poles of zeta functions in the manner of the Birch-Swinnerton-Dyer conjecture and the theory of heights. one sometimes writes q&(u) = c. . we mean an abelian variety A and an algebraic equivalence class c E NS(A) which contains an ample divisor. this group is not necessarily finitely generated. By a polarized abelian variety. If c is ample.c)+(A. The association induces an injective homomorphism NS(A) 4 Hom(A.5. [Cl 843 and Example 4 of $3) shows that over an algebraically closed field. Let A’ be the dual variety..6. be the translation of c by a. Then A. We define an isogeny cp: A + B of abelian varieties to be a homomorphism which is surjective and has finite kernel. Given a class c E CH’(A) and a E A. To each polarization we shall associate a special kind of homomorphism of A. We shall use other properties of the Picard variety later in several contexts. such that f*c. It is still conjectured that CH”‘(X. . as well as cohomological equivalence. then cpCis an isogeny. and one can define the notion of algebraic equivalence. The example of Clemens ([Cl 831.) is a homomorphism of abelian varieties f: A + A.

Note that the set of Zi is empty if and only if X does not contain translation of an abelian subvariety of dimension 1 1. Let X be a pseudo-canonical subvariety of an abelian variety A in characteristic 0.7.[I. such that every translate of an abelian subvariety of A of dimension 2 1 contained in X is actually contained in the union of the subvarieties Zi. see also Iitaka [Ii 821. in which case we have: Kawamata’s structure theorem ([Kaw SO]). An important characterization was given by Ueno [Ue 731. In particular. I. §61 SUBVARIETIES OF ABELIAN AND SEMIABELIAN VARIETIES 35 In general.13. [Ii 77]). then X is pseudo canonical. Then we have quite generally. since all curves of genus 2 1 belong to this class. SUBVARIETIES SEMIABELIAN OF ABELIAN VARIETIES AND Such varieties provide a class of examples for diophantine problems holding special interest. and let B be the connected component of the group of translations preserving X. If c is a polarization. and [Mori 871. Theorem 10. [Ue 731. Then the quotient f: X + X/B = Y is a morphism. let cp: A -+ B be an isogeny. Then there exists a jinite number of proper subvarieties Zi with Ueno jibrations fi: Zi + x whose jibers have dimension 2 1. any . Theorem 3. Then X is pseudo canonical if and only if the group of translations which preserve X is jinite. $6. A basic theorem describes their algebraic structure. The variety Y in Ueno’s theorem is also a subvariety of an abelian variety. and whose jibers are translations of B. see also Iitaka ([Ii 761. we are reduced to pseudo-canonical subvarieties.10: Ueno’s theorem. Hence to study the full structure of subvarieties of A and their rational points. whose degree is called the degree of the isogeny. who proved: Let X be a subvariety of an abelian variety over an algebraically closed field. We call f the Ueno fibration of X. A polarizais tion of degree 1 is a polarization c such that (pe is an isomorphism. and is called a principal polarization. For a proof. then the degree of cpC called the degree of the polarization. whose image is a pseudo-canonical subvariety of the abelian quotient A/B. Then cp is a finite covering. Let X be a subvariety of an abelian variety A. if X does not contain any translates of abelian subvarieties of dimension 2 1. Theorem 3.

Then C n r. then the set of rational points X(F) is not Zariski dense. as they imply Corollary 4. Ochiai besides Ueno. this conjecture corresponds to the general conjecture of $4 applied to subvarieties of abelian varieties. In the early days of the theory. Suppose that the genus of C is 2 2.36 SOME QUALITATIVE DIOPHANTINE STATEMENTS CJ 961 Although the above version way in the given references. I these references actually prove [Och 773 made a substantial actually Kawamata who finally abelian subvarieties. see Noguchi [No 81a].2 (Faltings [Fa 903). be a jnitely generated No direct proof has been found. to prove Conjecture 6. By Kawamata’s structure theorem. we have a clear description of this special set. So in the case of subvarieties of abelian varieties. I formulated Mordell’s conjecture as follows. Chabauty [Chab 413 proved the statement when the rank of I-. is finite. For an extension of the above results to semiabelian varieties. Then X is Mordellic. let C be a projective non-singular curve. Let r.3. combined with the Mordell-Weil and Lang-N&on theorems. so we call Kawamata fibrations in X when of Kawamata’s am indebted the structure contribution proved the the union of X is pseudo theorem is not stated that to Lu for pointing out that theorem as stated.1. but it was existence of the fibrations by the subvarieties Zi the Uenocanonical..1 it suffices to prove that if X is not the translate of an abelian subvariety of dimension 2 1. is smal- . imbedded in its Jacobian J. Theorem 6. Let X be a subvariety of an abelian variety. In light of the determination of the exceptional set. In particular. the Ueno-Kawamata jibrations in X constitute the special set dejined in $3. and suppose that X does not contain any translation of an abelian subvariety of dimension > 0. The following especially important case from [La 60b] has now been proved. Let X be a subvariety of an abelian variety over a field F finitely generated over Q. and the statement is today a consequence of Faltings’ theorem over number fields. Then X contains a jinite number of translations of abelian subvarieties which contain all but a jinite number of points of X(F). subgroup of J. The structure theorems constitute the geometric analogue of my conjecture over finitely generated fields [La 6Oa]: Conjecture 6. We now see: For every subvariety X of an abelian variety A.

Although he works over a number field.. Let X be a subvariety of A. (1) The variety X is a curve (Raynaud [Ra 83a. the equation x+y=l has only a finite number of solutions in I?. The method is to pass to coverings and use Hurwitz’s genus formula together with an extension of a theorem of Siegel. Then X contains a finite number of translations of semiabelian subvarieties which contain all but a finite number of points of X n r. be a finitely generated subgroup of A(C). see [KuL 751. By a linear its division group. For an example in the context of modular units. By a semiahelian variety. A. Let I-.1. be a finitely generated subgroup of the torus.CL061 SUBVARIETIES OF ABELIAN AND SEMIABELIAN VARIETIES 37 ler than the genus of the curve. The same question was raised by Mumford at about the same time. for some positive integer m. Let A be a semiabelian variety over C. is injinite then C is the translation of a subtorus. In particular. his proof makes no use of that fact. one can always raise a solution to the power p. Conjecture 6. one sees how it is easier to formulate the statement in characteristic 0. so I called this equation the unit equation.3 is a theorem in the following in an abelian variety cases. this is the only type of example one can concoct. but we continue to assume characteristic 0. and depends only on p-adic analysis. If CA I-. and let I. We define its division group I’ to be the group of all points P E A(C) such that mP E r. and one gets infinitely many solutions from one of them.4. Let F be finitely generated over Q. to make the results easier to state. Such theorems or conjectures apply to somewhat more general group varieties than abelian varieties. Chapter IX. one means an extension of an abelian variety by a torus. apply Frobenius. b]) . be a finitely generated subgroup of A(C) and let I. i.. because in characteristic p. We know that A(F) is finitely generated. Theorem 6. Manin’s proof of the analogue of Mordell’s conjecture in the function field case led him to ask whether the intersection of a curve with the torsion group of its Jacobian is finite. Basically. or torus for short. one means a group variety which is isomorphic over an algebraically closed field to a product of a finite number of multiplicative groups.. Conjecture 6. Theorem 3. and let r. One may take I. I formulated a conjecture which would cover both the ManinMumford conjecture and the Mordell conjecture as follows [La 65a]. to be the group of units in a finitely generated ring over Z. In [La 60a] I proved in characteristic 0: Let C be a curve in a torus.e. Let A be a semiabelian variety over C. In general.3.

= (0) so I. Theorem 6. its division group. Then mA(X) is algebraically equivalent to m2X. [Li 753 proved the conjecture for curves in toruses. be a finitely generated multiplicative group of complex numbers. stemming from [La 65a]. Let f(x. which we describe briefly for curves X on an abelian variety. In higher dimension. Or briefly put: If a curve has an infinite intersection with the division group of a jinitely generated multiplicative group. be a jinitely generated subgroup of the torus... then f is a polynomial of the f orm uX’+bY”=O or cX’Y”+d=O withu. If X = Xcm) then mA gives an unramified covering of X over itself. (Raynaud [Ra 83a. Hindry [Hin 881 proved his result by a method involving the Galois group of torsion points. that is the group of complex numbers z such that zm E r. = (O}. y) = 0 be the equation for a curve in the plane. (Hindry [Hi 883) In particular. is i@nite. y E I. the conjecture dates back to Chabauty [Chab 381.b.c. then X contains the translations of a Jinite number of subtoruses which themselves contain X n r. where f is an irreducible polynomial over the complex numbers. §61 (2) The group 1-. Theorem 6. be multiplication by m on A. and hence X is of genus 1. and X is a subvariety of an arbitrary commutative group variety. Iv # Xc’“) then X n Xc”‘) has at most m2(deg X)’ points by a generalization of Bezout’s theorem. We restate it in more naive terms. then the curve is the translation of a subtorus.. Liardet [Li 741. and let I.(X) = s. and let m. b]) (3) Again r. It is contained in the next theorem...38 SOME QUALITATIVE DIOPHANTINE STATEMENTS [I. defined over a field F finitely generated over Q.dEC. If there exist injinitely many elements x. Let m be an integer 1 2.= A.6 (Laurent [Lau 841). If X has an infinite . y) = 0. of degree m2. For the intersection with roots of unity my conjecture had been proved by Tate [La 65a]. If X A I-. As a cycle. so I. Let X be a subvariety of a torus and let r. so is an abelian subvariety. for some positive integer m. Raynaud proved the Manin-Mumford conjecture. for the intersection of a variety with a finitely generated group of units. Xcm) where s is some positive integer and Xc”‘) is the set-theoretic image of X by m. is the group of torsion points of an abelian variety A but X is the group of torsion points.such that f(x. Let I-.

then the set of is jinite (modulo the trace zB(k)). 2) be the F/k-trace of A.) over F. using (*): X)‘. then d(n)I number ~ c - of points on X n Xcd’) 5 d”(deg We note that q5(n) 2 n I” for sufficiently large n.4. [La 83~1 Chapter 4. then we use a Galois torsion points which I conjectured. A partial result which suffices for the above application has been anounced by Serre. where r is a positive integer bounded by c. Theorem 2. E X n Xtd”. Property (*) was proved by Shimura in the case of complex multiplication. This immediately gives a contradiction as soon as n is sufficiently large. Furthermore. if r is in the group of automorphisms z d’x. : G) 5 c. Let the translation of an rational points X(F) The more general case when X may contain a finite number of abelian subvarieties is still unknown as far as I know. be the multiplicative group of integers prime to n mod n. E X n Xcd”. Let G be the subgroup of G. = d’x. To apply (*). Let d be a positive integer prime to n. of period n.field of a curve over defined over F. If X # Xcd” we obtain the inequalities. consisting of those integers d such that dx is conjugate to x over F. The above results and conjectures concern the absolute case of finiteness for rational points of subvarieties of semiabelian varieties. there exists an automorphism o of F(x. Then (G.7. but no proof is yet published. namely: Theorem 6.) over F such that OX. The relative case for abelian varieties over function fields is due to Raynaud [Ra 83~1. suppose that there exist points x. The general case is not known at this time. namely: theoretic property (*) Let A be an abelian variety defined over F. n -+ co. Let X be a subvariety of an abelian variety A. lying on X. Let k = C and let F be the function k.CLWI SUBVARIETIES OF ABELIAN AND SEMIABELIAN VARIETIES 39 of intersection with A(JJO. that is there exists an element o of the Galois group over F such that dx = ox. If X does not contain abelian subvariety of dimension > 0. By (*). See [Hin 881. = d’x.. There exists an integer c 2 1 with the following property. of translates . (B. Let x be a point of period n on ArLet G. Then GX.. of F(x.

Then we may view X.40 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL 971 I. A rational point P E X.(F) for an abelian variety A F’ In the case when A. and say X. Let X.(F) + X. and it is a basic theorem that one can choose the family of Jacobians {J. For each y E Yo(ka) we get the fiber X.(F) + A. Then there exists a non-empty Zariski open set Y.(F) + X..@(y)) . and for y E Y. is a curve of genus 2 1. = X. Then we can choose 7c:X+Y as above. HILBERT IRREDUCIBILITY Let F be a function field over the constant field k. + JF is a canonical imbedding of X.(F) = J. be a projective variety over F.. the imbedding The map {y} c Y induces a point PH Q(Y) induces a map X. for such y the specialization map X. We ask whether there exist “many” points y E Y(ka) such that the specialization homomorphism isbjective on A.(W) called the specialization specialization map map.(W) is a homorphism.(F). If X = A is an abelian variety. namely there exists a morphism such that if r] is the generic point of Y then n-‘(q) = X.(F). then the A.(F) is induced by the specialization homomorphism on J. called also the specialization of X. into its Jacobian over F. of Y such that 71 is smooth over Y. sP(y) E X. $7. = JF is the Jacobian of a curve.(k(y)). Then we have an imbedding X.. is non-singular for simplicity. and that (PF: X. depending on our choice of rc.) (y E Y. as the generic fiber of a family.) for the family {X....) over y.(F) corresponds to a rational section sp: Y-+X. and the specialization mapping on X. (or X. Write F = k(Y) for some parameter variety Y. Suppose that X.) to be compatible with specialization.

. Thus we obtain the morphism whose generic fiber is again A. . .(W) is injective.(/C’) is finite for every finite extension k’ of k. One way is by means of Hilbert’s irreducibility theorem. f may become reducible over the algebraic closure k”. Let 7c:A+Y be the family whose generic member is a given abelian variety A.1.. be the Zariski open subset where rr is smooth. has genus 2 2.(F). of A”(k) we mean that set of elements (ti. Let Y.. Then exists a Hilbert subset S c A”(k) such that for t E S and y E Y. the specialization homomorphism A#) + A. T. of Ye which has a finite morphism into affine space and in fact such that the image of I+Gcontains a non-empty Zariski open subset of A” (where n = dim Y). Then k is Hilbertian. Hilbert’s theorem can then be stated in the form: Theorem 7. ... Suppose we know that X. Suppose that k is a number field. Let u be another variable. .(F) c J. We have to make more precise what we mean by “many”. with variables Ti . There always exists a non-empty Zariski open subset Y..(k”). Such a set may of course be empty. Let f(Tu)=f(T.} as parametrized by A”. Let A” be affine space over k. T.2 (N&on [Ne 521). By a basic Hilbert subset S.. and that X. Let k be a number field. .. It then follows that X. by using points y for which the specialization map is injective. and is irreducible over k. which we shall now describe. ~1 = WI Cul be irreducible. Thus we may view our family of abelian varieties {A. The relevance of Hilbert sets to our specialization problem for rational points of an abelian variety over a function field arises as follows.. u) E k[u] has the same degree as f as a polynomial in u.. . but not necessarily geometrically irreducible. .. We say that k is Hilbertian if every Hilbert subset of A”(k) (for all n) is not empty.) E A”(k) such that the polynomial f(t. Theorem 7.(F) is finite.. . By a Hilbert subset of A”(k) we mean the intersection of a finite number of basic Hilbert subsets with a non-empty Zariski open subset of A”(k). $(y) = t.CA §71 HILBERT IRREDUCIBILITY 41 is also injective since X. that is.t. 4 E MT.

thus giving the possibility of constructing extensions whose Galois group is the symmetric group.. especially Corollary 6.3. [La 83a]. This may be useful in certain applications. . gives a more general scheme theoretic setting for the specialization theorem including reduction modulo a prime. . This point of view was taken by Emmy Noether. for all but a finite number of points t E A’(k) the specialization map is injective. . are the elementary symmetric functions of variables ui. The exposition of [La 83a]. .). The proofs of Hilbert’s theorem give various quantitative measures of how large such sets are. Another application is to the construction of abelian varieties over Q with as large a rank as one can manage for the group of rational points A(Q). and has given rise in recent years to extensive theories by Belyi. Theorem 2.3. a much better estimate for the set of points where the specialization map is injective follows from a theorem of Silverman which will be given in Chapter III. . . then Hilbert’s theorem implies that one can specialize the variables in Q in such a way that one gets the desired extension over Q.. Then there exists a Hilbert subset S c A”(k) (non-empty by Hilbert’s theorem) such that for t E S and y E Y(ka) with e(y) = t the specialization map is injective. If instead of Q one can construct such an extension over a purely transcendental extension Q(t i. For instance. which was first attempted by N&on [Ne 521. .. . Cf.t. so is X. The oldest example is when the independent variables t. if dim Y = 1. .. Corollary in light of Hilbert’s irreducibility theorem. say subgroups of GL. On the other hand. of curves of genus 2 2 with n defined over k. See for instance Voloch’s theorem at the end of Chapter VI. $5. I also raised the conjecture that the specialization theorem would apply to non-commutative groups.(k(y)) is finite.J. who got rank 10 for elliptic curves by this method.. .) there may be various geometric methods which make the construction more natural than over Q. Chapter IX.42 SOME QUALITATIVE DIOPHANTINE STATEMENTS CL 971 In particular. which have only semisimple elements and are finitely generated. Thompson and others. Chapter IX.3. Corollary 6.3. .(F).t.. Let k be a number field. One application of the Hilbert irreducibility theorem which has been realized since Hilbert himself is the possibility of constructing finite Galois extensions of Q with a given finite group G. . we get: Let Z: X + Y be a family 7.(k(y)) Thus we see that if X. The point is that over a field Q(t i.u. Let I++: + A” be a Y generically jinite map. XD’) -+X.

so we consider P”(Q). x. take n = 1. . we may assume that (x. THE AND HEIGHT FOR RATIONAL RATIONAL FUNCTIONS NUMBERS We wish to determine all solutions of an equation f(x.. . .{co} can be represented by coordinates (1. we estimate the maximum of the absolute values of the numerators and denominators of x and y. . y) = 0. If x. . and after dividing out by the greatest common divisor. . Then we define the height (or logarithmic height) h(P) = log max Ixjl. for instance. and let (x0. Part of determining the solutions consists in estimating the size of such solutions. y are to be elements of the ring of integers Z. (x. Then these projective coordinates xj can be selected to be integers.) are relatively prime integers. written as reduced fractions. . $1. where f is a polynomial with integer coefficients. In other words. Let P E P”(Q). j For example. and first deal with the rational numbers. involving several variables and more general domains than the integers or rational numbers.. A point in P’(Q) . and the solutions are in some domain. . Consider projective space P”.) have no prime factor in common.x. x) where x is a rational number. Write x = x0/x1 where .CHAPTER II Heights and Rational Points II. 1~1). .x. For instance if x. in various ways.. .. lyl or better the maximum rpax(lx).) denote projective coordinates of P. Thus we are led to define the size in a fairly general context. The size will be defined technically by a notion called the height. then we can estimate the absolute values 1x1. . y are taken to be rational numbers.

. CE Z and ptac.44 HEIGHTS AND RATIONAL POINTS CK 911 x0. 4 = log max(lx. Let F = Q be the rational numbers. for all x. Then 41. y E F. we always denote u(x) = -loglxl. An absolute value u on F is a real valued function XHIXI” = 1x1 satisfying the following properties. AV 1. in terms of absolute values other than the ordinary absolute value. and which we denote by u.(x) = order of x at p. The absolute value which is such that 1x1 = 1 for all x # 0 is called trivial. We have 1x1 2 0 and 1x1 = 0 if and only if x = 0. and for each prime number p the p-adic absolute value up which we define as follows. Ix + YI I max(lxl. The ordinary absolute value. Then we define the p-adic absolute value I& = l/P’. We do this fairly generally. Let x be a rational number # 0.I. which is said to be at infinity. Example 1. Let F be a field.. lx1 I). The choice of relatively prime integers to represent a point in projective space works well over the rational numbers. or that it is non-arcbimedean. There are two types of non-trivial absolute values. AV 2. When we speak of the set of absolute values on Q we always mean the set of absolute values as above. lxyl 5 1x1+ IYI. This set of absolute values satisfies the . lxyl = lxllyl AV 3. where r = ord. but does not work in more general fields. and write x = p’afc with a. xi are relatively prime integers. Ivl) d then we shall say that it is a valuation. If u is an absolute value. If instead of AV 3 the absolute value satisfies the stronger condition AV 4. so we have to describe the height in another way.

If u = up. By the product formula. For each rational function q(t) E k(t) we write . For each u E V(F). x # 0.x. log maXIXjlv. . The product formula can be rewritten as a sum formula: 5 44 = 0 for all x E Q. We give another fundamental example besides the rational numbers. Let k be a field and let t be a variable over that field. . .. It is easy to see that the height can then be defined in terms of the set of all absolute values.. Then we can form the polynomial ring k[t]. ” j In that form. THE HEIGHT FOR RATIONAL NUMBERS 45 called the (Artin-Whaples) product formula: y IX”1 = 1 for all x E Q. the expression on the right-hand side is the same for all equivalent (n + 1)-tuples representing a point in projective space. . Let p = p(t) be an irreducible polynomial (of degree 2 1) with leading coefficient 1 in k[t].) E P”(F) same formula as before: h(P) = 1 u E V(F) by the m. . . Then h(P) = 1 log max IxjlU. x # 0. x # 0. We can then define the height of a point P = (x. and its quotient field is the field of rational functions F = k(t). Let P = (x 0. .u(x) = 0 “E V(F) for all x E F. let m. if we have 1 m. Let F be a field with a set of non-trivial absolute values V(F) = (u}. be a positive integer. then up(x) = ord.x. Example 2.) be a point in P”(Q).CK Yl important relation. We may also use additive notation. the height generalizes. We say that V(F) satisfies the product formula with multiplicities m. .(x) log p.

i. be the subring of k(X) consisting of rational functions which are regular at a generic point of W.f. elements. Example 3. Let Then one can normal- 4. On the other hand. ize the absolute values up as follows. have no common irreducible factor. E In general.f. .e. malize u. If W is a subvariety of codimension 1. . Then irreducible polynomials of degree 2 1 and leading coefficient 1 are simply the linear polynomials t . . Such a normalization brings the shape of these absolute values in even closer analogy to those defined on the rational numbers. .) where fe.46 HEIGHTS AND RATIONAL POINTS CK §ll We may define the where f(t). . We may define u. . . which we take to be algebraically closed for simplicity. Suppose the field k is algebraically closed. . non-singular in codimension 1. let 0. and defined over a field k.) is also a finite field.L) =.)).) = 4tlAd be the residue class field. g(t) E k[t] absolute value up by are polynomials and p Jfg. . Similarly. we define u.max degfj . with qdegp = d. and consequently if x E k(X) then we have the order ord. by Iqllm = edegq.. by GU) = (de f) log # (4 = (deg f) log 4. just as for the rational numbers. Then I!?& is a discrete valuation ring. are polynomials which are relatively prime.CI with CL k.(f) log #(k(u. Icplp = l/erdegp.. Let X be a projective variety.(f) Then k(u. We let V be the set of absolute values Us. j Suppose that k is a finite field with q elements. Then the height is given by Wo.(x) of x at this discrete . . up for all irreducible polynomials p with leading coefficient 1. where # denotes the number of elements of a set. a point of P”(F) can be represented by coordinates (fe. Then V satisfies the product formula. . we nor- Thus we multiply all absolute values by log #(k).

. Then h.) be a point in P”(F). is the divisor of polecof a rational function y. In fact.y. . Szpiro and others. I where (y). We may call this example the higher dimensional function field case.. some of the results proved for rational functions are still unknown for the rational numbers. (b) Let f: X + P” be the rational map given by the rational functions (Y0. (a) Let P=(ye.(f) counts the zeros off by giving each of them multiplicity one. their analogues for rational functions can be proved in an easier way than for the rational numbers.. The degree is the degree in the projective space in which X is imbedded. Frey [Fr]. whose codimension is dim(W). projective degree. We let F = k(X). .(x) deg( W). where W is a projective variety. Then hxU’) = sup deg (Yi)..(P) = deg f-‘(E) for any sufficiently general hyperplane E in P”. We shall see several examples of such results. Then we can define an absolute value by u&x) = ord.(f) = number of distinct zeros of J Thus n..). . The set of all such absolute values as W ranges over all subvarieties of X of codimension 1 satisfies the product formula. A significant diophantine example: the abc conjecture. The above analogy between rational numbers and rational functions has been one of the most fruitful in mathematics because certain results for the rational numbers can be discovered by this analogy. L) where L is a sufficiently general linear variety of projective space. in a very original work as follows. . We define n. .. and as such has a deg( W) = intersection number (W. Furthermore. y. Let f(t) be a polynomial with coefficients in an algebraically closed field of characteristic 0. The height arising from this product formula may be described also in the following ways. This conjecture evolved from the insights of Mason [Mas]. and in the past. Mason started a trend of thoughts by discovering an entirely new relation among polynomials.CK §I1 THE HEIGHT FOR RATIONAL NUMBERS 47 valuation.

c(t) be relatively polynomials not all constant such that a + b = c. [Da 651.1.g(Q2) 2 4 deg f + 1. and similarly replacing x by y and z on the left-hand side. Note that the left-hand The proof is quite easy. c. By applying Mason’s theorem similarly. or without common prime factor for a. These two possible assumptions are equivalent by the equation a + b = c. Masser and Oesterk formulated the abc conjecture for integers as follows. b. Then n 5 2. Let f(t). Let x(t). Let k be a non-zero integer. side in Mason’s inequality is just the height of Mason’s theorem.(abc) . deg(f(t)3 . b. Then g(t) be nonconstant we we get a theorem of Davenport polynomials such that f(t)3 . c). b. find n(deg x + deg y + deg z) 5 3(deg x + deg y + deg z) .1 (Mason’s theorem). b. Let a(t). see also the exposition in [La application.1. In the statement a+b+c=O. z(t) be relatively polynomials such that one of them has degree _2 1. and such that x(t)” + y(t)” = z(t)“. Influenced by Mason’s theorem. Adding. c relatively prime in pairs.g(t)’ # 0. we get deg x(t)” 5 deg x(t) + deg y(t) + deg z(t) . Also the statement is symmetric in a. y(t). and considerations of Szpiro and Frey. This yields a contradiction if n 2 3. let us show how Mason’s theorem implies: H 90~1. c} 5 n. b(t).48 HEIGHTS AND RATIONAL POINTS CK §ll prime Theorem 1.3. by Mason’s theorem. b. observe that it does not matter whether we assume a. As an Fermat’s theorem for polynomials. c and we could have rewritten the equation in the form h(a. Then max deg{a. Define the radical of k to be . prime Indeed. Davenport’s theorem gives a lower bound for the difference between a cube and a square of polynomials.

Given E > 0. we shall formulate the corresponding inequality multiplicatively. there exists an integer nl such that for all n 2 n.GIL 911 THE HEIGHT FOR RATIONAL NUMBERS 49 i. Instead of viewing the Hall conjecture as giving a lower bound for a certain expression. The determination of n. Then the conjecture states that 1x1 5 C2(E)lb12+“. Also the abc conjecture is unproved today. Then 1x3 . Thus for polynomials we had an inequality formulated additively. where b varies over the integers # 0. the equation xn + y” = zn has only solutions with x = 0 or y = 0 or z = 0 in integers x. taking E = 1 for definiteness. For any non-zero relatively prime integers a. there exists a number C(s) having the following property. y. then they have to be compensated by “large primes”. By a similar argument as for polynomials. In other words.y2 # 0.(abc)‘+‘. y be integers such that x3 .(&)~X~(1’2)-~.y21 2 C. there is no conjecture as to what C(l). The analogue of Davenport’s theorem is Hall’s conjecture [Ha 711. and the constant C(E) on the right-hand side. or C(E) is like. the product of the distinct primes dividing k. occurring to the first power. as do others of similar type. . lbl. itjf necessary to have the E in the formulation of the conjecture. Let x. one sees that the abc conjecture implies the Fermat problem for n sufficiently large. or how C(E) behaves as E -+ 0. Under the analogy corresponds to between polynomials and integers. depending only on E. or even C(l). or many primes. and that if some primes occur to high powers.(E).y2 = b. ICI) 5 C(&)N. This conjecture also follows directly from the abc conjecture. c such that a + b = c we have max(lal. The conjecture implies that many or large prime factors of abc occur to the first power. whereas for integers. of an integer.e. b. of a polynomial log N. depends on the constant C(E). Today. Hall’s conjecture. for expressions of higher degree. we can also view it as giving an upper bound for solutions of the equation x3 . z. n. Unlike the polynomial case. The abc conjecture. for some constant C.

X) by (4A = 132.27b2 = ((II.360t3 + 216t2 .1) X = 324t4 .27b2 # 0.60t4 + 45t3 . 88.1). Q = 9t2 .tlI)(x in the complex numbers. that k 2 2. x.2X).5.18(8t . B. Then for x. namely: QY2=X3+AX+B one example showing with cases for which x. as in the discussion following Conjecture 5. which yields integral points provided 2Q is a square. y. already provide lems.lot + 3. For the most general cubic.CI~)(X . there should exist positive all cases. Conjecture 5. The small factor 2-253-4 means that one sees the exponent k nearing 2 only for large values of the parameter t. in §4. y) to y2=x3+132x+b with x N 2-253-4b4. That equation is satisfied by t = 1 and thus by infinitely many t.27b2 is called the discriminant D of the polynomial x3 + ax + b. Elkies has found k > 513. D = -4a3 then . . b are (See [La 83b].21t2 + 6t .CI~)(C(~ . yielding an infinite family of solutions (b. cubic equations. If we have a factorization x3 + ax + b = (x . and in particular A = 33. By the way. We shall describe Vojta’s conjectures later. In particular.))“. and there is no conjecture at this time as to what would constitute all such families. we have Ix/f C max(la13.1. Consider the equation y2 = x3 + ax + b with a. The Lang-Stark conjecture is a consequence of Vojta’s conjectures [Vo 871.50 HEIGHTS AND RATIONAL POINTS CK 911 unsolved prob- Such equations. B = . one has: difficult Lang-Stark conjecture. b E Z and with -4a3 . . Elkies rescales the equation by replacing (A. y E Z we have except for a finite number of families of polynomials satisfying the above equation.~1~) .) numbers k and C such that in Ibl’)“. No other example of such a family is known.5. the expression -4a3 .CQ)(C~~.CY.84t + 15 Y = 36(54t5 . a.

M = maximal ideal of null sequences.(k)3+“.e. See [Fr]. Then IuI S C(A.. Szpiro made the conjecture not even quite in this form. The completion is denoted by F. The discriminant the conjecture of this polynomial is (abc)2. B. be the ordinary absolute value. THE HEIGHT IN FINITE EXTENSIONS Let F be a field with an absolute value v. and a. and let k = Au3 + Bv2 # 0. Generalized Szpiro conjecture. &V.} such that lim Ix. 82.a)(t + b). but as a function of a more subtle invariant N(D) instead of N.(D). +V.. we denote its algebraic closure by K”. This construction works just as well starting with any field F and any non-trivial absolute value v. For instance. II. Let u. v be relatively prime integers. let F = Q and let u = v. sequences {x. called the conductor. 5 It is an exercise to show that the generalized Szpiro conjecture is equivalent with the abc conjecture. to which we can extend the absolute value by continuity. We let R = ring of Cauchy sequences of elements of F. where D = -4a3 . Then R/M is a field. If K is a field. b are relatively prime. This conductor is irrelevant for our purposes here. To do this one uses Frey’s idea. It is a fact that the absolute value on the completion extends in a unique way to an absolute value on the algebraic closure of the completion.. The Szpiro conjecture has to do with this discriminant. Fix integers A.CK 921 THE HEIGHT IN FINITE EXTENSIONS 51 that the three We state it in Thus the condition D # 0 is equivalent to the condition roots of the polynomial are distinct. which is to associate with each solution of the equation a + b = c the Frey polynomial t(t .(k)2+” dnd 1111 C(A. We may form the completion of F. In fact. Szpiro actually only made IDI 5 cw%(w+“. a generalized form.. The construction of R can be generalized as follows.27b2 is a discriminant # 0. Then the completion is the field of real numbers R. Thus we obtain an . B # 0. B./. i. = 0.

which may not be complete. or if necessary to make more precise. . E V(F).. T.(t) be the field of rational functions. . If v = v.(f) On If a. then Q. Let v extend the absolute value q. Let F be a field with a set of absolute values V = V(F) satisfying the product formula.. The two standard examples are the cases when F = Q and F = k(t) as in $1. = completion There is an imbedding of the algebraic closure of F. on Q. for instance. the other hand. we let V(F) be the set of absolute values which extend v. consisting of all power series f(c) = f n=I a. # 0 then = v. = F.. and a finite extension of k(t) is called a function field. = FOO simfor plicity.52 HEIGHTS AND RATIONAL POINTS CK VI absolute value on F. ord. Let F = F.“. 2. a function F.c” where r may be positive or negative integer. . such that. . on Q. and the absolute values are in bijection with irreducible homogeneous polynomials in the homogeneous variables T. is called a p-adic field. . It is a fact that C. then Q. F.. is a finite extension of the completion F”. when F = k(x.x. . . Let F be a finite extension of F. We can describe elements v E V(F) as follows. is an infinite extension. . which we denote by C. Then the completion F. for some prime number p. This is the higher dimensional version of Example 2. Example 3.((t)). We also write F.. and its algebraic closure Q. Then F.. and let Then the completion F.(f) = r. = Q.((l/c)) field in one variable. . .. Example 1. or v. if v = up is p-adic.. is algebraically closed.((u)) = is the field of power series in l/t. We have the field C. = R and C. let v = v. is the field of power series Let v=vp. A finite extension of Q is called a number field. We let V(F) be the set of absolute values on F which extend those in V..) is the field of rational functions in n variables. We shall consider finite extensions of F. Put u = l/t.“. = C is just the field of complex numbers? However. Xi = T/T. Hence if F is a number field. Example p(c)=c. We may then form the completion of F. We may view F as the function field of P”.

c. in&ce the same absolute value on F if and only if there exists an isomorphism z of F. we assume that this is the case. or u corresponds to a pair of complex conjugate imbeddings of F into the complex numbers. We may therefore define the height of a point in projective Let P =(x0. . it follows that the set of absolute values V(F) satisfies the product formula with multiplicities [F.) with xj E F.x. VI THE HEIGHT IN FINITE EXTENSIONS 53 and v is induced by the absolute value on C.: &I = CF: Fl- The symbol vlv.] log max IXjl”* h(P) = [F : F] 0Ev(~) i space 1 The factor l/[F: F] in front has been put there so that the value h(P) is independent of the field F in which the coordinates x0. . CF.x. then we say that V is a proper set of absolute values. Two imbeddings cri . x # 0. lies in V) then [F. In general. 2 and 3.. For each absolute value v on F we have what we call the local degree We also have the global degree [F : F]. (leaving F.. From now on. F is a number field. and v extends the absolute value at infinity. . . . if every absolute value v0 of V has the property (*). cr2: F + C. . Then v corresponds either to an imbedding of F into the real numbers. fixed) such that Example. As a result. we let v. . Suppose F is a finite extension of Q..r?. are relatively prime integers.: FJW =0 for all x E F. it is not difficult to prove that if u0 is an absolute value on F (i. be the induced absolute value. . Note that if F = Q and x a. : F.e. . not all xj = 0.]. means that the restriction of u to F is z+. that is. These are the degrees of the finite extensions F. . : F. . : F. (*) E. CF. namely os. and the sum is taken over all u E V(F) such that u(ue. In the standard Examples 1. In any case. given such an imbedding CJ:F + C. we now view the height h: P”(F”) + R . then h(P) = log max /Xjl so we get the same height discussed in $1. lie. x. Conversely. We define _____ 1 IF.]. P”(F). over F. . and F over F respectively.

Given an algebraic number CY. We can take for c the leading coefficient in the irreducible equation for CIover Z. : F.EZ.]v(x) occurs quite frequently. or rank [F : Q]. nzl. j = uriQF)max Ilxjllu~ j Writing the height multiplicatively suggests more directly certain bounds for algebraic numbers. . Just as a rational number is a quotient of integers. + a.. Usually. we define H.. It can be shown that oF is a free module over Z.. there exists an integer c E Z. When F = Q. we shall use the notation llxll ” = IxI[F”:FJ .overZ. the height is a function on the set of all algebraic points. . Let F be a number field. generalizing the ring of ordinary integers Z. there exists a basis (a . letting d = [F : Q]. xn-l + . Number fields Suppose F is a number multiplicative field. This ring is denoted by oF. that = [F : Q]. can be reduced to the study of algebraic points for a single equation.e. Since the expression [F. called the ring of algebraic integers. points in the rational numbers Q. . = 0 with u.. Then rl + 2r. We shall see examples of this phenomenon later. one denotes by ri = ri (F) the number of real imbeddings of F.}ofo. i.lW = -lwllxll. c # 0 such that ctx is an algebraic integer. An algebraic number x is said to be an algebraic integer if x is a root of a polynomial equation X” + a. algebraic over the ground field F. In other words. we are interested in algebraic points because sometimes the study of rational points. paints whose coordinates are algebraic numbers.-.54 HEIGHTS AND RATIONAL POINTS m §21 as a real valued function on the set of point in P”. Aside from the intrinsic interest of knowing about algebraic points. It is sometimes useful to deal with the height relative to F.(P) = exp([F: Q]h(P)) = n usV(F) max (xjIL’“‘Qul.. for a family of equations. ” Then CFo:F. and by r. i.(F) the number of pairs of complex conjugate imbeddings. The set of algebraic integers in a number field F is a subring. = r. we have a similar representation for algebraic numbers as follows.

.1.x. called the ideal class group. Then we have the formula H. The map U++(. Thus the image is contained in a euclidean space of dimension r = rl + r. Then the discriminant of F is DF = (det aiaj)2r and Jh = IW We define the (normalized) logarithmic discriminant to be 1 d(F) = CF : Q3 log D. . on Q.~~~~ll~ll”~. . Let P = (x . and the regulator is the volume of a fundamental domain for this lattice. . Let {ai . be elements of or not all 0. :. ..a. . the ideals (non-zero. . be the set of absolute values at infinity of F.?nd write a . . b be ideals of oF. Let a. . and is denoted by h. . It can be easily shown that under multiplication. We say that a is linearly equivalent to b. . Dirichlet’s unit theorem states that the image is a lattice in this space.CK 021 THE HEIGHT IN FINITE EXTENSIONS 55 By a basic theorem of Dedekind. . .~ is a homomorphism of U into R’I+‘~ whose image is contained in the hyperplane consisting of those elements such that the sum of the coordinates is 0. vs.. and let a be the ideal generated by these elements. is the set of absolute values on F extending u. (Not to be confused with the height!) Let a be an ideal..b if there exists an element CIE F. If we have the unique factorization then it is a fact that Na = n Np”?. . We define the absolute norm of a to be Na = number of elements in the residue class field OF/a.(P) = Na-’ n max Ilxjll. by the product formula.& j where S.a. The order of this group is called the class number of F. .. always) of oF admit unique factorization into prime ideals. the ideal classes form a group.} be a basis for oF over Z. P Let x 0. ...~“. be the distinct imbeddings of F into C.) be the corresponding point in P”(F). c(# 0 such that b = aa.. Let U = of be the group of units. . .x. Let S. Let oi.

and we immediately two properties which are used frequently: c if F. Since . Theorem 2.) + d(F. for any number fields FI.56 HEIGHTS AND RATIONAL POINTS CII.). we have d(F. and the product is taken over all (non-zero) prime ideals p. F. then we obtain a bound for the coefficients of an equation for this algebraic integer over Z. Specifically: Theorem 2. is the class number. is the regulator. and there is only a finite number of ordinary integers in Z having bounded absolute value.) 5 d(F. The residue at this pole is 2”(27t)“hR wD”’ where h = h. is the absolute value of the discriminant. & = n (1= P Np-)-’ . The main point of this theorem is that it is uniform in the degree and does not only concern the points in P”(F) for some fixed number field F. The idea of the proof is that if we bound all absolute values of an algebraic integer of degree d. D = D. The zeta function gives sometimes a convenient analytic garb for some relations between the notions we have defined. F.) 5 d(F. is the number of roots of unity in F. R = R. Finally.1. 621 list The logarithmic discriminant will be useful later. then d(F. there is only a finite number of points in projective space P”(Q”) of bounded height and bounded degree. First we have a completely elementary result due to Northcott. More generally. the zeta function of F is defined for complex Re(s) > 1 by the formula numbers s with IF(s) . There is only a finite number of algebraic numbers of bounded height and bounded degree over Q. We shall apply further these notations of algebraic number theory to the height. The sum is taken over all (non-zero) ideals a of oF. c F.2. The zeta function has an analytic continuation to a jiinction which is holomorphic on all of C except for a simple pole at s = 1. w = w.).

and of units in oF.3. n. Theorem 2. It is interesting to give an asymptotic estimate for the number of elements of a number field F of height 5 B for B + co. we obtain tkt theorem for algebraic numbers. Let N(t) = lv(t. Assume that the boundary of W is (n . l]“-i + Boundary of W (n . In each case. In practice. $2. L) be the number of lattice points in tW for t real > 0. and such that each mapping satisfies a Lipschitz condition. not just for algebraic integers.1)-Lipschitz whose images cover the boundary of W. Let W be a subset of R”. the method of proof consists of determining the number of lattice points in a homogeneously expanding domain which has a sufficiently smooth boundary. The expression parametrizable means that there exists a finite number of mappings p: [0. that is with continuous partial derivatives. Let F be a number field.1.CK VI THE HEIGHT IN FINITE EXTENSIONS 57 the height bounds essentially both the numerator and denominator of an algebraic number.1)-Lipschitz parametrizable. of algebraic integers in or. This question can be asked of elements of F. let L be a lattice in R”. Then N(t) = ~ vol(L) where the constant constants. and using the following basic fact. Proposition 2. Chapter VI. Let r = rI + rz . The number of units u E oj$ with Hr(u) 5 B is (ii) y:(log B)’ + O(log B)r-’ for some constant yz > 0.4. see [La 641 or [La 703. . 5 B is (i) The number of algebraic integers x E or with height Hr(x) yOB(log B)’ + O(B(log B)‘-‘) for some constant y0 > 0. implicit Vol( W) t” + O(@)? in 0 depends on L. and the Lipschitz For the proof. w. and let Vol(L) denote the euclidean volume of a fundamental domain for L. such mapping exist which are even of Class C’.

(B) = ifo~~F. [We 511. we let F be a field with a proper set of absolute values satisfying the product formula. On the other hand. and is mostly due to Weil [We 281. The fundamental theorem about the relation between heights and divisor classes runs as follows. Theorem 3. (i. Schanuel [Sch 641./ 58 HEIGHTS AND RATIONAL POINTS [II> 631 Both parts come from a straightforward application of Proposition 2. following work of Siegel and Mahler. Let N.nBn The constant yr. For the counting of values of binary forms see [May 641. THE HEIGHT ON VARIETIES DIVISOR CLASSES AND Throughout this section. and both constants are easily determined. with Theorem 2.3. Let X be a projective variety. [Sch 793. The height of points in P”(F) for finite extensions F of F is then defined as in $2.: X(F”) well defined modulo -+ R.1. defined over the algebraic closure F”. II. $3. n = 2 otherwise.5. Theorem 2. satisfying the bounded functions .” has the value n’ + O(Blog B) O(Bn-lld) if d = 1. To each Cartier divisor class c E Pit(X) one can associate in one and only one way a function h.(B) be the number of elements x E P”-‘(F) height Hr(x) 5 B. has determined the somewhat harder asymptotic behavior of field elements of bounded height in projective space as follows. Then: hR/w N. For the setting of Schanuel’s counting in a more general (conjectural) context.5 generalizes the classical fact that the number of relatively prime pairs of integers of absolute value s B is fBz + O(B log B). see Chapter X.e. Let d = [F : Q]. mod O(l)). $3.

0 g + O(1). this height is the same as the height ho.1. then modulo h. let c be an ample class.. Thus if 9 is a line sheaf corresponding to the divisor class c. for the corresponding height. Choose h. Although I find it convenient here and elsewhere to use the language of divisor classes. we have As a matter of notation. Theorem 3. and we also write ho = h. Thus h. especially in Chapter VI. then h. By Theorem 3. We also want to describe the positivity properties of the height.2. we write this class as cD if we want to make the reference to D explicit. depends only on the Cartier divisor class of D. O(l). mod O(1). If c is ample. 2 -O(l). where D is a hyperplane.(P) = h(f(P)) This height association mod 0( 1). It is an elementary observation from algebraic geometry (to be recalled more explicitly later) that Pit(X) is naturally isomorphic to the group of isomorphism classes of line sheaves. then for c E Pit(Y) hgac = h. we shall also deal with line sheaves. . we also write hY instead of h. is a homomorphism for some hyperplane H. On any projective variety we have the following theorem. Remark./ CK 631 following THE HEIGHT ON VARIETIES AND DIVISOR CLASSES 59 properties: The map c H h. and let c’ be any class. property that if If f: X + Pm is a projective imbedding. The above theorem gives the basic properties of the height in its relation to divisors and the operation of addition. Observe that the height on projective space as defined in $2 is always 2 0. we have as described in $2. in other words we can choose h. as well as morphisms. in its class modulo bounded functions such that h. Furthermore. 2 0. and c is the class off -l(H) = height of f(P) satis$es the additional g:X-rY is a morphism of varieties defined over F”. if D is a divisor and c its class.

2 0. Let X be a curve. Proposition 3. Let c E Pit. [La 60a] and [La 83a]. Then one can choose h.60 HEIGHTS AND RATIONAL POINTS CK §31 such that h. roughly speaking. We shall see especially significant examples of such relations when dealing with abelian varieties. we have h(P)-w lim h. we may say that the association CH h. 2 0.(X) be algebraically equivalent to 0. Example. In fact. For classes cr. we have h. there is also a property corresponding to the weaker notion of effectivity. Geometric relations between divisor classes thus give rise to relations between their height functions.. and let E be an ample divisor. If the . The height h denotes the height associated with any ample class. P $ supp(D). Then there exist numbers yl. Then h.1.2(P) = 1. Proposition 5. Cf. yz > 0 such that Theorem 3. Chapter 4. c2 which are algebraically equivalent and ample. Theorem 3. preserves all the standard operations on divisor classes: the group law. for hE+a.4. in its class modulo bounded functions such that for all P E X(Fa)..3 is immediate from Theorem 3. The relation of algebraic equivalence gives rise to a height relation when the variety is defined over our fields F. Theorem 3. Thus.Wh. Let X be projective non-singular.3.2 translates the strongest positivity property of divisors into a property of the associated height.4. inverse images and positivity. Theorem 3. for P E X(F”). and Chapter 5. However. = o(hs) on X(F”). selecting h.(l). = 0(h22) + O. dejined over F”. Suppose that X is projective and non-singular and that the class c contains an eflective divisor D. Two divisors on X are algebraically equivalent if and only if they have the same degree.3..

c. with exponential growth.1. c(. we shall find logarithmic growth in the next chapter. For abelian varieties. it goes to infinity for every other ample class. and assume that the associated quadratic form is positive non-degenerate on r. B) = N(B) in X(F) such that H. One difficulty today is that one does not know in an effective way whether there is a rational point or not to start with. We have seen an example for projective space in $2. A key ingredient is the following theorem of Mumford [Mu 651. Suppose that c is ample. In the function field case. Theorem 4.} be a sequence of distinct points in C n I-. so there is only a finite number of points. Here we use the exponential height H. geometric methods give an explicit bound for the heights of rational points. are proved by showing that a height is bounded. N(B) grows like yB”(log B)“’ for some constants y. Let J be the Jacobian of C and let r be a finitely generated subgroup of J(F”).. and that X is defined over a number field F. one has some methods for bounding the heights of other rational points. ordered by increasing height. N(B) grows like y(log B)‘. cc’. II. Let C be a non-singular curve of genus 2 2 dejned over a finite extension of a field F with a set of proper absolute values satisfying the product formula. No such method is known today in the number field case. and as a result current proofs give an effective (albeit inefficient) bound for the number of rational points. Roughly speaking. = exp h. Let h be the height associated with an ample class. $4.In Q41 BOUND FOR THE HEIGHT OF ALGEBRAIC POINTS 61 height goes to infinity for one ample class. Roughly speaking. 44. the following cases emerge: N(B) = O(l). One can try systematically to give an asymptotic formula for the number of points N(X. once one knows the presence of one rational point. Let {I’. BOUND FOR THE HEIGHT ALGEBRAIC POINTS OF The finiteness statements for rational points on curves. both in the number field and function field case. 5 B for B -+ co. Then there is an integer N and a number b > 1 such that for all n we have Thus Mumford’s theorem shows that the points of Cn r are thinly . Some systematic conjectures due to Manin and others will be discussed in Chapter X. should be normalized in a suitable way. and h.

) 2 a.(k). Over certain number fields. For instance. using the Frobenius element. so H(P. Let X be a non-singular curve of genus 2 2 defined over a function field F of characteristic 0.L IYA. b”.2. so that WJ. over a number field one has a much stronger result namely the finiteness of the set of rational points. is isomorphic to X over F. Note that Faltings’ theorem is only vaguely related to Fermat’s problem. over the constant field k. Suppose X(F) has infinitely many points of bounded height.. Mumford already remarked that the rate of growth for the height that he indicated is best possible. say. the Fermat curve over the rationals Xd + yd = Zd with d 2 4. Then there exist numbers a. only a weak statement comes out.62 HEIGHTS AND RATIONAL POINTS GIL 941 distributed. In the function field case. > 0 and b > 1 such that log H(P. in relatively prime integers. dejined over k such that X. The hypotheses of the theorem apply when C is a curve over a number field F and I = J(F). Then there exists a curve X. in the sense that their heights grow rapidly. Indeed. Id. Theorem 4. in the concrete case of. by the Mordell-Weil theorem. of course. suppose we have a sequence of solutions P” = (-%I? 4 Yn. one has the following splitting theorem stemming from [La 60a]. and all but a jinite number of points in X(F) are the images under this isomorphismof points in X. But in characteristic p. We stated Mumford’s theorem deliberately under quite general hypotheses to show that it is valid without making use of specific arithmetic properties of the ground field. Faltings’ theorem applies to all number fields.) >=es@ grows doubly exponentially. Fermat’s problem has to do specifically with the rational numbers.) = max(lx. the Fermat curve will have many other solutions besides the trivial solutions with one of the coordinates equal to 0. . Since only weak hypotheses are used. albeit a useful one.

Observe that if P ranges only over X(F). The conjecture is stated here with a strong uniformity for all algebraic points.(l).(l) is a bounded function of P. In particular. is the normalized logarithmic discriminant already dis- Vojta’s conjecture 4. If 0. We define d(P) = d(W)) = d.fAp E k(P). We define k(P) = k(z. We shall describe some of the various geometric methods giving such bounds in Chapter IV. since for genus 2 2 the canonical class is ample. and Vojta himself sometimes feels it is safer to make the conjecture uniform only with respect to points of bounded degree. so Vojta’s conjecture implies Mordell’s conjecture at once. we have conjectures.. In the direction of bounding the height of points on curves. Let K be the canonical class of X. Let X be a curve defined over a number jield F. Although the constant 1 + E is conjecturally best possible.).GIL VI BOUND FOR THE HEIGHT OF ALGEBRAIC POINTS 63 The problem was to prove that the set of points X(F) has bounded height. .(l) is. for all algebraic points P E X(Q”) we have h. . where d(F(P)) cussed in $2. no conjecture exists describing how big O. . . d(P) + O(l) . the conjecture does not give an effective bound for the heights of algebraic or rational points. and therefore the right-hand side is bounded. The term O. . Suppose X is a curve defined over a field k. and let (zi .) be a set of affine coordinates for P. Suppose X is defined over a number field F.(P) 5 (1 + E) d(P) + O. Let P be a point of X in some field containing k. Given E > 0.(P).z. . . with a bound depending only on X and E. then we have a natural isomorphism O.z. is the local ring of regular functions at P and JZP is its maximal ideal. .3. As of today. as a function of E. it would already be a great result to prove the inequality Mf’) S C. then d(P) is constant.

Actually. Evidence for this comes from the function field case (Chapter VI.64 HEIGHTS AND RATIONAL POINTS GIL 641 with some constant C. z the point on the Fermat curve of degree 4 given by p = (p/4. There exists an integer n. See [Vo 871.e. the conjecture should be written bm i W) + ooog h. which is however usually much larger than the discriminant (see Chapter VII. Part of the importance of Vojta’s conjecture is that it sometimes allows one to reduce the study of rational points of a family of curves over a given field to the set of rational points of a single curve. We list some of these specifically. $5. Vojta’s conjecture (the weaker form for points of bounded degree) implies all the concrete diophantine problems mentioned in 51.(P)) + O(l) or still better h. with additional comments. but over algebraic extensions of this field. i. If it ever turns out that this constant can be determined. but Vojta has proved a similar result with an arithmetic discriminant [Vo 90b]. Theorem 5.(P) 5 d(P) + (1 + E) log h. end of Chapter V.. One family is that of all Fermat curves of degree n for n 2 3. $2).(P) + O. This reduction is carried out by associating to each solution of x” + y” = z” with relatively prime integers x. No such result is known today. of bounded degree. then one would have a solution of Fermat’s problem. The extent to which n. for instance. and Vojta’s appendix [Vo 881. denoting by E an ample divisor. .6). can be determined effectively depends on the effectivity of the constant in Vojta’s conjecture. and is sufficiently small that the remaining cases can be given to a computer. $2) and the holomorphic case of Nevanlinna theory (Chapter VIII.4. Corollary 4. Vojta’s conjecture reduces the study of their rational points to the algebraic points on the single Fermat curve of degree 4. y. Thus the difficulty of studying rational points is shifted.. such that the Fermat equation x” + y” = i? has only the trivial solutions in relatively prime integers for n 2 n. p/4).(l). y”‘4.

. Let F be gebraically closed jield k such that x + y = 1. E) n Np’+‘. E) such that for all a. Corollary 4.b = c and abc # 0 we have &(a. LA). Theorem 4.(x)) where u ranges over all the discrete valuations of F over k.5. ord. b. y E F but $ k be g be the genus of F and let s be the number of x. a function field of one variable over an alof characteristic 0. 4 = = i log max (I-4. c) 5 C(F. Then hF(x) and h. IYL 1~1))~ whence we obtain a bound on n by Vojta’s conjecture. as follows. Let of distinct zeros and poles [Mas]. The Vojta conjecture also implies the abc conjecture quite generally.CK §41 BOUND FOR THE HEIGHT OF ALGEBRAIC POINTS 65 from the which is an algebraic point of degree 5 64. Then immediately definitions we have h(P) ih(x. namely It is easy to get an upper bound for the logarithmic d(P) = Ws max (Ixl. Ivl. See [Vo 871. Let F be a number field and let E > 0. y. In the function field case.6. PWc The idea of the proof is again to fix n = 5/s (or whatever). The height here is normalized so that hF(x) = 1 max (0. Let x. c E or with a -I. the result is a theorem of Mason who gives an explicit bound for the constant C(F. b. There exists a constant C(F. Y. discriminant.(y) 5 s + 2g . E). [Vo 871.2. and to consider the point (a l/ny bli”.+n) on the Fermat curve of degree n. . We also see how a problem which appeared isolated until recently now finds its place in the context of extensive structural theories in algebraic geometry mixed with number theory. Cf. This example shows how more sophisticated conjectures imply classical concrete questions concerning diophantine equations.

Here we see the same quantity 2g . Vojta’s conjecture 4. . . should also be improved.. Given E > 0.+u. . Let s be the number of distinct zeros and poles of uO. and let h(P) = C -min 0 (0. . .uN be rational functions on Y such that u.(P) 5 d(P) + ehE(P) + O. $3. Next we come to Vojta’s conjecture for higher dimensions. From his point of view. both he and I now believe this factor is unnecessary. .7 (Voloch [Vol 851).= 1. 1 whose elements are linearly dependent over k. But the factor N(N + 1)/2 is not the best possible one conjecturally. Let P = (uO. see Chapter IX.u. Mason’s theorem was extended to higher dimensions by Voloch as follows. . Let u 0. there exists a proper Zariski closed subset Z such that for all algebraic points P E X(Q”) . . . . Assume that there is no proper nonempty subset of uO. according to conjectures which I made long ago in connection with diophantine approximations. 98. For a discussion in the context of Schmidt’s theorem.(l).) be the corresponding point in projective space PN. . but for complementary reasons.Z we have h.2 + s). the part of the error term involving the ample height h. Furthermore. .+. analogous cases of Nevanlinna theory have been verified to hold without this factor. In his Lecture Notes [Vo 87) Vojta actually puts the factor dim X in front of d(P). Let X be a projective non-singular variety defined over a number field. From my point of view. so that the full .un. Theorem 4.66 HEIGHTS AND RATIONAL POINTS cw §41 Observe that Mason’s theorem concerns the unit equation if we take the point of view that x. $2 and [Vo 891.8. See also Brownawell-Masser [BM 861. y lie in the finitely generated group of units in the subring of F consisting of those functions which have poles only in a finite set of places. ord. i Then h(P) 5 $V(N + 1)(2g . . .un.. .2 + s which will reappear Chapter VI.(ui)). Let E be an ample or pseudo ample divisor. whatever reasons he gives in [Vo 871 actually do not apply. . Let Y be a complete non-singular curve of genus g. defined over an algebraically closed field k of characteristic 0.

which we shall discuss in Chapter VIII. then I conjectured that the exceptional set Z in Vojta’s inequality is the same as the special set defined in Chapter I. Thus Vojta’s conjecture gives a quantitative estimate for the heights of points in the case discussed qualitatively in Chapter I. or ample. Vojta also formulates another conjecture for coverings of X. Thus the only possibility to have infinitely many rational points when K is ample is when these points lie in a proper Zariski closed subset. Note that when the canonical class is negative. when K is ample. so we preferred to state the simplest case first. The nature of the exceptional set Z has to do with more qualitative conjectures. Suppose that K is ample or pseudo ample. Furthermore. and the Vojta conjecture implies that the set of rational points P over F which do not lie in the Zariski closed set Z has bounded height. then the term d(P) is bounded. or the conjectures in the compact case. to deal with quasi-projective varieties. If we look only at rational points P E X(F). or -K is effective or when -K is ample. the conjectures as we have stated them in one and higher dimension may be called the absolute conjectures.(P) 5 d(P) + f(1 + E) log h.CK 041 inequality BOUND FOR THE HEIGHT OF ALGEBRAIC POINTS 67 should read h. It is an important problem to determine the exceptional set 2. Then we may take E = K.(P) + O. $3. and we postpone the most general formulation until we discuss Weil functions in Chapter IX. namely the Zariski closure of the union of all images of non-constant rational maps of group varieties or abelian varieties into X. Furthermore. which contain another positive term on the left-hand side. then the inequality is vacuous. The inequality has content as stated only when the canonical class is effective. whence is finite.(l). . To define such a term requires other definitions. Vojta states generalizations. so with the non-compact case.

Let X be a projective non-singular variety over F.. and having good reduction at u. a description of the torsion. We describe the reduction somewhat non-invariantly by having assumed the projective imbedding. T.. then we say that X has good reduction at u. Let F be a field with a discrete valuation u. isomorphic to X’ over F. . Thus we collect such results in a separate chapter.. We first discuss a naive notion of reduction of X at u.CHAPTER III Abelian Varieties The presence of a group structure on a variety gives rise to numerous additional relations for the height.. If IkCVJis the prime in ideal defining a non-singular variety XkCVJover k(u). Then we can reduce the coefficients of polynomials in Z.[T. with a description of generators.]. . T. $0. be the ideal in o. Let k(u) be the residue class field. valuation ring D.. so we start with a summary of basic facts and terminology about such families in a fairly general context. If X = A is an .. III. BASIC FACTS ABl>UT FAMILIES AND NERON ALGEBRAIC MODELS We shall be dealing with algebraic families of abelian varieties and of curves. all emphasizing the group structure. We let I be the ideal in the projective coordinate ring F[T. . the group of rational points can be analyzed as a group. ZJ defining X over F. gives rise to a quadratic function associated with every divisor class. and maximal ideal m. mod m.‘. . . Furthermore. .] consisting of those polynomials which have coefficients in 0. we say that it has good reduction at v if there exists a projective variety X over F. . and in particular.. to get an ideal ZkCV) k(u) CT’. bounds for the heights of generators. . . and we let I. Given any variety X’ over F. . .

Let X be an abelian variety or a curve of genus 2 1. A is smooth over S.27~. over o we mean a group scheme A satisfying the following properties. NM 3. The generic fiber A. yj E o. where A is the discriminant. 3 and y2. NM 2. then it is a basic fact that for all but a finite number of the discrete valuation rings o. and the graph of the group law on Ak. Good reduction is expressed more invariantly in the language of schemes as follows. and X.x .). We can take this latter property as the definition. over k(v).. if the characteristic of k(v) is # 2.) which is smooth and proper. A) . NM 1. An early theorem of Chow-Lang asserts that if X’ is another projective variety over F isomorphic to X over F.“) is the reduction of the graph of the group law on A. X’ have good reduction at u. But often in practice a variety may be defined by actual equations. WI BASIC FACTS ABOUT ALGEBRAIC FAMILIES 69 abelian variety. the natural map Mor. Let A. By a N&on model of A. so we do want to know what reduction means in terms of such equations. as when we represent a curve by an equation y2 = x3 . Indeed.y. We shall usually say almost all instead of all but a finite number.(X. Then X. and XicO. a morphism X. 4) .. then the elliptic curve has good reduction if and only if A is a unit in o. A = 16(41/z . Let X.W. quite independent of a projective imbedding. For every smooth morphism X + S. If o is a Dedekind ring with infinitely many prime ideals. and has good reduction at u. and such that the generic fiber of this scheme is the given variety X. Let S = spec(o) where o is a Dedekind ring. with quotient field F. be an abelian variety over F. is the given abelian variety A.. a non-singular variety X over the quotient field F of o has good reduction at u.. again be a non-singular variety over F.y3 in affine coordinates.m. extends uniquely to a morphism X -+ A over S. has good reduction at v if and only if there exists a scheme X + spec(0. corresponding to these maximal ideals. then the isomorphism between X and X’ reduces to an isomorphism between Xkcv. In other words. then &) is then also an abelian variety.Mor. -+ A. Thus we may say that the good reduction is uniquely determined.

It is proper if and only if A. the Ntron model is not necessarily proper over a given discrete valuation ring 0. In particular. see [Art 863 and [BLR 901. has good (resp. semistable) reduction at every local ring o. and therefore any Weil divisor on A is locally principal. For proofs of facts concerning the N&-on model going beyond Net-on. Fix the discrete valuation ring o. We denote by A. Corollary 4. all the local rings of points on A are regular. (For good reduction see Chapter IV. we know that Ai is an extension of an abelian variety by a linear group. is an algebraic group over k. then we say that A. of o. extension over which it We shall now discuss the general notion of conductor for an abelian . If this linear group is a torus (i. has semistable reduction. The following are basic facts about good and semistable reduction. has semistable reduction at v if and only if Br has semistable reduction. Net-on [Ne 551 proved the existence of N&on models. Then: A. that is. Of course. A group scheme over spec(o) whose generic fiber is an abelian variety is proper over spec(o) if and only if the abelian variety has good reduction at all primes of o. then we say that A. If A. semistable) reduction over o or over spec(o) if A. If o is a Dedekind ring with quotient field F. has good reduction at u. see Grothendieck [Grot 703. we mean the open subgroup scheme of the N&on model whose fibers are the connected components of the N&on model. a product of multiplicative groups over k”) and so At is a semiabelian variety.) Let Br be an abelian variety over F.2. By the general structure theorem for group varieties. the special fiber over the residue class field k.e.70 ABELIAN VARIETIES cm §Ol and obtained by extending the base from S to F is a bijection. has semistable reduction then taking the connected Ntron model commutes with base change. In particular. hence an isomorphism of abelian groups. By the connected N6ron model. not necessarily connected. over F. has good reduction at o if and only if Br has good reduction. A. For a more recent discussion. and let k = k(u). has good (resp. Then A. isogenous to A.. Thus Ai is a group variety over k. Note that the smoothness assumption implies in particular that A is regular. or A. Grothendieck also proved the semistable reduction theorem: Given an abelian variety over F there is a jinite has semistable reduction. denoted by A’. and specifically involving semistable reduction. it is trivial from the definition that taking N&on models commutes with smooth base change.

and applications to other theories (for instance. B) . the association c?++u*c for tl E Hom(A. THE HEIGHT AS A QUADRATIC FUNCTION Let A. then u = t = 6 = 0. For a discussion of the discriminant and conductor of curves in general. and Raynaud [Ra 64-651 in higher dimensions.. B be abelian varieties. III. If A has good If A acquires degree prime d = dim A. which can also be described explicitly. $1.CIK VI THE HEIGHT AS A QUADRATIC FUNCTION 71 variety over F. As we know. See [Grot 721. Arakelov theory) see Saito [Sai. SGA 7. This occurs if p > 2d + 1. relations between divisor classes give rise to relations between their associated heights. At is an extension of an abelian variety by a linear group. We let: u = dimension t = dimension of the unipotent part of A. This description gets lengthy and technical. be the fiber of the N&on model over our and suppose for simplicity that k is perfect. 3 for elliptic curves. In particular for d = 1. which I got from Serre-Tate [SeT 681. semistable reduction over a Galois extension of F of to p. including genus bigger than 1.4+ t + 6. in terms of Artin and Swan conductors related to the wild ramification of 1. where 6 is an integer 2 0. and this linear group is an extension of a torus by what is called a unipotent group (a tower of extensions of additive groups). Let A.. then 6 = 0. The principal relation between divisor classes is that given c E Pit(B). Then as already mentioned. or reduction. In fact. Grothendieck proved the independence from 1. of the torus in A. 6 is defined for each prime 1. we see that the condition p > 2d + 1 is precisely the condition that p # 2. at the valuation u as in Then one defines the order of the conductor Serre-Tate [SeT 683. where if p = 0. 881. to be f(u) = 22. but it already gives insight to list the following properties satisfied by 6. For a systematic treatment of 6 see [Ogg 671 for elliptic curves.

.. by will be called the height pairing associated with the class c. = qC + 1. (in its class mod O(1)) by the limit q.. /I) is bilinear obtains [Ne 551: in (a.(P) = lim h.2. From this fundamental relation. Let c E CH’(A). for instance in [Dem 683 and [Zi 761 for elliptic curves. and is called the N&on-Tate height. that is (. such that h. which is uniquely determined by c will be denoted h. If c is ample. Q). and a linear form 1. Some authors normalize the bilinear form with an extra factor l/2 in front. = (a + /3)*c. 911 is quadratic in ~1. Estimates for the difference between the N&on-Tate height and the naive height obtained from a projective imbedding have been given. who expressed the height as a sum of local intersection numbers (but also used the limit argument locally). The bilinear form V’.(2mP)/22m. + O(1) us functions Zf c is euen. then I. the associated quadratic get a seminorm for form qC is semipositive. then its associated semipositive bilinear form A(F”) x A(F”) .l)*c = c.a*c . There exists a unique quudrutic form q. = 0.. m+m This was Tate’s fast way of getting the form. 8).1 (N&on-Tate). Then we P E A(F”)/A(F”). on A(F”). As an immediate consequence of the positivity properties of the height. one obtains N&on’s theorem: Theorem 1. The quadratic form can be obtained directly from a choice of h. In other word. The sum qC + I. replacing more elaborate arguments of N&on. Let A be an abelian variety dejined over a finite extension of a jield F with a proper set of absolute values sutisfying the product formula.h(Q) = <P. Q>. height pairing is a In particular.72 ABELIAN VARIETIES L-111. one Theorem 1.p*c 8) then D.R.(a.b’) . if we let Dc(a.fic(f’ + Q) .

Then we get a norm IPI. we have a conjecture of mine [La 781: Conjecture 1. with minimal discriminant AA. If c is ample. A fundamental problem is to find a basis for the lattice consisting of points of minimal height. then class induces sional vector subgroup of A(Q). Theorem 1. There exists an absolute constant C. and which concerns the possible lower bounds for the N&on-Tate height on non-torsion points. again by counting lattice points in a homogeneously expanding domain as in Proposition 2. Let A be an abelian variety defined over a number Jield. in the cases of number fields and function fields. = h. Here we mention a related question which also has independent interest. We start with the number field case. how to give an upper bound for the heights of points in such a basis. This norm extends to a norm on R @ A(F).(P)“2 for p E A(WWh.0'7 Q>c is the torsion field F. Q). the asymptotic formula for the number of points with height h=(P) 5 log B has the shape N(B) = y(log B)1” + O((log B)(‘-I)“) for some constant y. a non-torsion . We shall discuss the set of P for which IPI. is a homomorphism from Pit(A) into the group of real valued functions on A(Q) whose kernel is the group of torsion elements in Pit(A). Say on an elliptic curve. . Let c be an even ample class.3. then the kernel of the bilinear form U'. which we may call the N&on-Tate norm associated with c. One may ask for abelian varieties the same question about the asymptotic behavior of the number of points of bounded height. (i) (ii) The map c++h. We shall discuss this much deeper aspect in $4. = 0 below.3 of Chapter II. Zf A is defined over the number the Net-on-Tate height associated with an ample even a positive definite quadratic form on the finite dimenspace R Q A(F). Since the height here is logarithmic. The group of rational points modulo torsion can then be viewed as a lattice in R @ A(F) 511 THE HEIGHT AS A QUADRATIC FUNCTION 73 which we call the N&on-Tate seminorm associated with c. > 0 such that for all elliptic curves A over Q. and is a positive definite quadratic form. This kind of asymptotic behavior was already given by NCron [Ne 521.

containing an ample divisor and even. see the end of $5. Then we have the height h. For some results on lower bounds of heights elliptic curves and abelian varieties. We let be a morphism of projective varieties over a jield k. From [LaN 591 we get: Proposition 1. and we let P(Y) denote its Zariski closure in A. that is P .4 with upper bound problem. [Mass [Mass 851.. The notion corresponding to log [AAl in higher dimension the Faltings height. If P(Y) and Q(Y) are in the same algebraic family in A.3 then reads: Theorem 1. The analogue of the structure Theorem 1. Q are congruent modulo tB(k). be the F/k-trace of A. and we let F = k(Y) be the function Jield. for the generic point q of Y is an abelian variety. 3 of 01. For a discussion of the relation of Conjecture 1.. the is We now pass to the function field case. see Masser [Mass 811. Let P.74 point P E A(Q).(F) as a rational section P: Y-+A..5. Remark.. which will be defined in the next chapter. This section is a morphism on a non-empty Zariski open subset of Y.6.Q E zB(k). $5.. which is thus defined over the function field k(q) = k(Y) = F. We assume Y non-singular in codimension 1. abc on 841. then P. log IAA Hindry-Silverman showed that this conjecture is implied by the conjecture [His 883. Q be rational sections of 7~:A + Y. which we take to be algebraically closed for simplicity. Let z: B + A. Then for any point P E A(F) we have . defined as in our standard Example We suppose that the generic fiber n-‘(q) = A. satisJies ABELIAN VARIETIES cm §I1 h(P) L C.. Let c be a divisor class on the generic jiber A. We view a rational point P E A.

Let L. If we want the theorem to be formulated for all points P E A(Fa). then one must assume the stability of the trace. or Picard variety.(P) = 0 if and only if P E A(F). form The N&on-Tate height h. &JX) = f&(x.7 (1) (2) The height (x.e. 6) be the dual variety. and found that one gets certain classical lattices from the theory of linear algebraic groups. y) H f&(x. Let F’ be a finite extension of F.l]*(s) = 6. We let (A’. Y). Theorem 5. + zB(k). y) is a bilinear form on A(F”) For x E A(F”) and y E A’(F”) we have x A’(F”). we have the associated height &.c. Then it may happen that the F/k-trace is trivial but the F/k-trace is non-trivial. which can be normalized so that it is Z-valued. which we encountered in Chapter I. Theorem 1. b] has investigated this form and the lattices which arise from it. that is [ . erties of & are summarized in the next theorem.. Applications. Having described the quadratic and bilinear forms associated with a divisor class on an abelian 011 THE HEIGHT AS A QUADRATIC FUNCTION 75 h. extends to a positive dejinite quadratic on the jinite dimensional vector space Remark. So we let F again be a field with a proper set of absolute values satisfying the product formula. (3) Let c E Pit(A) and let (PE:A + A’ be the homomorphism such that cp. One may thus view R @ A(F) as a lattice in R’ (r = rank of A(F)). be the . we conclude this section with a description of another form arising from the dual variety. with such a form. Over a finite field k the group A(F) is finitely generated and modulo torsion admits the N&on-Tate positive definite quadratic form. Shioda [Shio 89a. that we have picked F sufficiently large so that the trace remains the same for finite extensions. The basic prop- Since 6 E Pic(A x A’). rational over the algebraic closure. It is a fact that 6 is even.(a) = element a’ E A’ such that ‘6(a’) = c.3.. and we let A be an abelian variety defined over F”. . i.

be a height of Y associated with a very ample divisor class on Y.k% %Y). ALGEBRAIC FAMILIES OF HEIGHTS The theorems in this section give some description how the height can vary in an algebraic family of abelian varieties. cm @I (4) If c is even..fiom ii.(k) we have the N&onTate height which is a quadratic function. We work under the following assumptions. Y) = . corresponding to this imbedding. Then .. varieties defined ouer a x-‘(q) = A. Let k be a jield of characteristic 0 with a proper set of absolute ualues satisfying the product formula. Let smooth ouer Y. Then we have associated a height function h. differing by a bounded function from the restriction of h.(u. is an Y0 be the non-empty Zariski open subset where II is for 011 y E Y.(k). which is given with a projective imbedding. k) be a divisor class.(u + u) . Ill.(u) . We shall give a bound in terms of the height on Y. . We are interested in seeing how this bound varies for YE Y. then We call h. c has a restriction to each fiber. to A. $2.16 bilinear form derived . Let k = k”.: A(k) + R well defined mod O(1).w. that is L. We let h.(k). the NBron height pairing on A x A’. On the other hand. such that the generic fiber Let c E Pic(A. so variety.. the fiber n-‘(y) = A. For each ye Y. which we denote by cy.h. Let n:A+Y be a jfat morphism finite extension of abelian variety. is an abelian of projective non-singular k. 0) = h.(k).&(“).

= Yl b> cl suck that [Sil 847). 1)) + YZ. There exist YZ = YAY. k) we huue + Y2 for all y E Y. where 0 is the zero section on an affine open subset of P’.(k) .1 (Silverman-Tate class. I.1 says that for all y = (a. Let b E Pic( Y) with deg b # 0. For the next theorems.1. k) be a diuisor function and associated with c.h((a. we shall also assume that the NCronGjeveri group of Y is cyclic.k(W). and h.. For abelian varieties. and is the most important case for applications.Theorem 2.1 is made only because so far the proof uses resolution of singularities. Then h. We let k. b as variables. b.(k) with P # 0 we have lky(P) . The restriction to characteristic 0 in Theorem 2. the height associated with a class algebraically equivalent to 0 has lower order of magnitude than the height .&(Y) Example. b. Y(P). Let c t Pic(A. the projective degree gives an imbedding deg: NS( Y) + Z. a choice of height numbers Y. This imbedding allows us to normalize heights on Y a little more than previously. This is the case when Y is a curve. and thus affine coordinates of P’. y2 are absolute constants.)1 i y. Let c be the divisor class of 3(O). Manin-Zarhin give an estimate of the same kind as in Theorem 2. 1) in P’(k) with 4a3 + 27b’ and all P E A. # 0 where y.. Theorem 2. Otherwise the arguments are rather formal.W)I 5 Y. viewing a. Under this hypothesis.=-~ ’ hb deg b By Theorem 3.5 of Chapter II.(!s) I&(P) and P E A.(P) = h((xv73 Y(P)> 1)). with respect to “canonical coordinates” [MaZ 721. Consider the 3-dimensional variety A defined on an affine open set in characteristic 0 by y2 = x3 + ax + h.

[n]P section. denoted by PH P. 2.. is independent of the choice of b with deg h # 0.2 (Silverman [Sil 841). If dim Y = 1. Thus we are making a more severe restriction than when dealing with rational sections earlier. because any rational map of a non-singular curve into a projective variety is a morphism. we shall deal with sections P:Y+A ofn. Therefore asymptotically. T) be the k(q)/k-trace of the generic fiber A.(kh)) + A.(k(d)/W4 is finitely generated. is a morphism since Y is assumed non-singular. Then We shall now give an application by Silverman of his height theorem. so the family splits. this is a relatively rare occurrence.. A.. (This condition is satisfied ij Y is a curve.(k). we shall write P. because for any other class b’ with deg h’ # 0.xY is a product. Theorem P: Y + A is also a c E Pic(A. but when A=A. In the higher dimensional case. Y is a curve. Let (B. WC have that deg(h’)h is algebraically equivalent to deg(b)b’. instead of P(y) for YE r.wume that NS(Y) is cyclic.(k). that is. This is a theorem of Weil. then again any rational map P: Y-A. .associated with an ample divisor. then any rational map P: Y-t A such that R o P = id generically is necessarily a section in our restricted sense (morphism). To suggest families of points. by which we shall means morphism such that K o P = id. we know that 4. morphism For each y E Y. Let be a section such that for arbitrarily large integers n.(k) we have the specialization homo- q: A.) Let k). By the Lang-Niron theorem. For the next theorems.

We let be the N&m model. In particular.2. [Man 691: Corollary 2..) = 0. But Silverman actually proves that this set is finite in fields of bounded degree. and $ rank r > rank A. If NS( Y) is cyclic. As already mentioned in Chapter I.3. Lrt Y he a projective non-singular variety ouer F. be an abelian variety over F. and let A.(k).3. one then obtains a result of Demjanenko-Manin [De 661. We want how often this specialization homomorphism is not injective. if’ k = Q”. x Y. Let Y be a non-singular cum ouer k.(k) of hounded degree over Q such that CS~ nor injective is on 1-. wer Y in light of the uniqueness of a N&on model on the intersection of the two affine open subsets. Let r be a finitely generated free subgroup of A. and follows by applying that theorem to A = A.(k) such that Ok is not injectiue on r has bounded height in Y. We can cover Y by two (or whatever) atline open subsets which are specs of Dedekind rings. we can then define a N&on model of A. $7 Nbon had obtained a weaker specialization theorem using the Hilbert irreducibility theorem. and let yO E Y(F) be an F-rational point. and A0 the connected N&on model. Let r he the group of morphism f: Y + A. Let A. such that . and that rational sections are morphism. As a corollary of Theorem 2.(F) then Y(F) is finite.under the assumption that sections are morphisms. We shall now obtain a refinement of Theorem 2. The corollary is essentially the split case of Theorem 2.f(y. to know Theorem 2. with I(: A + Y being the projection. In [La 83a] I .. Since we defined a N&ron model over a Dedeking ring. to show that the set of points where the specialization homomorphism gP is not injective is thinly distributed. A SSUWI~ that NS(Y) is cyclic.(k(q)) which injects in the quotient Then the set of ye Y. there is only a finite number (f points y E Y.3 (Silverman [Sil 841). Let F = k(Y). Let k = F”. be an ah&an variety ouer a number jield F.4.

transposed to varieties of dimension 2 1 a theorem of Tate [Ta 831 concerning heights on a family of elliptic curves. (For a treatment of related results depending on moduli spaces and Faltings techniques, see Green [Gr,,, 891.) In the proofs I gave in [La 83a] I assumed the conjectured existence of a good completion for the Nt%x~ model. However, Chai has pointed out to me that one could argue in essentially the same way without using this still unproved existence, by making use of other technical means, namely cubical sheaves on the N&n model, thereby proving the results unconditionally. I shall therefore follow Chai’s suggestion. If P: Y + A is a section, we let r,:A+A be the corresponding translation. Let 0: Y + A be the zero section and P: Y + A any section. trivially 70 0 ip = Tp 0 To = Tp and P = TTp 0. 0 The first relation is between automorphisms tween maps of Y into A. To simplify the notation, let us abbreviate as follows:

We have

of A, and the second formal


linear combinations

ITP, T*l = T.P+cJ TP- tp + 70, CP> = (f’ + Q, - (0 - (Q, + 0. Ql
Then for instance if c E Pit(A), [P, Q]*(c) we have by definition

= (P + Q)*(c) - P*c - Q*c + O*c,

and similarly for [ip, T,]*c (inverse image). These two inverse images are related in a simple way, if we look only at the part in the connected N&on model. Proposition 2.5. Let cy E Pic(A,). Then there exists an extension cF to an element of Pic(A”) which satisfies the .following property. P, Q E Sec(T, A’) we have [rp, ta]*c = n*[P, Q]*c. c of For

This is a variation of Proposition 5.1 of [La 83a] due to Chai. As he pointed out, this reformulation of the proposition follows as a direct consequence of Breen [Br 831, $3, pp. 30-31, 01 also of Moret-Bailly [MorB 851, Chapter II, Theorem 1.1, p. 40. Indeed, by those references, a line sheaf S?F on A, extends uniquely to a cubical sheaf 9 on A’. The

essential part of the definition of cubical theorem of the cube, i.e. the line sheaf $$ (m:9)(m”“’ on

sheaf is that 9



A0 xI. A0 xT A”

is trivial, where the tensor product is taken over all subsets I of {I, 2, 3}, and m,: A0 xr A” xr A” + A” is the addition taken over the factors in I. Pulling back to A0 xr T X~ T = A0 via idAo x P x Q we get Proposition

Thus WC call a class c E Pic(A”) cubical if it satisfies the formula of Proposition 2.5. The existence of such a cubical class is all that is needed to deduce the following result, as in [La 83a], which is the higher dimcnsional version of T&e’s theorem for elliptic curves. Theorem 2.6. P E Sec(Y, A”).
restriction qf points Let Let of c to thr y such that ?I: A + Y he the N&-on model of A,. Let c E Pic(A’) he cuhicul, and let cy = CIA, he the jiber A, for y E Y,(k), where Y, is the open subset A is proper ouer y. Let: height quadrutic on A,(k); part

h, = hey = N&on-Tute qy = homogeneous Then for y E Y,(k) we have &YY)) = fh,~,~,.cb)

of A,.

+ ‘Ml),

Remark 1. Note that [P. P]*c E Pit(Y), whence the height hlp,pl*L makes sensesince Y is a complete non-singular curve. Remark 2. Since [ - I],, is an automorphism of the N&on model, one can write 2c as a sum of an even class and an odd class, and one can therefore catch the linear part of the height similarly. In the version of [La 83~11this posed a difficulty since one could not necessarily extend [- l] as an automorphism of the good completion. From the theorem, one gets a factorization families. of the height in algebraic

Corollary 2.7 (Silverman’s conjecture). Suppose Y = P’ is the projective line, und h, is the standard height on P’. Let 7 he the genericapoinl <f Y, and let q,, he the quadratic form of the N&on-Tate height h, on the generic fiber. For P E Sec(T, A) (and not necessarily in Sec(T, A’)), P fixed and y E Y,(k) variable, we hum &‘(Y)) = q,W,W + O,(l)

Corollary 2.8. Let Y be an arbitrary Let a E Pic( Y, k) have degree 1. Then &‘(y)) = q&%(y)



of genus 2 1.

+ 4hhP)

+ 4(l).

Both corollaries follow as in Tate and [La 83a], using Chai’s versions of the preceding results, given here as Proposition 2.5 and Theorem 2.6. Hence whereas the corollaries were proved before in dimension > 1 only conditionally by the use of good completions for the N&on model, they are now proved unconditionally.






The group of torsion points on an abelian variety comes in the theory in several ways. First, one wants to determine the group of torsion points rational over a given field, for its own sake. But in addition, the torsion points give rise to representation spaces for the ring of endomorphisms and for the G&is group, thus intervening in a much more extensive way in the general diophantine study of abelian varieties. We shall collect here some basic facts which will be used in this chapter and even more importantly in the next chapter. In light of this multiple use, the present section will provide a common reference for those facts. Classically, if A is an abelian variety of dimension d, defined over the complex numbers, then A is a compact complex Lie group, isomorphic to a complex torus, and thus for each positive integer WI, if we let A[m] be the subgroup of points of order m, then A[m] zz (Z/~IZ)*~ as abelian groups.

By a theorem of Hasse in dimension 1 and Weil in arbitrary dimension, the same structure holds for all abelian varieties, provided m is prime to the characteristic of a field of definition. We assumethroughout that A is defined owr a jield k and that 1 is a prime numbrr prime to the characteristic of k. We define the Tate module 7;(A) to consist of all sequences

of points a;~ A(k”) such that la, = 0 and lai+l = ai. Then T,(A) is a module over Z,, since for every p-adic number c E Z, we define its action Z,, Z, c(a,,a, ,... )=(ca,,cu, )=(ca,,cu, c(a,,a, ,... ,.__ ).

A basic theorem

of Weil (Hasse-Dewing

in dimension

1) asserts:

Let d = dim A. Then T(A) is a We can identify 7;/1”T = A(k’)[f”]

free module of rank 2d ouer 2,
= A[I”]. We can also write

7;(A) = l&l A[[“], where the limit is the projective limit. We shall also consider the Q,-vector space UA) = QI 0 T(A) = Q,W), which is then a 2d-dimensional vector space over Q,. Let Gk be the Galois group of k” over k. It is a basic fact that the extensions k(A[l”]) of k are separable and so Galois. Then T,(A) is a representation module for Gk, by defining for each LS G, the action E componentwise: +,,a* (...) = (UUl,mz2, ...). Similarly, if f: A --t B is a homomorphism over k, then ‘J(A) gives the representation

We let

such that


,... )=(f(a,),f(a,)

,... ).


rib4 + K(B)

be the natural extension to the vector spaces over Q,. Since dim v(:(A) = 2d, after picking a basis, we get corresponding representations in GL,,(ZJ and GL,,(Q,). All these representations are called the I-adic representations associated with A over k. Suppose next that k = F is the quotient field of a discrete valuation ring 0,. Let w be a valuation of F” extending v. By the decomposition group G, of w we mean the subgroup of GP consisting of all elements such that n+. = w. Then G, can be identified with the Galois group of the completion, that is G, = Gal(F;/F,). If w, w’ are two absolute values of Fa extending u, then w, w’ are conjugate (i.e. there is an element of GF sending w to w’) so that the decomposition groups G, and G,, are conjugate subgroups of G,. Let k(v) and k(w) be the residue class fields of o,, and the COTresponding valuation ring (not discrete) o, in F”. Then there is a natural homomorphism G, + Gal(k(w)/k(u))





which is easily shown to be surjective. The kernel of this homomorphism is defined to be the inertia group, denoted by I,. Thus the inertia group is the subgroup of G, inducing the identity on the residue class field extension. Again, any two inertia groups I,, I, are conjugate. Suppose that k(v) is a finite field with y elements. Then there exists a unique element o E Gal(k(w)/k(u)) such that gx = 9 for x E k(w).

This element is called the Frobenius automorphism. Then there exists a unique element, denoted by Fr,, in G,/l, such that Fr, induces the Frobenius automorphism on k(w). This element Fr, is also called the Frobenius element. In general, it is well defined only module the inertia group. Once more, Frobenius elements Fr, and Fr,. are conjugate. In light of these conjugacies, we frequently denote the decomposition group and the Frobenius element by G, and Fr, respectively, because we are interested in their properties only up to conjugacy. In particular, the characteristic polynomial of an element in a finite dimensional representation of the Gal& group depends only on the conjugacy class. Next, suppose that A, is defined over the field F, quotient field of the discrete valuation ring 0,. Suppose that A, has good reduction. We let A be the proper smooth model over spec(o,) whose generic fiber is A,. Let w be an extension of u to F” as above. Then we have a natural homomorphism W”) + A(W), and therefore a natural homomorphism on the I-adic spaces

commuting with theorem that:

the action

of the decomposition


G,. It is a basic

Assuming I not equal to the characteristic of k(u), then the extensions F(A[l”]) are unramified for all n, and the natural reduction homomorphism gives G,-isomorphisms

Let k be a finite field with q elements, and let A be an abelian variety defined over k and of dimension d. For each positive integer m there is a unique extension k, of k of degree m. The set of points A(k,) is finite. The Frobenius automorphism of k” which sends XHX~ for x E k” induces an isogeny Fr,: A -+ A, the Frobenius isogeny, whence an automorphism Fr,: A(k’) + A(ka)

called the Frobenius automorphism. In fact, if Fr E G, is the Frobenius element in the Galois group, then the representation V,(Fr) on V,(A) is the same element as the representation of the Frobenius isogeny on V,(A). Let a,, . . ..qI1 be the roots of the characteristic polynomial of &(Fr). Then #A(k,) = fi (1 - a,“) = degree of (id - Fr,“) i=,

because A(k) is characterized as the set of points in A(k”) which are fixed under the Frobenius automorphism, so A(k,) is the kernel of id ~ Fr;. We have: The numbers mi are algehruic and hove absolute ualue q’12. The characteristic polynomial n (T - q) has coeficients in Z, and is independent of I. These properties are due to Has% for d = 1 and to Weil in general.






By an infinite descent one means a procedure by which, supposing given a rational point with some height, one finds another point with smaller height. The iteration of this procedure either shows that there was no rational point to start with, or ends up providing only a finite number, or a finite number of generators for the group of rational points if a group structure is involved. Let A he an abelian variety mm a number field F, The infinite descent is most classically applied as follows to prove the Mordell-Weil theorem of Chapter I, Theorem 4.1. Let m be an integer 2 2. One proves first that A(F)/mA(F) is finite. The proof effectively bounds the number of generators, but gives no indication as to their possible heights. To obtain a finite number of generators, one then has the following result whose proof is one version of descent. Proposition 4.1. Let r be an abelian group such that r/mr is finite. Suppose yioen a real valued norm / ) on r. Let a,, .,a, be coset representatives of ~/WC in r. There exists a number c, and a subset B of I- such that: (1) IPI < c, for all P E B, i.e. B is a bounded. (2) For any POE r, there exist integers n,, n,, ..n, and a point P E B

= P..+.lPol+l: ( which concludes the proof. k).+. and in the number field case.. Iterating IPljnf. and such that mP. We start with a discussion of non-singular varieties over a field k which become isomorphic to an abelian variety over a finite extension of k. we get of the elements a. We denote by PHS(A. Then X is a principal homogeneous space of an abelian variety A over k.+. then X is isomorphic to A over k. B is therefore a finite set. > . E A(@) such .) in r by starting where c is a bound for the norms this estimate. and such that for each x E X the map a-ax is an &morphism A with X. Theorem 4. with our point PO. PI. Let X be a principal homogeneous space of A ouer k.UK §41 such that Proof. We construct a sequence of points (PO. k) the isomorphism classes of principal homogeneous spaces of A over k. For each point x E X(k”) and LSE G. WC must now assume that the reader is acquainted with basic facts about the cohomology of groups. Then the set B in the proposition is a set of bounded height. . But the method does not bound the heights of generators. .. 1+. which is an action of A on X. Let X be such a variety..a. We remind the reader that k’ denotes the separable closure of k. but not otherwise. We apply the proposition when r = A(F). Thus A is the Albanese variety of X.2. . By this we mean that there exists a morphism defined cwer k.Ui” By hypothesis. but which may not have any rational point over k. . If X has a rational point. We shall now describe another type of descent. . there is a unique a.. thus showing that A(F) is finitely generated.. more sophisticated. with the seminorm equal to the square root of a N&on-T&e height (quadratic form) associated with an ample divisor class. We then have the following cohomological description of PHS(A.

. Weil defined the group law in geometric terms [We 551. 0 + A[m] + A(P) + A(k’) gives rise to the cohomology in the form O+ A(k)/mA(k) -H’(G. Schmidt. We now come to the study of descent in a cohomological context.. who over a finite field has a rational point. which al- . whose + 0 beginning can be rewritten A[m]) --t H’(G. then H’(G. In particular. Perhaps the first result of this showed that a curve of genus 1 and so is an elliptic curve in our Let m be a positive integer 2 The short exact sequence kind was due to F. Convenient references for proofs are [Ta 741 and [Sil 867. X) is trivial. sequence.. for every group variety (not even necessarily commutative) one can define a principal homogeneous space and a first cohomology set as we have done above..$eld and X is a group variety.. 2. A) = 0. A(k”))[m] +O See [LaT 581 for this as well as a general discussion of principal homogeneous spaces over abelian varieties. A(P)) group. For a comprehens&c account. The determination of this group is one of the standard diophantine problems. A(k”)) = H’(G. If k is LI finite . if X = A is an ah&m variety.. and prime to the characteristic of k.3 ([La 56~1). One has: Theorem 4. In particular. A principal homogeneous space of a group variety ouer (I finite field always has a rationul point. Chit&t originated these constructions in special cases. sense. is a I-cocycle. then H‘(G. the principal We define the ChBtelet-Weil WC(A) k) + H’(G. but this is irrelevant for us here. The association XH {a*) in- PHS(A. A) = H’(G. homogeneous spaces form an abelian group to be = H’(G. I refer to Colliot-Th&l&ne’s survey of Ch&let’s works [COT 8X].. Tate [Ta 581 gave precise duality theorems over p-adic fields in this connection.. K. A).UK 541 that PRINCIPAL HOMOGENE”“S SPACES 87 and the function duces a hijection: o-a. Actually.

because we are going to refine these exact sequences by using the absolute values. ACql) -+ ~‘(G. In particular. A)[cQ] +O. 0 + . We also generalize them by considering an arbitrary isogeny ‘p: A + B defined over F. we assume that k = F is a number field.4[q] --t A(F”) + B(F”) -+ 0 ES 2. 2. such that (TO= v. will also be viewed as a group of automorphisms of . of G. Then u on F” is induced by an imbedding We defined the decomposition group to be the subgroup G. and hence to an exact and commutative diagram: The two groups S’@(AF) and III(&) are by definition the kernels of the central and right vertical maps.4(Fe) and B(F.P) give rise to restriction homomorphisms on the cohomology groups. Then G.B). 0 + B(F. and are called the Selmer group and Shafarevich-Tate groups respectively. c G. It is easy to see that they do not depend on the choice of extension of u to the algebraic closure.WJ + H’G.)/rp. we obtain the local exact sequence: ES. consisting of those elements o E G. and A(P) c A(F.. For simplicity. 4Cvl -+a The natural inclusions G. Then we obtain the two exact sequences: ES 1.88 ABELIAN VARIETIES UK WI though written for elliptic curves. A[rp]) -+ H’(G.. apply without change to abelian varieties. For each absolute value u on F denote by the same letter u an extension to the algebraic closure F”. but . 0 -+ B(F)/c+A(F) + H’(G.

Under the jiniteness of III (Conjecture the two group III(A) and UI(A’) are dual to each other. The injection S: B(F)/rp(A(F)) is given by the coboundary operator -+ S’-‘(A. We write A”“’ for the Selmer group in that case. . (multiplication by m on A). Theorem 4. Next WC describe the role that Ill plays in making infinite descents. + S@‘(A. but does not have a rational point in Q. Then the finiteness of #“‘(A. The direction of Kolyvagin and Rubin gives one connection between diophantine problems and the theory of cyclotomic fields.6 gives a generalization of the weak MordellWeil theorem. spaces. Cassels [Cas 621 for elliptic curves followed by Tate for abelian varieties [Ta 621 defined a natural pairing III(A) x IU(A’) + Q/Z. Further insight was given by Kolyvagin [Koly]. we can describe III by saying: in terms of the homogeneous The group III is the subgroup of principal homogeneous ouer F which have a rational point in F. it follows that #UI(A) is a perfect square (if finite). so A is an elliptic curve. III(A) and IU(A’) can be identified. By definition. take for ‘p the isogeny m. Rubin [Ru 871 gave the first examples when it was proved that a ShafarevichbTate group is finite. When A has dimension I.4) See also [Gas 651. Extensive computations have shown this to be true experimentally. for all v.4.) implies the finiteness of A(F)/mA(F). for all v.) Observe that Theorem 4. then A is self-dual. the Selmer group S’“‘(A. Let ‘p: A t B be an isogeny of’ abelian mrieties Then we have an exuct sequence 0 --t B(F)/q+i(F)) Furthermore.) -+ IU(A. Theorem 4. The curve 3x3 + 4y” + 52-’ = 0 has a rational point in Q. spaces of’ A Example (Selmer [Se1 511).) from the long cohomology sequence. + 0. Indeed. over F. Conjecture 4.6 follows immediately from ES 3 and the snake lemma. and because of a natural skew-symmetric non-degenerate form on Ill(A).6. The Shaj&reuich-Tate group III is finite.)[q] is finite. Let A’ be the dual (Picard) variety of A. 4. Theorem 4.5 (Cassels~~ Tat@.depend only on A and F.

namely The finite group P)(A. $“. but the problem lies with UI(A..)[m] + 0..4 that Ill is finite. .) We reach the fundamental problem of finding generators have on the one hand the decreasing Selmer groups: $+“(A. The exact sequences with m and with m” fit into a commutative exact diagram -L+‘“)(AJ III(A.) is effectively computable. Then for some n the last term in ES 4 is 0. The infinite descent takes place by using powers of WI. we obtain the exact sequence + m”-‘UI(A.““A.(AF) where Rc.) 2 $“~Z)(A.) groups c .I. 47 1 S’“‘(A.) is effectively 4 Scm.) is the subgroup with height of #“‘(A.) generated by all points P E A(F) h(P) 5 s.) + uI(A.)[m”] + 0. “(A. we have the increasing Ro~.) = &n. The height is the N&on-T& height associated with an ample even divisor class. 0 --t A(F)/mA(F) + S’“‘(A.“)(AF) this allows us to find generators computable.” =.) = R&A. so we get an isomorphism A(F)/mA(F) Since S’“+‘(A..) z..) to define: in Scm’(A. = image of S’m”‘(A.) in the above diagram. Assume Conjecture 4.(A. We ImP.. 0 + A(F)/mA(F) + S’-“‘(A. It is now convenient S’ms”‘(A.) Then directly ES 4.)[m”] 0 49 idJ. and on the other hand.)[m].id W%)Cml - 0.) from the definition.) = 3’“. of A(F). and so a positive definite quadratic form on the group of rational points.(A.We rewrite the exact sequence with m for the applications.

We let one of the that A has we let: roots of Fr.) also would give a bound for the height of generators of A(F). and all u where A does not have good reduction.4. and Nu = #k(v). Suppose good reduction nt I). the determination of the bound s for the height of points generating P+)(. In particular. Ill.). guarantees and In any case. “8 . The Euler product converges for Re(s) > 3/2 because it is dominated (MS . if the finiteness of III would give an effective bound (to be determined in each case) for the heights of generators of A(F). = Frobenius element in G.yA. acting on A(k(u)B). i=l Let S be a finite set of absolute values on F containing all archimedean ones. the local ring at discrete valuations.for A(F).2d) = characteristic F... and k(u) the residue class field. However.” (i = 1.)‘: by the product for .T).t. Fr.) number of steps one obtains an equality S’“%b) that after a finite = R. Alternatively. one could take for m the order of III&) then get the easier isomorphism A(FymA(F) : S’“. We let G. to my knowledge. the zeta function. To give such estimates one has to dig deeper into the BirchbSwinnerton-Dyer conjecture. the finiteness of III(a. P”(T) = fi (1 . the procedure does not give a closed form for estimates of such heights. D. THE BIRCH-SWINNERTON-DYER CONJECTURE Next let A be an abelian variety defined over a number field o = oF be the ring of algebraic integers.. be a decomposition group. 65. It is not even clear from the literature. rxi.. and their associated .

such that for all but a finite number of v the ring of local integers 0. Compare with [ShT 671. The first Birch-Swinnerton-Dyer conjecture is: Conjecture 5. Ulmer [Ul 901 has constructed elliptic curves over a rational function field with finite constant field. see Serre [Ser 69-701 and the subsequent article by Deligne [De 701. for v E StiO.> has measure ~~(0.. and the finiteness of the Brauer group which plays an analogous role to the Shafarevich-Tate group would imply the rest of the conjectures in the function field case. The second BirchbSwinnerton-Dyer conjecture concerns this constant..(S)(s Then L.. Let r be the rank of A(F).. We must introduce other factors to get some constant independent of S. and satisfies a simple functional equation. Then conjecturally. We shall not go any further into this case.1. would also have such an analytic continuation.. to form a complete L-function LA(S) = n 1 u +s.s has a zero of . is a locally compact field.. except for indicating the following application. One can also define polynomials P. they hang together more tightly than in the number field case.(s) = C. we have . Then for x E r. Although these conjectures are not proved today.(Nu-7 We shall give the factors for bad reduction below.where & is the Dcdekind zeta function of F. such that the L-function of the elliptic curve vanishes of arbitrarily high order. The BirchSwinnerton-Dyer conjecture would imply that these curves have arbitrarily high rank. but no construction is known today exhibiting explicitly independent points having this rank.(S) # 0. LA has an analytic continuation to the whole plane. In particular.) = 1. and every open set U of K. and to formulate a coherent set of conjectures for arbitrary complete non-singular surfaces over finite fields. At lirst we follow Tate’s presentation in [Ta 66b] to avoid having to define the bad factors for v E S by using another device as follows. on F. One of the advantages of T&e’s formulation is that it allowed him to transpose the whole set up of the conjecture to the function field case. order I at s = 1. Thus L. At each absolute value u the completion F. however. P. has the Taylor expansion L.1)’ + higher terms: with some constant C.. LA. For each v we choose a Haar measure pc.

$2.. for S = SO. is the normalization of the absolute value described in Chapter II. We define the regulator R./F is compact.” By definition A. The measures vu define a measure ~=fl~. ..> We define the norm 11~11 P&IF) = It is easy to show that there exists a set SO (depending on A.. (. Y and hence on the factor group A.. = P. We now define onA.} is a basis of A(F) is a . the function L: is independent of the choice of w and p.1 we get L$&) = C..&’ on A(F.) is a compact analytic group over F. ..(Nu-‘) for v+$s. we have n. for all v. w. Choose an invariant differential form w of degree d on A over F. where Cz is independent of S = S.1)’ + higher order terms.. Pi} where {P.*(s . module torsion. consists of Let A. p) such that for all finite sets S containing S. . determine a Haar measure lol.. as s -) 1 is independent of Consequently the asymptotic behavior of L:(s) S.__. and using Conjecture 4. The group A(F. M”P. = Id&(& ?)I. (( ) indexed by the absolute values of F. . P. Then o and p. Then the field F is imbedded discretely on the diagonal in A. Define the u-adic period X” = s . . and x0 E F.where IIxiI..). Furthermore. be the additive vectors group of ad&s./F. {Pi. such that for all but a finite number of D. 3X”. and it is a basic theorem that A..W”. the u-component x0 lies in the local ring O&.

inducing the identity on the residue class field extension at v. be the Frobenius element in G. Let A be the Niron model of A. Hence I shall also describe the more precise version worked out by Gross [Gross 821 for abelian varieties over any number field. P’) (P. be the Gaiois group of Q” over F and let 1. The constant C. over L+. The above form of the conjecture is convenient because it did not force a consideration of the factors in the L-function corresponding to the bad places c. be the inertia subgroup of G. On the other hand. we define Thus c. Let G.2. For u finite. P’) is the N&on pairing of Theorem 1. namely an explicit value for 1 I! L$‘(l) ’ which is used to estimate the heights for the points in a basis of the Mordell-Weil group.5. it does not give something else we want. and as before let 7.. with connected Nltron model A’./I.(A) = lip A(Q”)[I”].* has the value where III is the Shafarevich-Tate group discussed in $4.. We then define GJi” = n C”. and = &(P.J. Let 1 be a rational prime unequal to the characteristic of k(u).. ” cSa. equal to the index of the subgroup of points in the r&due class field on the connected component of the special fiber. As before. let Fr. . in the full group of k(c)-rational points on the whole fiber of the N&on model. is an integer.basis of A’(F) module torsion. We then have: Conjecture 5.

.} be a basis and define A..“(S). .where A(Q”)[I”] is the kernel of multiplication by I” on the group of algebraic points of A.(7. be the absolute value of the discriminant of F over Q. The superscript I. and define The determinant of the 2d x 2d matrix is non-zero and depends only on v> 0. . I Then A WA = v. Ii Let W’.. Let H’ denote the submodule of H.(A. a”) and let ?j = A q. in G. where a.. Let (al.)‘“)-‘. Let H = H. Let D. including those with bad reduction.(C). Then H ’ is free of rank d.@‘). so z(A) is a free Z. .. We define the local L-factor of A at I) by the formula L..’ IHomz.. Then rank(W.} be an F-basis of #(A.. a”). denote the projective o. Z.__. Let u be a real place of F corresponding to the imbedding g: F + R. Thus we have defined the local factor for all t). in H”(A. . is a fractional ideal in F.4 = n L. and LA.No-“Fr.-module of rank 2d. Let {yI.-module of invariant differentials on the N&on model A. while is the real connected component. .(A. The . Z) which is fixed by complex conjugation.(s) = det(id . Let {q.q.(C). .. Let D be a complex place of F and let o: F + C be an imbedding inducing u on F. indicates the submodule of elements fixed by all elements of the inertia group I.yzd} be a basis of H..o.(R)” Again the determinant is non-zero and depends only on q. Then H is a free Z-module of rank 2d.) = d and A”“” W’ is a submodule of rank I of H”(AF. Z) be the integral homology of A. o. ..(A).

The discriminant is given The origin of the group law is at infinity in PI. . by A = 16(4y.y+ with Y*. Suppose p is a prime such that p4 divides yz and pb divides y3. which we shall now describe in almost complete generality. y3 E Q.. y3 E Z.96 ABELlAN “ARlETlES CIK WI product is independent of the choice of q.3. we can always achieve that y2. Finally we define Then C. and we can formulate the precise value of the constant in BirchSwinnerton-Dyer: Conjecture 5. We now have all the local factors defined. over Q for elliptic curves in that form is given by a c # 0 with the following effects on alike coordinates yccm3y. An isomorphism rational number (2 = 1): xwc-2x. is a positive real number. letting Yz HP-‘?* and h-P-% . Under these circumstances. $6. = o. Y2++ ce5 > y3 H c-by3 By such an isomorphism. We suppose that the elliptic curve is defined by the homogeneous polynomial in P’: $2 = 2 .27~:). there exists a “minimal” differential form w. THE CASE OVER Q OF ELLIPTIC CURVES Both the fact that we deal with an abelian variety of dimension 1 and that we are over the ration& is significant for the structure of the L-function. Then we can change the elliptic curve by an isomorphism using p = c. We haoe Ill.yg2 .

Letting A be this minimal model. = 1 + p . = I.. one then has (see [Art 861): The Nt%m model A is the open subscheme regular points.(. The L-function is then defined by us)=np. following Dewing [Deu 411. of A consisting qf the With respect to the minima1 model.After having is a minimal discriminant. we define the period at infinity The factors for the L-function can be described explicitly in the present case. If n(p) denotes the number of irreducible components of A.t.f(p) is 0 if p[A and 2 1 if pIA.. then 1.(number of points of A. For each prime p let A.p-“) &a 1 ’ n I t. To describe the fully genera1 situation including the primes 2 and 3 requires longer formulas which the reader will find in [Ta 741. in k(p)) If A has good reduction at p. over the algebraic closure of . On the other hand. be the reduction of the minimal model for A mod p.. We let dx ““=G For p # 2. = c(~ + a. a factor of curves over done so repeatedly until no further possible. or 0. if pIA. 3 we can then reduce the equation of the minimal model mod p.p-’ + p’-ZS ways. . and then A is called a minimal denoted by A”. and is a positive integer where the exponent .. It is a measure of We can define the conductor NA in various bad reduction. is the sum of the characteristic roots.A(1 . and is an invariant of an isomorphism class of elliptic Q.(.1. we obtain what model for the curve over Z. then t. Let t. The cwve has good reduction at p if and only if p j AA.. This minima1 discriminant is defined up to + I.

(s) and satisfies a functional equation L(s) = cL(2 . f(p) = ord.(A) + 1 . As before we let R. Since dim R = 1. Let &&) = Jv. multiplicative reduction.. be the regulator. Then the Birch-Swinnerton-Dyer conjecture as formulated in [Ta 741 is: We want to estimate +VUR. A”(FJ). or additive reduction. can be small. occurs here much as the product of the class number and regulator occur together in the residue of the Dedekind zeta function of a number field. already mentioned at the end of Chapter III. cp = (W. and if (P. and a lower bound must be determined. The conjecture is a precise version of a conjecture of Hasse. having C~integers helps us. y3 in the minimal model. with E= *1. and as before. However. _.1 of Chapter II. If pIA. Theorem 2. where ‘1 = 0.4 is h&morphic in the whole s-plane. 1 or 2 according as A has good reduction.n(p) = v + S. The function t. the elliptic curve is self-dual. q)l. . the period c. Conjecture 6. 3 and p is not a common factor of y2. but p # 2.1. and 6 is the integer defined by means of representation theoretic formulas like Artin and Swan conductors.P. then RA = Idet(P. We call E the sign of the functional equation..‘2(2n)~“r(s)L”(s). So J(p) has a tendency to be equal to 1 for pIA.): so cp is an integer for each pIAAA. We let A0 be the connected N&n model. then according to [Ogg 671. then f(p) = 1. All in all.k(p). so we don’t need to know more about them. and for this purpose. the product #(IUJR.} is a basis for a(Q) mod torsion. Cf. from above. $1. .

I arrived at the conjecture [La 83b]: Conjecture 6. a conjecture of Montgomery for the leading coefficient of Dirichlet L-series. Let H(A) = max(ly. gives a bound for the product of the heights &p... In fact.3.) << H(A)“‘Z*NE’N’(log h(P..(log N log log N)-“’ Since #(III..} for A(Q) sion.} is almost orthogonalized. . E(N) may have the explicit form c(N) = b. . Here we set h(P) = (P.y2x . such that: h(P.)R. the conjectured bound applies just to the regulator R. . Iy31”). There exists a basis {PI.7 and Corollary 7.P.P. and E(N)+O N + m.(3. Chapter 5. . Then #(IU. by a general theorem of Hermite.P...8 for precise statements and proofs.)R. Conjecture 6. mod& tor- where the constant implicit in CC. I conjectured that one can estimate #(UI.). Suppose we have a minimal equation yz = x3 . ordered by ascending height. (See [La 83a].)-. ore absolute. .) for A(Q) module torsion.) As a result.H(A)“‘2Nc’N’b.. we cannot quite find an orthogonal&d basis for this lattice.) of points in a basis {P..) is an integer.By analogy with estimates for the residue of the Dedekind zeta function of number fields due to Landau. and c.) if {P. arxd N = NA.e. The point is that the regulator is a determinant. as follows.h(P.. Theorem 7. b. 5 b.. are universal constants independent of A. independent of A. vector space Since A(Q)IA(Q). The bound for the smallest and a bound for R. P). but we can find an almost orthogonalized basis..y3 j&r A ouer Z. We actually want bounds for the heights h(P. . .4.(10g N) as where b. is a lattice in the finite dimensional R 0 A(Q).) << H(A)“‘2N”N’(log N) N)c’(‘-‘I”.. and other considerations described in [La 83b].

) comes from the bound for the product. which in twn came from the bound for the regulator by using only linear algebra..4.) comet from dividing by fi(P. the upper bound for &P. Such a lower bound is given by my Conjecture 1. . and so we need a lower bound for these heights.AN “ARIETlES CIK WI height h(P.100 ARFI. On the other hand.) for i < I.

The proof (due to Zarhin) that Finiteness I implies the semisimplicity of certain l-adic representations and a conjecture of Tale will be given in $3. while Finiteness II is a statement about finiteness of the number of certain isogeny classes. These theorems were outstanding conjectures regarded as having independent interest. which we shall call Finiteness I and Finiteness II. Parshin reduced the Mordell conjecture to a conjectured property of abelian varieties. following ideas of Tate and Faltings. Finiteness 1 concerns finiteness of isomorphism classes within an isogeny class. it is hard to remember. which we shall describe in $2. It was not entirely made clear in several expositions of Faltings’ work. that the first part implies the second part in a more elementary way. his modular height. Faltings proved them all simultaneously with the Mordcll conjecture. for instance. In the late sixties. He also used results of Raynaud on p-power group schemes. The .CHAPTER IV Faltings’ Finiteness Theorems on Abelian Varieties and Curves This chapter gives an account of Faltings’ finiteness theorems. I shall therefore give details for this implication. that the isogeny theorem for elliptic curves was not known before Faltings. just in this special case. The Shafarevich conjecture actually splits into two parts. and that a proof of this theorem would have been regarded as a major result by itself. Faltings’ proof that it implies Finiteness II will be given in $4. After that I shall briefly discuss Faltings’ method used to prove the first part. and using a new notion of height. In retrospect. after a preliminary recall of Torelli’s theorem in 51. and structure theorems for I-adic representations. Faltings gave a proof for Finiteness I based on an extensive theory using the moduli spaces for abelian varieties and curves. the Shafarevich conjecture.

@. Let C. the degree of q<p.discussion of the Faltings height and some of its properties proof of Finiteness I will be given in 55.2.e. and imbedded in their . leading to the showing how terms of the This gives an IV. . is an isomorphism.)is equal to 1. Torelli’s theorem asserts the converse.Jarobians J1.1 times). and let f:C-J be a canonical morphism into its Jacobian. 1. where the polarization is detined by the classes of the theta divisors. An isomorphism of curves induces an &morphism of their Jacobians.)+(J.. class of 0. In a last section 1 shall give results of Masser-Wustholz to bound the degree of isogenies of abelian varieties in Faltings height of the varieties and polarization degrees. but even more induces an isomorphism as principally polarized abelian varieties. We defined c to be a principal polarization if the degree of c (i. C.f(C) + ‘. so ‘p.I. Let C be a cuwc (complete. We shall now give an example of a principal polarization.c. 0) the prim+pally polarized Jacobian of C. TORELLI’S THEOREM We recall from Chapter I. we obtain a divisor on . + f(C) Let 0 denote the algebraic Theorem equivalence (y ..e. We also recall the isogeny cpc: A + A’ such that &x) = cd . A proof of the following version can be found in Weil [WC 571.G) . 51.. an element of NS(A)) containing an ample divisor. The class 0 is a principal We call (J. nonsingular as always). namely 0 = . J> ooer k. $5 that a polarized abelian variety is a pair (A. Theorem 1. see also [Mil 86a].1. c) consisting of an abelian variety A and a divisor class for algebraic equivalence (i.I times (where 9 is the genus). be cunxs defined over a jield k. By taking the sum of f(C) with itself g . It is a basic fact that: polarization. alternative approach to prove Finiteness I. Let a:(J. called the theta divisor.

also as in [NaN 811. with an isomorphism <f principally polarized abelian varieties d&ed Then u restricted to C. It is not true in general that the &morphism class of J (without polarization) determines the isomorphism class of the curve. In particular. then the fiber X. is the extension of the base from spa(R) to F then X. Let A be an abelian variety defined mer an algebraically closed . Let y be the closed point of spa(R).[IV> @I he k. There is only a finite number of isomorphism classes of curves ouer k. some point a E J(k). Proposition 18. see Milne [Mil 86a]. Using basic properties of abelian varieties having to do with the existence of a certain involution (the Rosati involution) and its positive definite trace. for C. [La 591 or [Mu 701). obtained from X by reduction module M is then also a non-singular curve.3. a curve will be assumed complete and non-singular. over k. Let A he un oh&m variety over a jield k.2. then such a scheme X is unique up to isomorphism over spec(R). The first examples are due to Humbert [Hu 18991. with iC. For a reasonably detailed account of the general algebra concerning semisimple algebras with involutions implying the finiteness statement. 52.g. and having the same genus as X. there is only ~1 finite number of principul polurizutions. see [Lange 871. = C. We say that C has good reduction at y if there exists a scheme X OKI spa(R) which is proper and smooth. For an investigation of the actual number of principal polarizations possible.4. The standard books on abelian varieties develop the theory of the Rosati involution (e. is isomorphic (us u curve) ouer k. THE SHAFAREVICH CONJECTURE Let R be a discrete valuation ring with maximal ideal M. imbeddable in their Jacobians ouer k. C. which yields: Corollary 1. Let F be the quotient field of R and let C be a complete non-singular curve over F. gives un isomorphism of C. and in particular. Then there exists only n finite numher of polarizations of given degree. one can prove a finiteness statement about principally polarized abelian varieties due to Narasimhan Nori [NaN 811. . If k is the residue class field R/M. IJnless otherwise specified. IV. whose Jacobian is isomorphic to A over k.. and such that if X. An early theorem of ChowLang asserts that if the genus of C is 2 1.jield k of characteristic zero. Theorem 1. What we really want is the combination of this finiteness theorem with Torelli’s theorem. THESHAFAREVlCH C”NJECTVRE 103 mer + a. following Narashiman-Nori.

Number field case. [Ar 711 ten to fifteen years before Faltings. Let C be a curve over F. up to isomorphism. R are Dedekind rings. but the same works in the function field case mutatis mutandis. For concreteness.finction field of a curve Y ouer an algebraicully closed field k of characteristic 0. which is a discrete valuation ring. we let R be the integral closure of k[t] in F and we let R’ be the integral closure of k[l/t]. Then there exists a finite extension F’ cf F and a jnite set of points S’ o/ spec(o.Suppose R is a Dedekind ring. with quotient field F. be a curve over F.y. proved by Faltings [Fa 831. The implication was given by constructing certain ramified coverings of the curve. and in particular. Parshin at the end of his 1968 paper observed that the Shafarevich conjecture implies the Mordell conjecture. Given g > 1 and a finite set of points S of spa(R). Let C be a complete non-singular cuwe of yews g 2 1. Let S be o finite set of closed points of Y. The Shafarevich conjecture [Sha 631. and let Y = spec(R). Then there is only a jinite number of cur-uesof genus g over F. if t E k(Y) is a non-constant function. which are not split. We say that X. For instance. Let k be a field and let F = k(Y) be the function field of a complete non-singular curve Y over k. We say that C has good reduction at a closed point y E Y if C has good reduction over the local ring 0 y. F The proof made extensive use of intersection theory on the surface X whose generic fiber is a curve X. Let g be an integer 2 1. and may be summarized as follows. Let R he the ring of integers of a number field F. with good reduction outside a finite set of points S of spec(o.).1. Then Y is covered by two open sets spec(R) and spec(R’) where R. and which huue good reduction at ull points of Y outside . Function field case. in the function field case implies Manin’s theorem. Let X.. defined over a number field F. The function field analogue of the Shafarevich conjecture was proved by Parshin and Arakelov [Pa 681. Now let us pass to the function field case. runs as r0110ws.) (cuntaining all points lying owr 2) . Let F = k(Y) be the . there exists only a fide number of CUIWS of genus g over F (up to isomorphism) having good reduction outside S. has good reduction at y if X has good reduction at the local ring Op. of genus y. Lemma 2. we express the construction over number fields.y. Let ye Y be a closed point. and is formulated as follows.

$21 THE SHAPAREVICH CONJECTURE 10s hauing the JiAlowing property.f on X does not have any fibral components. and ramified . The idea of constructing all abelian coverings of a curve by pull back from &genies of commutative algebraic groups was first used in connection with the class field theory in the geometric case [La 56~1. Let X. Unramified coverings come from the Jacobian. and if we let W. at enough primes to include those where dividing (PI) .) might ramify. b]. Then the function J above. with good reduction outside S’.22” and ramification index 5 2 ouer P. so after extending the ground field again to a finite extension F’ we may assume that D is rational over F’. which can be changed by an element of F.2) + z (eg . be the normalization of X in the function field F’(f”*). If W is a covering of C of degree n. and also enough primes so that R is a principal ring. The lemma can be proved as follows.(P. with [W. Remark.V='d+ (J) where f is a rational function on C’.(P2) by 2 is unramified outside a fixed set of primes. We first extend the base field F to a finite extension over which the points of order 2 on J are rational. then W. By a covering we mean a possibly ramified covering. The covering J + J obtained by multiplication by 2 restricts to a covering C’ of the awe which is unramified. then 2dW) ~ 2 = n(2dC) . be the generic fiber of X. Let D be a divisor of degree 0 on C’ such that 7~ = (PI) . Now let R be the localization of oF. The fact that the genus of W. of C’ = C. independent of the choice of P. can be selected without loss of generality so that C’ is the generic fiber of a scheme X proper and smooth over spec(R). Then X. so has unique factorization. is bounded independent of P follows from the Hurwitz formula for the genus of a covering.w. and [W: C] denotes the degree of a covering W of C. und so of bounded genus. defined over F’. For euery rational point P E C(F) there exists a covering W. We suppose the curve C imbedded in its Jacobian J.:C’] unramifed 5 2. Dividing the divisor class of (PI) . satisfies the desired coditions.. Let P.1) of W and eu is the ramification where the sum is taken over all points index of such a point. Pz be two distinct points on C’ lying above P. and such that the divisor of . is smooth over spec(R). . ouer all other points of C.

Theorem 2. Then the Shafarevich conjecture for abelian varieties yields a contradiction of the above elementary fact. B are isogenous over F. we shall mean the conjecture for abelian varieties unless otherwise specified. Given u number Jield F and an abelian uuriety A over F. We may now give Parshin’s application. a given abelian variety has only a finite number of polarizations of given degree. By Torelli’s theorem. The proof of the Shafarevich conjecture is reduced to the a h alogous theorem about abelian varieties. their Jacohians provide infinitely many isomorphism classes of principally polarized abelian varieties with good reduction outside S. one goes through two parts: Finiteness I.jinite number of isomorpkism clusses of abelian varieties over F which are isogenousto A over F. or say to a fixed curve W. Shafarevich conjecture for abelian varieties. which therefore has infinitely many rational maps onto C’. isomorphic to each other. Then there is only a finite number of isomorphism classesof abelian varieties A mer F with good reduction outsidr S. Proof: Suppose for concreteness that we are in the number field case.coverings come from the generalized Jacob&s of Rosenlicht. Suppose we have infinitely many isomorphism classes of curves over F with good reduction outside S. when referring to the Shafarevich conjecture. Thus good reduction depends only on an isogeny class. one must use mostly Torelli’s theorem. The Shafurevich conjecture implies Mordell’s conjecture. and in particular has only a finite number of principal polarizations. there is only a . there are infinitely many W. The same argumest applies mutatis mutandis in the function field case. . By Shafarevich. For the rest of this chapter. By a basic elementary fact recalled in $1.1 below. which reads as follows. Suppose that there are infinitely many rational points P of X in F. then A has good reduction at u if and only if B has good reduction at v. and C is a curve over a number field F. Let F be a number field and let S be a jlnite set of primes of op.2. We recall the basic fact that if two abelian varieties A. To get from abelian varieties to curves. See also Theorem 4. Parshin’s original construction used the generalized Jacob&s. contradicting the theorem of de Franchis. and concluding the proof. To prove the Shafarevich conjecture. but several peOple noticed that the above Kummer construction can be used instead.

CITY 107 Finiteness II. 93. THE I-ADIC SEMISIMPLICITY REPRESENTATIONS AND In Chapter III. $3 we defined the l-adic representations of the Galois group and of the endomorphisms. We consider these representations now in greater detail. and its cokernel is torsion free because a homomorphism which vanishes on the points of order I factors through multiplication by 1. which contain both geometric aspects and diophantine aspects. Indeed. we suppose that I is a prime number not equal to the characteristic of the field k.. We let Gk = Gal(k”/k) be the G&is group.[IV% §31 THE I-ADIC REPRESENTATLONS AN” SEM. For em-y abelian variety A ouer k.. Among other things. and other theorems concerning abelian varieties. Note that the tint isomorphism (1) is equivalent with the second (2).2. the map in (1) is inject%. Conversely.MPI. Consider the following properties of a field k and abelian varieties cwe~ k. Property 3. there is only a finite number of isoyeny classes of abelian varieties mm F of given dimension and huving good reduction outside S. on K(A) is semisimple. we obtain a seemingly more general statement which is an alternative form of the Tate property [Ta 661.2 (Tate naturul maps (1) (2) are ismorphism. property). all modules involved are torsion free finitely generated. the over k. As before. Property 3. As a consequence of Property 3. Hence if (2) is an isomorphism. we shall prove: Finiteness I implies Finiteness II IV. For simplicity we omit the sign @ when tensoring with QI or 2. Given (I number field F and a Jinitc set cf primes S.1 (Semisimplicity). so tensoring an &morphism (1) with Q! yields an isomorphism (2). the . For every abelian variety representation of G. The next sections describe the properties of abelian varieties used to deal with these steps. it follows that (I) is an &morphism.S.

Multiplying by a suitable I-adic integer..3. Let A. we can write cp = i: ziai B) i=. If V. and so is an isogeny.. B be abelian varieties defined over k satisfying 3. Let rzi be ordinary integers I-adically very close to zI.(B)). we also get the immediate consequence: Corollary 3. for vector spaces over QI. B) + H*m. T.(A) x Hom.a.K(S) be a G. with zi E Z. Ho&4 4 + HomGk(7U). over Z.2.2.(B) of finite cokernel.3.(A. and use the formula End&4 x 8) = End.4 (Isogeny theorem).Hom. if c+.V. has finite kernel in A. .c(~ is a basis of Horn&.(E) of as well as the analogous formula for the ring of G. Note that the argument actually proves that Corollary 3. are isomorphisms. K:(B))> Q.-isomorphic. Proof. then A. B) x Hom. Proqf: As in Property 3. A) x End. We then apply (2) of Property 3. B are isogenas suer k.. we may assume without loss of generality that we have a homomorphism rp: T(A) -+ 7..3 implies Corollary 3.2. The two finiteness properties may be formulated for other fields besides number fields over which they may be true.1 and 3.(A)..2 to the product A x B..(B. it sufices to prove the second statement. We denote them by Finiteness I(k) and Finiteness II(k) . B be abelian varieties over k satisfying Properties 3.-endomorphisms l$(A x B) = K:(A) x v.(B) In addition. Then Xn.~(V. .-isomorphism.(A) and v(B) are G. thus proving the theorem.4. Let A.1 and 3. Let rp: &(a). Then Ihe nutural maps Z.Corollary Properties 3. By Corollary 3.

/I”IT.]-submodules W of 1/.5 ([Zar 751).2 from Finiteness I are not very long. Then trivially side But B. . If the I-isoyeny rkm of A over k contains only a Jinite number of k-isomorpkism classes. then f&r .Y. Part of them stem from [Ta 661.: T + 7.. = /. = w 7.: A --t A/W.. and the corresponding . And second. Let: W”O = wo + l”T c ?y K = Y”vw We can identify where y. Corollary 3.(A) suck that t&:(A)) Proof wo= Abbreviate Wn7. by definition W.4. Both a.1 (semisimplicity) and Property 3.1 and 3. and hence also the isogeny theorem ouer k. Because the arguments used to prove properties 3. as a subgroup the isogeny a. These arguments are contained in two lemmas.2 (Tate proprrty) for the field k. = B.. is the canonical of A[I”].. We claim that fln(7. Lemma 3.+A such that @.(A) and F = V&4). and bn are defined over k. for n 2 1. we shall give them in full. Then we have a theorem of Zarhin: Theorem 3.U-Wl) = ker E.. = ker a.4.[G.(A) these exists u E Ql End.111Q. We consider W. in a[[“]. The first gives us projection operators on the way to semisimplicity. Finiteness I(k) for I-isoyenirs implies Property 3. Let us also define an I-isogeny to be an isogeny whose degree is a power of the prime 1. first B”m4) = i%W) = m(A). map.(Bn)) = Wz It suffices to prove that both the right-hand side and left-hand contain I”T and have the same canonical image mod l”T.for a field k. = 7.:B.

Therefore ui is actually a” endomorphism of W”“. End. End. §31 Conversely. denoted by the same letter.(A). There exists y E B such that x = &y since 8. and isomorphisms vi: B.(A) in End V&4).(A)) = W.] in End V.(X) = 0.110 FALTINCS’ FlNlTENESS THEOREMS [IV.[G.“) = r+$” Thus finally u(l/.7. for iEI. is surjective. But Q. End.‘) is compact.[I”]) thus proving the claim.End.$I&’ Then E Q End.6. Let E = image of Q. let W.W’J’ = P&4).(A). every element x 6 II(@) is a limit x = lim xi with xi E ui(W. and the lemma is proved. Hence the sequence (ui} also converges in Q. Let ui = .(x) = q3.-semisimple. Let again A be an abelian variety defined over a field k. c &(B. how to apply the conclusion of Lemma 3. Since End(W. Suppose that for each Q.. Since Wz is compact.kW Then E is the cornmutant .[G.(A). after taking a subsequence of I we may assume without loss of generality that the sequence {IA.‘. Q.(A).(A) = E”d. we can find a sequence I of integers with smallest element n. + B. End&4) such that u(K(A)) = W Then: (1) (2) (3) K(A) is G. The next lemma shows Lemma 3.]-submodule W of l&4) there exists an element u E Q.} converges to an endomorphism u of W.y = I”y so y E B[l”] whence ker a. By the hypothesis that there is only a finite number of isomorphism classes in the isogeny class of A. But then 0 = a. of Q. so we may view u as the restriction to W”” of an clement of Q.

Take for W a simple factor of 6. All of Faltings’ theorems which we are stating for number fields have also been proved by Faltings for finitely generated extensions of Q. Given x E W there exists s E S such that cx = sx.5.e.(A) was proved by Sure [Ser 721. Theorem 3. and therefore cw commutes with elements of End. Let r: = v(a). Let k = F be a number field.1 and 3. will prove that Finiteness I implies Finiteness II.(A) is open in GL. the isogeny theorem. let c E C. it will now suffice to prove Finiteness I for a number field. whence cw=cuv=ucvc w so W is stable under c.(A) is semisimple. let W be a G. Since elements of G.1 and 3. that is End&l) has rank 2 over Z. = W. The right ideal u(Q! End.(A). To prove Theorem 3.(Z.3 and 3. on V. End. Indeed.2. Theorem 3.2). Remark 2. the (. or the image of Gk in Aut T. End.(A) such that uV = W’. and so implies the Shafarevich conjecture.-scmisimplicity of K follows. Then abelian varieties over F sntisjj semisimplicity and the Tate property (i. It follows that c E E.). Remark 1.Prwf.(A) is a semisimple Q. so (3) follows directly from (2) and the basic theory of semisimple algebras. and therefore their corollaries. it follows that cw is the same as multiplication by an element of QI. we follow Tate.(A)) is generated by an idempotent e and eV. Then for every G. as in my Algebra. To prove (2). there exists u E QI End. thus proving (2). Properties 3. the semisimplicity of . By hypothesis.(A). the semisimplicity of the representation of G. e. To prove (I). When dim A = I. [Use for instance Jacobson’s bicommutant theorem.(A) such that u(K) = W.8. who proved that either A has complex multiplication. Let C be the cornmutant of Ql End. Since W is S-simple.-algebra. there exists u E QI End.2. commute with elements of End.(A) in End(v). Conversely. corresponding to a simple factor S of E.4.submodule W of r: we have CW c K’.. since Q.1 Combining the two lemmas proves Properties 3. Chapter XVII. Thus we have proved Theorem 3. and therefore also their consequences Corollaries 3. which is more elementary. End. It is a basic fact from the theory of abelian varieties that Q. specifically: the conjecture of Shafarevich.-subspace of F. it follows that F is scmisimple over Q.8 (Fakings). Then E c C.g. The next section. This will be described in $5. As for (3).(W).

the Mordell conjecture followed from the number field case in the more general case by specialization.. THE FINITENESS OF REPRESENTATIONS FINITENESS I IMPLIES The previous section consisted CERTAIN FINITENESS I-ADIC II field. The decomposition group G. then it is part of the basic general theory that the reduction map x4 + wb. --t Aut(V) be a representation (continuous homomorphism). be the associated valuation. for w/u. it is the stabilizer of w.ings’ Chapter VI in [FaW 841. We let D I a prime Let V be a finite dimensional vector space over the I-adic field QI. Since all the decomposition groups G. e. and not eyuul to the ckaracmistic of k(u).e. and let p: G. the isogeny theorem. 54. IV. I In addition. Equivalently. The kernel of p is a closed normal subgroup.) . then all w/u are unramified.. the Tate conjecture. k(u) the residue class field. whose fixed field is a possibly infinite Galois extension K of F. The kernel of this homomorphism is called the inertia group. it follows that if scnne w is unramified. If A has good reduction at v. Such a representation is called an I-adic representation.I-adic representations.. i. Let w bc a valuation of K extending u. Let A be an abelian variety defined over F. which leaves w fixed. and we thus get a homomorphism G. for wlu are conjugate in G. is the subgroup of G. induces an automorphism of the residue class field k(w) over the residue class field k(u). the function field case over finite fields was known by work of Tate [Ta 661 and the sequence of papers of Zarhin [Zar 74-761. If the inertia group is trivial. the representation p is said to be unramified at u if its kernel contains the inertia groups of G. so we simply say that u itself is unramified in K. See Fal.“. $7. Next: of theorems over an arbitrary We let R be a discrete ualuation ring and F its quotient field.. c G(k(w)/k(v)).g. An element of G. then w is said to be unramified wer u.. via the Hilbert irreducibility theorem. 4 c. As we already pointed out in Chapter 1.

. Corollary 4. Let A . The proof of Theorem 4. If p is unramified a: II. to p’. unrumified outside S.. if k(o) has 4 elements..3 is based on the following lemma. such that Fr. E G. then E”(X) = xq for all x E k(w) Hence the trace All such elements Fr. Let F he a number field and let d be a positive integer.e an abelian variety defined over the field F which is the quotient Jield @’ a discrete valuation ring. Let S be a finite set of$nite places of F.)). Up to isomorphism. p’ are two semisimple l-udic representations of dimension d. Lemma 4.) = tr @(Fro) then p is G. but see Serre-Tate [SeT 6X]. such that if p.[IV. there exists only B finite number of semisimple representations of G.-isomorphism F&4) + b(B). and WC can apply Theorem 4.4. and mostly N&on in general.-isomorphic for all u E s’. then there exists a unique element Fr. there exists a finite set s’ of places disjoint from 1 and S. inducts the Frobenius automorphism on the residue class field of w. to A owr F and A Proof An isogeny induces a G. and such that fir u $ S the traces of Frobenius elements Fr. then B has good If B is isogenous reduction also. and for each v 6 S let Zu be a . of G. . and that the representation of G. Much deeper is the converse due to Ogg for elliptic curves. Given u finite set S of finite places of F.1 to conclude the proof. Let 1 be a prime unequal to the characteristic of the residue class field. of dimension d wer QI.jinite set of elements of QI.3 (Faltings). depends only on v and not on w. This Frohenius automorphism is the unique element such that... Thus we shall write it tr(p(Fr. lie in Z.(A) is unramified. und if tr p(Fr. §41 T”E FlNlTENESS OF CERTAlN I-i\orc REPKESENTATIONS 113 k an isomorphism. let F be a number field and let u be a finite absolute value. now have a basic result. are conjugate in G.1.. unramified outside S. on 7. From now on. Then A has good reduction $ and only if the representation of G. We Theorem 4.2 (KoizumikShimura). has good reduction. on &(A) is unramifed. Theorem 4.

3 follows directly from Lemma 4. We have a natural homomorphism G. Hence Theorem 4.(T) Then dim. By a theorem of Hermite..-algebra generated by the image of G. This proves the leIllIX+. p’ as in the lemma. . representatives of RjlR generate R over Z. Thus in fact. where q.(Fr. such that for all representations p. in (R/IR)*.3 applies to the I-adic representations with which we are concerned.(Fr. Hence the cardinality of RjlR is bounded. Since G. Remark. is compact. in the finite group (R//R)* is unramified outside S. be the representation on K(A) for some abelian variety A of dimension d over F. unramified outside S. Now by Tchebotarev’s theorem.) = tr p’(Fr. by Nakayama’s lemma.) is an ordinary integer.(R/W* / and #(R/IR) 5 ldtmn.114 FALTINGS’ FINlTENESS THEOREMS [IV. and R/IR is itself generated over ZJZ. in (R/IR)*.)l 5 2dq:“. 641 Proof.. (An I-adic lattice is a free Z. The representation of G. in the I-adic spaces V and V’ of p and p’ respectively. in the sense that they can take only a finite number of values for each v. there exists a finite set S’ of absolute values such that the Frobenius elements Fr. there exist I-adic lattices T in V and T’ in V’ which are G&able. by the images of the elements of G.jR) x End&).4. in End. there is a subgroup H which is closed and of finite index in G. By Weil’s theorem (Hasse in dimension 1) the trace of Frobenius tr p. Let p. Example. If we now apply the hypothesis that the traces of Frobenius are bounded.-module of rank d. then we obtain tr p(a) = tr p’(a) for all c( E R. then Theorem 4. and we are actually representing the finite group G/H in (R/IR)*. If tr p(Fr. Furthermore. satisfying Itr p. whence p is G. there exists only a finite number of extensions of a number field of bounded degree. for v E S’ have images which cover the image of G. the kernel contains H. Faltings’ lemma is remarkably simple. 5 8d2. is the number of elements in the residue class field. The difficulty lemma lay in discovering it! of the .) for Iat S’.-isomorphic to p’ since the representations are assumed semisimple..) Let R be the Z.

whence follows Finiteness I.)’ = (*@(a.) = /j”‘“‘Cl’(A. THE FALTINGS FINITENESS I HEIGHT AND ISOGENIES: Isomorphism classes of abelian varieties are parametrized by what are called moduli varieties. is smooth over spec(o). IV. by Theorem 3. Unless otherwise specified.4. WE have the I-Finiteness und Tote’s property fir I and Isogeny theorem for I Proof. $4 we let u be a Dedckind ring with quotient field F and we let S = spec(o). and we let A. Let R’ denote the usual sheaf of diflercntials. there exists only a jinite number qf isogeny classes of abeliun varieties ooer F which have dimension d und good reduction outside S. be an abelian variety over F. II./S) by Q’(A. be the zero section. Actually..5. the dual of the Lie algebra Lie(A. 55.(A) is a semisimple representation of G.). a number field F and a jinite set of primes S. Faltings defined the height of an abclian variety directly.Corollary followiny 4. This proves Corollary 4. and by Tat& property in the form of the Isogeny Theorem 3. We Ict A. be its N&on model over S. we conclude that there is only a finite number of isogeny classesover F with good reduction outside S. it follows that a’(A. To prove this. the results of this section are also due to Faltings. WC then have the determinant det @(A. If A is an abelian variety of dimension d and good reduction outside S. then V. We let S./S) is locally free of rank equal to the relative dimension of A.5. we fix a prime 1. We conclude that there is only a finite number up to isomorphism.5. We abbreviate @(.). Let [: spec(o)+ A. Recall that Finiteness II states: Given a positive integer d. As in Chapter III. Fini1eness I implies Finiteness implications: I a a a Semisimplicity Semisimplicity Finiteness II.4. He proved that the heights of abelian varieties isogenous to a given one are bounded. and we can apply Theorem 4.). over S. consist of S together with all the primes of F dividing 1. Since A. and what we call the . and showed how it was related to the height of the associated point on the moduli variety with respect to some divisor class.3 to the family of such representations.

let cr: F + C he an imhedding inducing u on F. For s E L. to be deg 2’ = deg L = -1 ” log/ls~l. s # 0 we have /sI.4.. Each isomorphism of F. + L. on L. The elements s t L are just the sections of 2. Then we Ict IISII” = 0. The right-hand side is independent of the choice of s by the product formula. s # 0 we have IlsIl. The co-lie which is the determinant of the dual Lie algebra determinant is a line sheaf over spec(o). : Q... transports the absolute value of F. By an o-form of L we mean a module L. Suppose given a hermitian structure.. = [F.]... . Two complex conform ( .-vector spaces F.) = [* det R’(A. z L.. = 1 for almost all v. L. jugate imheddings give rise to complex conjugate spaces L. Let N. and we assume that for s E L we have (s. so that F&L..). s). [*n’(. and let llsll.) = det Lie(A. w. s>i” lbll” = (a s>. Then given s E L. i.. for u real for 1) complex. = C SF. satisfying s E L.)=“~” Il47’~ For u at infinity. Suppose given for / iI.. Let u be a finite absolute value on F. F. and the degree of the line sheaf 2 = Lassociated to L.. = (5 S). = 1 for all but a finite number of v.e. We define the degree of L. to L. We also assume that for s E L.)‘.co-lie determinant &4. We now define an Arakelov metric and degree spa(o) in the case of a number field F whose ring First let L be a vector space of dimension 1 over each absolute value u on F a u-adic absolute value 14” = l4”lSl” for all a EF and for a line sheaf over of integers is D = Do. to I. >.. = ISI?. a positive definite hermitian on the C-vector space L. We pick such an isomorphism sending the local ring D. locally free of rank I over o.

. Then Lie(A. starting with our abclian variety A over F. and the above formula detines the Faltings hermitian structure on Z(A. h. d. be an abelian variety over the complex numbers. At the beginning of Chapter III we defined N&on models and semistability. Let F be a number field. we have We define the hermitian such that for o E Z’(AC) Globally.)” = H”Q’(Ac) structure and Z(A. has semistable reduction. of dimension d.. and if F’ is u finite extension of F. we can define the Faltings height 1 Theorem 5. We now have: Proposition 5. we may define the stable Fakings height .1 for all abelian varieties in a given isogeny class. positive integers. if A. c) over F.(A) 5 h.). Let H”R’(Ac) be the space of holomorphic I-forms.The above constructions give rise to a set of absolute values on L of the type which then allows us to define deg L. Let A. We also mentioned the basic fact that taking connected components of N&on models commutes with base change under the hypothesis of scmistability. n. Note that for the application to the proof of Finiteness I. there is only a finite number of points of P’“(F) of bounded height. Then there is only a finite number of isomorphism classes o/ polarized abelian wrieties (A. are equal.1. As a result.1 is of course analogous to the theorem that in projective space. Faltings compares his height with an ordinary height associated with an ample divisor class on a variety which parametrizes isomorphism classes of certain abelian varieties. To prove Theorem 5. Consequently. for each imbedding o of F into C we obtain an abeiian variety A. We apply this degree to the Lie determinant.. und Ap. and polarization degree n such that h. Let F be a number field.2. Theorem 5.) at the infinite absolute values. it suffices to prove the finiteness of Theorem 5.1. over C. then the Fakings heights qf A.) = A”‘“” H’R’(A.

such that all components have the . and a level N structure with a positive integer N divisible by at least two primes 2 3. and whose complement in pa.N which is non-singular. since we are dealing with a smooth family of abelian varieties over M.+.).. Theorem 2. V. Given a positive integer d and u positive intryer ible by at least two primes > 3.. there exist: N diuis- A finite disjoint union qf projective varieties Md. The isomorphism is to be a Galois isomorpbism.N (for a proof. &xl = f -‘W for x E J%.N.N has codimension 2 2..118 for any finite extension F’ such that A. dejined ouw Q in the sense that this union is stable under the action of the Galois yroup G. see for instance [CbF 901.. over Q which is an abelian scheme of relative dimension d. in the following sense. a principal polarization c. Let N be a positive integer. has codimension d.N. an open Zariski dense subset Md. such that for every extension field k of Q the association XHLL cx.N in Md.. E) consisting of an abelian variety of dimension d. c. a proper smooth morphism f: A + M. Then there exists a moduli space M. defined wer Q as above. and when d 2 2 this extension is unique because the complement of Md.5(l)). In addition. By a level N structure variety A of dimension d we mean an isomorpbism 8: (ZjNzy + A[N].. It follows that such a triple has no automorphisms other than the identity. is semistable. for such triples. This co-lie determinant corresponds to a divisor class in Pic(aa. for an abelian Let A be defined over the field k of characteristic 0 which is all we want here..sume dimension d(d + 1)/2.3.. a he1 N structure E: (Z/NZ)2d + A[N].. Theorem 5..+. an alyebruic equiualence class e on A. This line sheaf has an extension to a line sheaf over Md. we can define the co-lie determinant as before as a line sheaf 2 on Md. points The algebraic set A& is called the Baily-Bore1 compactification of Md..(k) and the iso- is a bijection between the set of rationul morphism clusses of triples ouer k. We consider triples (A.. and consequently this means that all the points of order N are rational over k. which we shall call the co-lie determinant class .

c.N --t R: (up to O(t)) which we call the Lie height on M..N. This is the analogue of Theorem 5. because j. These properties summarize the relevant facts about the moduli variety for the moment.N). also called the canonical on M.1 lo Theorem 5.. It is a fact that in a given triple (A.. we are actually dealing with the semistable height on the left-hand side. corresponding class... 4)). If (a. K are ample.I a height function hi: Md.. the error term with a big 0 on the right makes sense. The elementary fact that there is only a finite number of points in c. To reduce the general case of Theorem 5. Proofs can be found in Chai Faltings [CF YO]. for every number field F.4. 4 the associated point in the moduli space M. we denote by m(A c. . even in the non-semistable case. we can associate to .(m(A c. then Raynaud has proved in general.4. are the essential ingredients which go into a full proof. Theorem 5. c. [Zar 771) to deal only with principal polarizations. is ample. to be positive. As such.. c) over a number field. that &dA’) = h&A) These and Zarhin pointed out that (A x A’)4 is principally polarizahle.. Faltings then established a connection between his height and the Lie height as follows. c. Furthermore. the abelian variety A is semistable.4. we have the relation (d + l)i = K. so in Theorem 5. = hl(&4 Since J. E) is a triple.. now implies immediately the finiteness statement for triples with bounded Faltings height over F. because we can take h. and I an much indebted to Chai for his guidance on moduli facts. if A’ is the dual variety of A.i.(F) of bounded height.1 for triples..4. is ample. Finally when d 2 2 it is a fact that the canonical sheaf A”“” f&o has a unique extension to a line sheaf to a class in Pic(M.1 to Theorem 5. 4) + Oh h. and one can use a theorem of Zarhin ([Zar 741. one has to adjoin points of finite order to create semistability.. Indeed. It is a technical malter to reduce Theorem 5. For ewry h&(A) triple (A. and both E. E) OWT a number field we have c. and denoted by K.

) is injective. is a finite set.5. [Fa 84bJ Theorem 5. extending Faltings’ results to the non-semistable case. is semistable.).). if available. Faltings stated explicitly that it would require an entire book to justify properly the use he was making of the moduli spaces. These properties have since been proved. together with the next result from [Fa 831. and Deligne pointed out certain properties (a). and will be found in the book by Chai-~Faltings [ChF 903. is its extension to the N&on models. Indeed.) and S?(B.) On the other hand. If A. (# 0 as usual) such that 2” = 0(&J. Then Raynaud’s formula is: = deg(%) = .$[F : Q] deg(u) + log@ : b. If t+: A. d-z %(A . We call b.) + Z(A. Raynaud ([Ra 851. and for the diffcrcnces by a modification and extension of Faltings’ methods. Let F br a number field and A.For a mope dctailcd exposition of Faltings’ theorem. + B. . has semistable reduction. the reader can consult Faltings’ own write up on heights and moduli spaces [Fa 84a. if A. the The Faltings metrics on L&4. an abelian variety over F with srmistable reduction.P. Faltings used another expression. and u: A. would simplify the exposition considerably. We now come to the second ingredient which enters into Faltings’ proof of Finiteness I. Indeed. Theorem 4. which is what Faltings used in his original formula.deg JW.) . b]. tensor product P(B.) give rise to the metric on r. over F.9) has given an eflective bound for the height of abelian varielies isogeneous to A. This finiteness comes by putting together Theorem 5. There b. Hence we get is the sheaf associated with the ring of integers o. (b). (c) which. define The canonical homomorphism a canonical injection where 0 = o‘ exists an ideal different of u. then (o : b.4. let i be the zero section and let G = ker u.) = rank 5*0&. then he gives an explicit formula for the difference of the heights.1 for a given isogeny class. then instead of log(o : b. as well as Del&e [De 831. + BP is an isogeny over F. Then the set of Fakings heights of abelion varieties BP which are F-isogenous to A..

d) und k = k(n) having the following properry. and Wustholz [Wu 841. Let d. IV. among other things. Let A. n. Schappacher [Sch X4]. This is the approach taken by Masser-Wustholz [Maw 917 who solve part of this problem effectively as in the next theorem.(A))‘.The finiteness ing result. Then there exist effective c = c(m. One basic isogeny problem then is: . If abelian varieties of dimension n over a number field F ~>f huuing polarizations of degrees aI most m. Different types of finiteness bounds occur here. such that for my isogrny up: A. or in this case a bound for the degree of the isogeny. . One approach runs as follows. There exists a finite set S qf primes. h&. Making explicit that one can extend the result without mentioning the polarization was done by Zarhin [Zar 851. isogenous integers. We want to bound in some senw isomorphism classesof abelian varieties isogenous to A. We shall deal with this from another point of view in the next section.5 comes partly from the follow- Theorem 5. Theorem 6. and if they are ouer F. Szpiro [Szp 85b]. n be positive constants A. It uses other results of Raynaud [Ra 741. using his product trick. + Br qf degree not divisible by any primes in S. namely a bound for the height of the solutions. B are degree d. One of them is a bound on the number of solutions of a given diophantine problem. See also the expositions of Faltings’ work in Deligne [De 831.jind another isogeny degree is bounded only in mm of A. THE MASSER-WUSTHOLZ TO FINITENESS I APPROACH Let A be an abelian variety over a number field. Let a:A+B 0: A + B whose be an isogeny. of heights in Theorem 5. then rherc exists an isogeny ouer F of degree ut most c. have semistable reduction. and to relined properties of ramification.but another requires more.max (I.1. $6.6. we haue This result reduces Theorem 5. WI. Finally we note the technical point that Faltings in [Pa 831 proved the Shafarevich conjecture for abelian varieties for a given polarization degree.5 to the study of p-isogenies for a finite number of primes p. in this case the number of isomorphism classes.

if we are given a polarization on A with known degree.1. for various reasons. We note here already that the proof in the manuscript I have seen.122 FALTINCiS’ FINITENESS THEOREMS [IV> 561 As in Faltings’ proof the above result implies Finiteness I by using the polarized.1 in Chapter IX because this proof in some ways follows a pattern whose origins lie in the Baker method of diophantine approximation. We shall return to a discussion of the proof of Theorem 6. and also uses an explicit version of Theorem 5. the above approach leads to the following questions raised in their forthcoming papers: Bound the degree of some polarization on A in terms of its Faltings height. I shall not go any further into them. Bound the degree of an isogeny as stated at the beginning of the section. Zarhin remark that (A x A’)4 and (B x B’)4 arc principally However although Theorem 6. For instance. Bound the discriminant of the ring of endomorphisms of an abelian variety in terms of the Faltings height. .1 is effective. such a reduction requires ideal class estimates for endomorphism rings of abelian varieties which are not known to be effective today. As Maser-Wustholz have observed. In particular. we do not know a priori a polarization on B whose degree is bounded in terms of k&(A) and the degree of the polarization on A. like Faltings’ proof also uses a good dose of moduli theory as in Chai-Faltings. at the moment the reduction of the basic problem to the case of principally polarized varieties is not. Since current results concerning these questions are very partial at the moment. the Maser-Wustholz proof replaces Raynaud’s p-adic theory by arguments at infinity. Of course one can require more than just bounds.4 which makes effective the constants implicit in that theorem. namely bounds having roughly a form similar to the estimate in Theorem 6. However.

Frey had the idea of associating an elliptic curve to every solution of the Fermat equation in such a way that the curve would exhibit remarkable properties which would contradict the TaniyamaShimura conjecture.CHAPTER V Modular Curves Over Q Among all curves. It turned out that there were serious difficulties it) carrying out this idea. there is a particularly interesting family consisting of the modular curves. which parametrize abelian varieties with other structures involving points of tinite order. In this case. They form the prototype of higher dimensional versions. These parametrize elliptic curves with points of order N. of modular CUWCS. or quotients of the Jacobian. Here we concentrate on elliptic curves and points of finite order. over Q. in Chapter IV. or cyclic subgroups of order N. Mazur was able to describe completely the rational points in the most classical sense. and more specifically of the Shafarevich conjecture. and we shall describe Mamr’s results in $2. . These results also involve the determination of certain MordellWeil groups for the Jacobian. as parametrized by modular curves. We shall indicate the main idea in $3. Ribet succeeded in proving a result on such representations which was strong enough to show that TaniyamaShimura implies Fermat’s last theorem. A famous conjecture of Taniyama-Shimura asserts that every elliptic curve over Q is modular. i. having to do with the Galois representation on points of finite order of the curve. modular varieties. $5. lying on the modular CUTV~S. in the sense that it is a rational image of a modular curve. We have already seen the use of such varieties in Faltings’ proof of the Mordell conjecture. which we shall describe in $1.e.

The coset space has a representative fundamental domain with a well-known shape pictured below. Then r(l) operates on F. At the present time. V. . which parametrizes divisor classes. stemming from Tunnell. it is not known what to do when the rank is greater than I. $1. and also showing how more specific results have been proved in the case of curves defined over the ultimate base field. or isomorphism classtx of line sheaves. We start with a complex analytic description of the situation. so we get a faithful representation of r(l)/& 1 in Aut. Finally. in $5.(Z) be the modular group. following work of Gross-Zag&. by and k 1 operates trivially. that is the group of matrices with intcgcr coefficients and determinant 1. and also establish a connection with the theory of cyclotomic and modular units. Here we shall be concerned with isomorphism classes of elliptic curves and some additional structure arising from points of finite order. Thus the present chapter can be seen as giving concrete instances of the more general theory of previous chapters. we show in the case of rank 1 how one can construct a rational point of infinite order. BASIC DEFINITIONS A general pattern of algebraic geometry is that isomorphism classes of certain geometric objects are parametrized by varieties. Let T(l) = SL.($j). The modular construction and the insight via the Birch-Swinnerton-Dyer conjecture provide an explicit solution of a diophantine problem. that is the set of complex numbers r=x+iy with y 2 0.In $4 WC give an application of modular theory to a classical problem concerning Pythagorean triples. or what to do in higher dimension. the rational numbers themselves. Let $j be the upper half plane. We have already seen an example of this situation with the Picard variety. The general pattern behind these special phenomena is also unknown at present. with additional information due to Kolyvagin.

w2] by [CLU. = ?.j(oo) = a). we shall always suppose that wl/wz lies in the upper half plane. thus obtaining a compact Riemann surface isomorphic to P’(C).[V> 511 125 r(l) There is a classical function. co2 over the integers. while . In terms of q. and the above argu- . In addition. Rut C/A is a complex torus of dimension I. This means that A is an abelian group generated by o1 and uZr and that these two elements are linearly independent over the real numbers. CWJ for any complex number c # 0. + 744 + 1968849 + higher terms. If as a local parameter at infinity. with basis co.... Thus we may define j(A) = j(r).. This is rather ad hoc./w. and in addition is the same if we replace [w. and we have j(cA) = j(A). Let A=[q. A better way to conceive of j is in terms of isomorphism classes of complex toruses as follows. then one can compactify r(l)\B by adjoining the point at infinity.{co} infinity under with the afiine line (projective one takes from which is deleted). One can characterize j analytically by stating that j(i) = 1728 and j(e*“‘“) = 0. the function j has a Laurent expansion 1 j = . which gives a holomorphic isomorphism j: r(l)\% + P’(C) I-space . q] be a lattice in C. so we put w. Then the invariance of j under SL.(Z) shows that the value j(r) is independent of the choice of basis as above. h&morphic on $5 and invariant called the j-function.

1I. Let N be a integer. such that y. classes of ~OVASL’S Furthermore.TV) gives a hijection together between T. be the cyclic group generated by l/N. of finite index. Now possibly in some positive where A = g.(N) = subgroup d b arbitrary We may view I/N as a point of order N on the torus C/[q 11. If we put g#) = 60 x.(N) and Y. The association and isomorphism classes of ~oruses T+a(C/l? 11. Let Z. a. 11. . which WC now describe.(N) = subgroup 1 0 D . defined over Q.wd and Y. urn4 and g&i) = 140 1 u-0.(N)\$ together with a point of order N. The value 1728 is selected for usefulness in arithmetic applications. where the sums are taken for w E A and o # 0.(w) = r. WC have the following parametrizations: The association T++ (C.(N)\@ and isomorphism with a cyclic subgroup qf order N. then j = 1728g:/A.merits show that j is the single invariant for isomorphism classes of such toruses.27~: let r be a subgroup of r(l).(N).‘Cr. Then we may cnnsider the pair (C/CT. ! A i > o a mod N. l/N) &es a bijection between T. so that the q-expansion of j has integer coefficients. We define: of elements y s ( of elements y = ( of elements y E i l-(N) = subgroup r. We shall be specilically interested very special subgroups r. Z. Then r\b is a finite ramified covering of T(I)\$. mod N with arbitrary mod N with 1 h. b. there exist affine curves Y. T.(N)(C) = rdN)\S.) as consisting of a torus and a cyclic subgroup of order N. .

and the points at infinity are called the cusps. Then the only rational points of X. The set of cusps on X. 52. Mazur gives a complete description. MAZUR’S THEOREMS We consider the modular curves X.(N) --t Y.(N) parametrizes isomorphism classes of pairs (E. 67.“(N). in the following theorems [Maz 771 and [Maz 781.(N) is denoted by X. P) of an elliptic curve and a rational point of order N by the points on X. then a point of Y. This covering is Gal&. The ramified covering Y.(N) + X. Similarly.(N) is z 0. 19.(N) + X.IY. in the following sense. Mazur gives a complete table describing X. In light of the representation of pairs (E. 781.(N) extends to a ramified covering X. we have the completion X.. Its completion is denoted by X. 17.(N)+ X.(N)(k) corresponds to such a pair (E. 37. that is. to (Z/NZ)*/k 1. 43. or 163. P) algebraically. one gets an interpretation in terms of elliptic curves as follows. except when N = 11. Similarly.(N) can be compactified by adjointing the points which lie above j = co.(N) parametrizing pairs (E.(N) with its cusps X.“(N). N = I I or N 2 17.. then we have natural finite morphisms X. P) with E defined over k and P rational over k. @I MAZUR’STHEOREMS 127 and such that Y. Let N he ~1 prime For the exceptional cases of N listed above.(N) cwer Q and ask for a description of the sets of rational points. If p is a prime dividing N.(N) of Y.. for Y. Z). The afine curve Y. Theorem 2.(N) and X.(N) defined over Q. numhrr such that the yenus f>f X.(P) The essential problem is to describe the rational points when N is prime. under the association a+brsa = a i 0 0 a-1 ! with Galois group isomorphic mod N for a t (Z/NZ)* V. .1. where E is an elliptic curve and P is a point of order N.(P) and X. where E is defined over k and Z is invariant under the Galois group G.(N) in Q are at the cusps.(N)(Q) in [Mu.(N).(N). If k is a field containing Q.

Z/mZ Let E be an elliptic cur-w over Q. PI Theorem 2.(N) and J. It turns out that there are precisely two cusps on X.(N) by giving generators. We map these curves into their Jacobians over Q. 61. It is a fact that C is finite. Let N be a prime ow Q admits a cyclic subgroup group G. which corresponds to a point in the Jacobian.(N) gets imbedded in its Jacobian. We define the cuspidal (divisor class) group to be the group generated by the associated point in the Jacobian. 12 WC suppose from now on that the genus of X. 31. subgroup E(Q). 19. 5. is isomorphic to one of the folkwing with with Then the torsion groups: 15 m 5 10 or 1 5 d 5 4. If we let Q. 13 when X.(N). Mazur gets: Theorem 2.(N) is z 0.128 MODULAR CURVES OVER Q cv. which arc denoted by J. I.(N): X. and send the cusp at infinity to the origin of the Jacobian. Now take N prime 2 3.(N). OY 163 otherwise As to rational torsion points on elliptic curves.) is a O-cycle of degree 0 on X. N = 11... Then: number such that some elliptic curve of’ order N stublr under the Gnlois N = 2.2. and denote this group by C. We shall describe the algebra of endomorphisms of . in terms of the modular representation via elliptic curves and cyclic subgroups of order N.. 2122 x Z/ZdZ Part of the investigation of these sets of rational points lies in an analysis of the rational points on the Jacobians of X. and it is easy to see that the covering is unramified over the cusps. 43.. The degree of the above covering is given [X.I. so X. and that N-l order of C = n = numerator of -..(Q. 1 I. Q. be the two cusps (lying above 0 and m respectively in the complex representation of the upper half plane). then (Qd . 3.(N).(N)] = y.(N) lying above j = co..3.(N) and X.. Such endomorphisms are induced by cor- . m = 12..(N) has genus 0.

so the Eiscnstein terminology had precedence over cuspidal terminology. Z) H (E/Z. We call T the Hecke algebra.(N) which (E. n=l Then I” is an ideal of T.(N)/I”J. denoted re- We let T be the subalgebra of End J. Let I” = fi I”.(N) generated by w.g (E/H.(N).5). (Z + H)IH) IjN.. and 1 + I . Second.(N). Then Mazur proved that the Eiscnstrin ideal is equal to the cuspidal ideal.7..(N) (proof in [Maz 771. and all T with l)N. We now describe these correspondences. (for IJN). These correspondences spectively by induce endomorphisms wN. However. Chapter II. this involution is induced by the map TH -ljNx. Mazur defines the Eisenstein ideal I to be the ideal of T generated by I + w. = WI and T. We let the Eisenstein quotient be B = J. of order 1.(N) . J. It is a fact that: T = End. In the representation from the upper half plane.(N). and a correspondence is described by associating a divisor to a point. 9. On the other hand. for a prime number by (5 Z) . First there is an involution consists of the association (automorphism of order 2) on X. we define the Hecke correspondence where the sum is taken over all cyclic subgroups H of E(Qa). it is possible lo define Eisenstein ideals (related to Eisenstein series) in other contexts when there are no cusps. Recall that E[N] is the kernel of multiplication by N on E. of J. as Marur points out. ~IWZ). define the cuspidal ideal of T to be the ideal annihilating the cuspidal group C. and [“J. of J.(N) is an abelian subvariety defined over Q.rv> A21 MAZUR’STHEOREMS 129 respondences on X.

Theorem 2. Theorem 1. 53. C. In terms of the complex variable on the upper half plane covering Y. We redefine the Hecke operators in a manner suitable for the current application.2 (Conjecture of Ogg) and Theorem 3. Let S(N) be the complex vector space of differentials of first kind on X.(N). arc called the Fourier coefficients of the form.(N)) gives an B(Q)> For the proof see [Maz 771. We call N the level. (u.(N). = The natural isomorphism map of C into B(Q) C = so in particular.(N)(Q) is finite. showing that the only rational points on X. where f is holomorphic can write this differential on 8.. After proving that B(Q) is finite.1 above. B(Q) is finite.1. on S(N) whose effect on the Fourier coefficients of a form as above is: .. Let C be the cuspidal group on J. Chapter 111.(N)(C) by the map r. as In terms of the parameter q = Pi: we The coefficients ~1. one sees as an immediate corollary that the set of rational points X. V.. For each prime pjiy there is an operator T.4.(N) are at the cusps..s quotient oJ’ J. Mazur’s proof engages in a very jazzed up form of descent.(N) defined over Q. MODULAR FERMAT’S ELLIPTIC CURVES LAST THEOREM AND We consider again the curve X.(N)\5 such a differential can be expressed + YchV(Q in the form w = f(r) dr. Then JubWQ). but Mazur still had to do substantial work to prove the full strength of Theorem 2.

If a conjecture of Hasse is true for cc(s).(N) might be necessary.3 (mod 4). If so. Hecke). that. Taniyama thought that functions other than the ordinary modular functions parametrizing X. of some special type (cf. Taniyama expressed this conjecture roughly in problems at a conference in Kyoto in 1955. then the Fourier series obtained from L. Then there is un riyenform in S(N) such that for . and L. in the sense of isogeneity. A precise statement runs as follows. and especially to decompose the jacobian variety J of this function field into simple factors.(s) may be obtained. applied to differential forms.131 If PIN. by going back this considerations. Namely. astonishing at the time. and said that he expected that the ordinary modular curves sufficed. Taniyama-Shimura conjecture. Let E be an elliptic curve over Q. number field is the zeta function of C over k. such that An eigenform in S(N) is a non-zero form which is an eigenvector for each of the operators Tn. the effect of T. it is very plausible that this form is an elliptic differential of the field of that automorphic function. is the same as the effect of the correspondences defined in $2. our new problem is to characterize the field of elliptic modular functions of “Stufe” N. around 1962. and by finding a suitable automorphic form from which L. The problem is to ask if it is possible to prove Hasse’s conjecture for C. Conccrning the above problem.(s) denote the L-function of C over k. among others. Shimura. there is also an operator T. made the conjecture more precise. have now been adopted. J contains elliptic curves with complex multiplication. It is well known. in conversations with Sure and Weil. One way to express the Taniyama&jhimura conjecture is to say that every elliptic curve ova Q is modular. See [Shi 891. satisfying q . For pjN. Let C be an elliptic curve defined over an algebraic k. Is this true for general N? In his reference to Hecke. in cast N = q is a prime number. when he wrote: 12. and let N be its conductor. He explained the role of the rational numbers Q as a ground field (as distinguished from an arbitrary “algebraic number field k” as in Taniyama’s Problem l2). 13. These ideas.(s) by the inverse Mellin transformation must be an automorphic form of dimension -2.

(N) in the 1. are integers. and how one c. It also follows that all coefficients a. Thus up is also the trace of Frobenius addition. since they are expressible as u.: 6. say of residue characteristic 1.)...lN the eigenvalur aP qf T.) = I + p .=p+l-#E(F. p”.JFr.) = T.CQ. 3. In + E defined owr Q. are integers. such that ij w~. but as Weil writes: “Nach einer Mitteilung van G. from the point of view of the functional equation of the zeta function. See for instance [Shi 71b]. which is equal to the dimension of S(N). Then T is a free Z-module of rank equal to the genus g(N) of X.GW. for this firm that #E(F. Let T = T. km) having the properties: 1. there i. When there exists an eigenform as in the statement of the conjecture.adic representations. now be the subring of End&(N)) generated by the operators TP over Z. = x. then x*<+~ is the above eigenform for the Heckr operators. §31 has the pwperty each prime p.132 MODULAR CURVES OVER Q cv. I shall follow Ribet [Ri 90b] in describing how one obtains certain representations of the Gal& group G. Then the residue field k.. We note that the coefficients a.s a rational mup II: X. det {I”.(N). is unramified at all primes p not dividing N. = T/m is a finite field. and where uI = 1 and up is the eigenvulue as ahue The special role played by the conductor arose in Weil [We 671.. mod m for all pkN. Let m bc a maximal ideal of T. Shimura” the differential of lirst kind on E corresponds to the Heckc eigenform on X.(N). . tr p. . one says that E is modular.ln combine a theorem of Ribet with Frey’s basic idea to deduce Fermat’s lasl theorem from the Taniyama-Shimura conjecture.~ is a suitably normalized d@rential cf first kind on E. 2. One can show ([DeS 741 or [Ri 9Oa]) that there is a semisimple continuous homomorphism P. the cyclotomic character (definition recalled below).

a homomorphism are isomorphic. The action of T which takes each T E T phism is in fact Z-valued. and let w whose eigenvalue under TD is ap for each prime on w is given by the homomorphism ‘p: T + C to the eigenvalue of o under T.)) = a(‘&) and det y(Fr.(Fr. + GL(2. together with the fact that a semisimple 2-dimensional representation is determined by its trace and determinant. we have o[ = gx’“‘. the direct sum of the Jordan-Hiilder This semisimplification factors of p. We say that y is modular of level N if there is a maximal ideal m of T and an imbedding I: T/m + F” such that the F”-representations Go GQ 2 Pm GL(2.. 54. + Aut(E[I]) on the points of order I in E(QB). The Frobenius element Fr.) is the semisimplification p: G. This homomorIf I is a prime number..133 Recall that the cyclotomic character is the character such that for every CI E G. then pm: GQ + GL(2. is defined to be p” of the representation Next let F be a finite field and let y: G. Fa). GL(2. and m = rp-‘(u)). F”). Let E be a be an eigenform in S(N) p!N. Equivalently. F. F) c GL(2..). which implies that all elements of the image of pm are conjugate to Frobenius elements p. F) be a continuous semisimple representation. modular elliptic curve of conductor N. for v lying above p. This follows The representalion P.)) = @ . was defined in Chapter IV. from the Cebotarev density theorem. is unique up to isomorphism.. such that tr(yFr. Example. and an I-th root of unity [. T/m) & one requires ti: T + F” GL(2.

dividing out by any common factor. After permuting these integers. Let p be a prime number.)-module. Proposition Indeed. Let y be an irreducible 2. and is weaker than what is proved in [Ri 90a]. The decomposition group D. and p is the image of p in F. c are relatively prime. we consider .) on Y(Q. One needs a test for finiteness. Suppose there is such a triple. Fr.134 MO”“LAR CURVES OVER Q P. Then y is modular of leuel N/p. with an action of F on !g making g of rank 2 over F. This theorem immediately gives: Conjecture of Taniyama-Shimura a Ferntat’s last theorem.). Let p be the reprcsentution of G. b. Without loss of generality. for some uip. suppose that conjecture true.dimensional representation OJ GQ oeer a finite jield of characteristic 1 > 2. this means simply that y is unramified at p. we may suppose that b is even and that a E 1 mod 4./Q. It suffices to show that there is no triple (a. Let E he a semistable elliptic curve OWI Q. The next theorem is the principal result of [Ri 9Ob]. Suppose further that p is a divisor IIJ N at which y is finite.1. Here.. denotes Fr.2 (Ribet [Ri 90b]). see [Ser 871. b. put in minimal model over Z. given by the next proposition. By restriction. Let A br the minimal discriminant. We say that y is finite at p if there is a finite flat group scheme !o over Z. It proves a special cake of conjectures of Serre [Ser 871.some prime 1. $31 for all but a tinite number of primes p. q # I such that y is not Jinite at q./Q. where I2 5 is prime./Q. und that there is a prime q[N. but easier to understand and sufficient for the application. Following Frey [Fr]. taking into account the primes 2 and 3. and choose I+. c) of relatively prime non-zero integers satisfying the equation a’ + b’ + c’ = 0. on Eli] for .(A) = 0 mod 1.) are isomorphic. If p # I. Note that the proposition uses the minimal discriminant. we view F* as a Gal(Q. Fir&y. and that the fully general minimal model must be used here. Let p be a prime. The representation p is finite at p q and only ij ord. such that y and the representations of Gal(Q”. we may awune that a. 3. is then the G&is group Gal(Q”. Assume that y is modular oJ squure free level N. For a proof. It shows how the level of a representation can be diminished. we need to detine the notion of a representation y being finite.. Theorem 3.

a non-zero rational number is called a congruent number if it occurs as the area of a right triangle with rational sides. and the minimal discriminant in a minimal equation to be A = (abc)“/2’ We then obtain the representation p of G. V. which is proved to be irreducible by results of Mawr [Mu 773 and Sure [Ser 871 Proposition 6. The problem whether a number is congruent or not can be reduced to a problem about elliptic curves and modular forms as follows. 1. and we shall give a brief summary of some of his results and the context in which they occur. APPLICATION TO PYTHAGOREAN TRIPLES Classically. z) with L) # 0.1. By using Proposition 3. Dickson in [Di 201 traces the question of whether a given number is congruent back to Arab manuscripts and the Greeks prior to that. these two hypersurfaces in PJ intersect in a smooth quartic in P” which contains the point (l. so tinite at p. we deduce that p is modular of level 2.~ + b’). so the Hecke algebra is 0. 1) to the plane z = 0 gives a birational isomorphism with a plane cubic whose Weierstrass form is ED: ~y2= x3 ~ D*x or also Dy2 = x3 . Frey already had noted that if p # 1 and p divides N but p # 2. then p is unramified at p. it is clear that a rational number D is the area of a right triangle with rational sides and hypotenuse h if and only if (h/2)’ + D are both rational squares. One can compute the conductor of E to be N = ahc. $4. Applying Theorem 3. v. Tunnel1 [Tu 831 showed that this old problem is intimately connected with coefficients of certain modular forms. From the Pythagoras formula. Geometrically. This is impossible. because S(2) has dimension 4(2) = 0.O.O.2 inductively. w.Dv2 = z2.x. Hence D is a congruent number if and only if the simultaneous equations u= + Do2 = d: u2 .1. have a solution in integers (u. on E[l]. . $4. 1. one sees that this representation is not finite at 2. and we are done.the elliptic curve E defined by the equation y2 = x(x ~ a’)(. 1). and projection from (l. The intersection is thus an elliptic curve over Q. the representation is modular. By TaniyamaShimura.

(q) associated with beginning of $3.-(E) has rank 2 over Z.The points on the space curve with v = 0 correspond to the points where y = 0 and the point at infinity on ED. Thus we see that: D is the area qf a rational right triangle if and only if the group E”(Q) is infinire. We have the corresponding function half planc. and satisfies an equation as above.J~ class number 1. We recall that an elliptic curve E is said to have complex multiplication if End. We bit nmre explicit. Suppose a function f on .!j is defined by a power series with u1 = 1. at s = 1. see [Sh 731. but with exponent 3/2 instead of 2. Shimura [Shi 713 proved that over Q with complex multiplication is modular. then the L-function L&) has the form . If L. Then we say that . along the lines of the Birch-Swinnerton-Dyer conjecture [COW 773: Theorem 4. On the other hand. normalized as in the statement of the Taniyama-Shimura conjecture in $3. If E is a modular elliptic curve over Q. A curve ED define&by the above equation with D # 0 has complex multiplication by Z[& 11. and be used to analyze the behavior of L. A test whether ED(Q) is finite comes from applying a result of Coates-Wiles.f((ar + h)/(cT + d)) = (CT + d)‘f(T) for Thus one says that / is a form of weight 2. and uly is a differential of first kind.1.(l) # 0 then E(Q) is finite. It is easy to see by reducing mod& primes that these points on E”(Q) are precisely those of finite order. and so is a subring of an imaginary quadratic field over Q. and this function satisfies the equation every elliptic curve the modularity can make the matter a a form as at the f of t in the upper .f is a form of weight 3/2. Consider the power series j. Let E he un elliptic curve over Q with complex multiplication hy the ring of integers qf a quadratic field I. and a sign factor which is relatively complicated.

which describe the group of rational points module torsion when this group has . Thus the value of the L-function at 1 is a non-zero multiple of coefficients of gH. Theorem m (x3 . we consider conjectures and results related to the BirchSwinnerton-Dyer conjecture for elliptic curves over the ration&.3. and go.62205 As Tunnel1 also remarks. Furthermore. let 9 = %A) = 4 “gl (1 . lar properties the behavior of LE(s) at s = I can be deduced from moduof E..2. MODULAR ELLIPTIC CURVES OF RANK 1 In this section.P”) and for each positive integer t.. we have Tonnell’s theorem. we haue Theorem 4. gii. 1) = b(d)‘/l(2d)-“‘/2. are modular forms of weight 3/2. Putting everything together. L&P’. where /I = !-. and go. V.P)U . the main result of [Tu 831: 4. let Let Waldspurger [Walp 811 showed: integer.x)-l” dx = 2. the Birch~Swinnerton-Dyer conjecture implies the converse of the statements in Theorem 4. For d a square free odd positive L(Ed.In particular. If b(n) # 0 then 2n is not the area of any right triangle with rational sides. $5. Indeed. If a(n) # 0 then n is not the area of any right triangle with rational sides. 1) = a(d)pd-“‘/4. is the real period of E.3 when n is a square free posilivc integer.

we .Q(wJ = 1 dwJ. so We begin with a construction of a certain rational point D be the discriminant of an imaginary quadratic field K = Q(. Then there exists an ideal n of oK such that oK/n = Z/NZ.(N)(C) represented over the complex numbers by the pair (C/a. o. We shall now consider when Q. we define the Heegner point on E to be the trace of ITX.(N) + E curve over Q. the point Q.* is normalized as in that statement. = T*I. depends up to sign only on the parametrization n: X. but not on the ideal n and the class a. -‘%Q It is easily shown from the basic theory of modular curves and complex multiplication that in the group E(Q) mod torsion. an-‘/a). to the ration& that is Q. with ideal class a.(N) + E is normalized so that the cusp (m) goes to the origin in E. that is z(c4) = 0. So we let E be a modular elliptic parametrization n: X. n) (depending also on the choice of n) to be the point in X. is a torsion point. and in particular.. = ~(a.(N) + E and on the choice of D. where H is the Hilbert class field of K. with a over Q as in the Taniyama-Shimura conjecture stated in $3. In the canonical model X.. in E(Q). if ob E GNIK is the element of the G&is group corresponding to the ideal class b. By the theory of complex multiplication and class field theory. Let We assume throughout that D (and so K) is chosen so that every prime p(N splits completely in K.rank I. As before.b). The differentlal form wE. it follows that We assume that the rutiunul map II: X.(N). Let n be an ideal of oK. We define the point x.Jn is cyclic of order N. E WVW). Then following Birch. it can be proved that x.

x . Let w. then Q. in E(Q) has infinite order if and only if the following three conditions are satisfied: g.2 ([GrZ 861). and in par- .. If E = . and the rank of E(Q) is 1.1. then E D is defined by the Weierstrass yz = x3 .x .1 (Gross-~Zagier [GrZ 861).) is jinite.2. 0 L(EB 1) # 0.1 essentially comes from the factorization formula of Theorem 5.1(c). If the Heegner point Q. Theorem 5. UI(E. The canonical height h(Q. = w be the involution Hecke operator as described in $2. is a torsion point. It will play a role in the next conjecture where the order of the finite group ED(Q) enters in the formula. Let E be the sign of the functional equation as in Conjecture 6. The point Q. and under these conditions.. -4.. has infinite order.-I.define the twisted elliptic curve ED: If is a Weierstrass equation equation for E.D3y. if E = 1. one can describe E also as follows. The three conditions are natural if one thinks in terms of the factorization of the L-functions associated with the elliptic curves. Thus we get necessary and sufficient conditions under which QD has infinite order.J = -&J Theorem 5. then jtir D # -3. Gross-Zagier have constructed a non-torsion rational point. namely one has the formula: Theorem 5. so Theorem 5.3 (Kolyvagin [Koly 881).y..1 of Chapter III. The theorem is predicted by the Birch-Swinnerton-Dyer conjecture and the condition in Theorem 5.D’y. then E”(Q) is finite. or DyZ = x3 ~ y. Then W.) is 0 if and only if Qn is a torsion point. In terms of the modular curve. (c) . This conjecture describes more properties of the Heegner point. In particular.

2.2 and also [Gros 901). Most of Conjecture 5. one is led to the following conjecture on the index of the point PD = Tr. depending on the action of Galois on JYWP(O To the extent diophantine geometry is concerned with rational points.140 MODULAR CURVES OVER Q = dx/2y. then [GrZ 861. The integer c(n) is interesting independently of the context of elliptic curves of rank 1. If (E(K) : ZP”) = #WE. Conjecture 5. 1) with the Birch--SwinnertonDyer conjecture.. see Tate [Ta 741.... We recall some notation..n there is also the form o+min. as before be the integers in the BirchSwinnerton-Dyer conjecture (Conjecture 6.4 has been proved by Kolyvagin [Ko SS].. the minimal (N&n) differential associated with the minimal model of E over Z. See Remark 1 below.).2 of Chapter III). For the definition of this minimal form in general. in an appropriate sense. then Wg. because Manin [Ma 721 conjectured that in every isogeny class of modu- . To my knowledge. If one combines the formula for L’(E. we see that the construction of Heegner points provides an explicit way of getting part of the group of rational points when the rank is 1.4 (Gross-Zagier PD has infinite order. V. 551 titular gives the index of the subgroup generated by this point in the group of all rational points. If this minimal model has the Weierstrass form. then (E(Q) : ZQD) #ED(Q) = #UI(E. ‘2’ PIN where t is an integer with ltl 5 3. no construction is known or conjectured today to give a subgroup of finite index when the rank of E(Q) is greater than 1 for modular elliptic curves.r&~.. A corollary would be that when the three conditions of Theorem 5.)%(n) g cp. It can he shown that there is an integer c(n) such that The integer c(x) describes the extent to which the parametrization x is not minimal. Remark I. It is sometimes called the Manin constant.1 are satisfied. For each p[N we let c. Aside from the form uE. see also [Gras 901.)“2c(n) n c.

there is a curve whose Manin constant is 1.f if for all primes lJ4N.. The explicit construction of the Heegner point corresponds to the explicit construction of cyclotomic units in the cyclotomic case. Stevens [St 891 refines this conjecture to the case of parametrizations by X.5 ([GrZ 861). Taking the trace corresponds to a similar operation in the theory of modular units.(N) rather than X. in which case he conjectures that every modular elliptic awe admits a parametrization by X. Theorem 5. for h. which is Shimura correspondent n = IDI is given by to f. g is an eigenfunction of T. E Z he the integer such that N splits completely in Then there exists n modular ~vrn g of weight 3/2 invariant by r. The Gross-Zag& construction depends on the choice of an auxiliary D. = m.(N) with Manin constant equal to I. (the 7. 1) # 0. which give rise to a subgroup of the cyclotomic units as in [KuL 791. so that E(Q)/torsion = ZQ. For convenience. Remark 2. or the more recent formulas pertaining to the modular units. .cv> §51 MODULAR ELLlPTlC CURVES OF RANK 1 141 lar elliptic curves over Q.-eigenvalue off). and such that the coeficient b. I shall reproduce one simple statement due to Gross -Zag& which explains some of this dependence.. with Q(fi) some generator Q. It is not clear today in general under which conditions such formulas should exist. with eigenvalue a. Suppose we have a modular form of weight 3/2. Furthermore g # 0 if and only if L’(E. but the reader should be aware of a very broad formalism concerning such index formulas in which the Mordell-Weil group plays the role of units and III plays the rble of a class group.. we make a definition.. and of modular units. with q-expansion We say that g is Shimura correspondent to .(N). Let N be a prime number..(4N). For each D such that lel m. Assume N prime and rank E(Q) = 1. The question may be raised about more precise information on this dependence. The index formula above follows the same formalism as the classical index formula for cyclotomic units in the group of all units.

142 MODULAR CURVES OYER Q cv> §51 Although this section has because it gives prototypes some of which are not even Gross for his advice and help been fairly specialized. I am much indebted to in writing this section. . I have included it for possible much more extensive results. conjectured today.

The group of Cartier divisor classes Pit(X) is isomorphic to the group of isomorphism classes of line sheaves on X. and a morphism onto a curve n:X-tY defined over an algebraically closed field of characteristic 0. which are of independent interest since they exhibit the diophantine geometry in a context independent of more refined arithmetic invariants found in the number field case. BASIC GEOMETRIC FACTS In this chapter we assume that the reader is well acquainted with basic properties of intersection theory on surfaces. Let X be a complete variety. there is a geometric analogue to the theorem that a curve over a number field has only a Iinite number of rational points. VI. There have been several methods in the function field case to obtain such bounds. The original proof of finiteness (without explicit bounds on heights. This isomorphism is . The height of these sections has a geometric definition. We consider a projective non-singular surface X. so the generic fiber is a non-singular curve over the function field of Y. and we want to give bounds for those heights.CHAPTER VI The Geometric Case of Mordell’s Conjecture As we saw already in Chapter I. The purpose of this chapter is to describe some of these methods. $0. conjectured in [La 60a]) is due to Manin [Man 631 and the ideas of this proof will be given in $4. Rational points of this curve over finite extensions of k(Y) amount to sections of this fibering over finite coverings of Y.

(D) induces the isomorphism. Z). By a vector sheaf 8 we mean a locally free sheaf of finite rank. if D is represented by the pair (U. In particular. The degree on a the degree on the pull back to the is the normalization of X. A is denoted by (3. Then the intersection number (c. then we define Z(D) = 9 0 Q(D).(D) over U are given by The association D H @l. q?p) an open set U. itself. So if f: x’ +X definition de&) = Suppose that X has arbitrary dimension. or the class of D.A!). Then O.f *(c). on then the section of O. Z) is defined as (c.3’) on a surface. .f: z’ +x be the finite morphism. The intersection symbol extends bilinearly to a pairing between Pit(X) and the group generated by the cuwes. and is non-singular. Z) = deg .p is a line sheaf in the class c. where c is the class of Y in Pit(X) singular curve X is defined to be normalization X’. so X is what we call a surface. We often write (9*) instead of (9. we let the determinant be de(J) = A’ 8 = A’““” 8. we define (9.144 THE GEOMETRlC CASE OF MORDELL’S CONJECT”RE cw WI given as follows. Z) = (c. then we get a pairing between line sheaves into the integers called the intersection pairing. The tensor product is taken over 0. Suppose X is a curve. If 1. Let D be a (Cartier) divisor on X. but Z is a curve on X.(D) is the sheaf such that. then by deg f*(c). such that Z’ is the normalization of Z. If this rank is r-. The intersection number of two line sheaves 3. If 9 is a line sheaf on X. If 9 x O. if X has dimension 2. Let . . and f is the composed morphism Z’ + z c x.(D) we define the degree de@) = deg(D) = de&).

and such that X is a local complete intersection in S. We denote this generic . we must go around this possibility. then .$. Since n is not necessarily smooth. such that X is a complete non-singular surface. if U is an open subset of X such that the restriction of n to U is smooth. The determinant is the maximal exterior power. = x. It is easily shown that the canonical sheaf is independent of the imbedding. If 71 were smooth (which it usually is not). THE FUNCTION ITS CANONICAL FIELD CASE SHEAF AND Let Y be a complete non-singular curve algebraically closed field k of characteristic of genus 0.y = det j*R&. up to a natural &morphism. then this line sheaf would just be CI.. Since X is assumed nonsingular. function field he the function The first thing to do is to describe a canonical lint sheaf on X. $1. VI. where field q is u yeneric point of K and L = k(X) of Y. WC then define the canonical sheaf of the imhedding j to be w x. where +Z& denotes the dual sheaf. The conormal sheaf is the vector sheaf 6. and we suppose that g 2 2. Let q dcjned oum an be u jlat proper morphism. This means that the ideal sheaf a defining X in S is generated at every point by a regular sequence.s = BIP. We let g be the genus of the generic fiber.. the ordinary sheaf of differential forms. 0 det q.Such determinants already occurred in the discussion of the Fakings height. but they will occur from now more frequently and systematically. Furthermore. there exists an imbedding of X into a smooth scheme S over Y.jiber by c = n-‘(q) We let F = k(Y) of x.&.

. . The line sheaf (Oxi. In fact. A point morphism so that a.... is a subsheaf of the constant sheaf with fiber 61xiv.. correspond to a divisor class which is called the relative canonical class. is independent of the factorization X c S + Y.Z z R:. making the Y’ - / A Y and such that sP is generically an injection. For such a covering Y’. The image of O.. = @AK).. let z be the generic point of X.z = at..... of k(Y) corresponds to a P of C in a finite extension s = sp: Y’ + x where Y’ is a (possibly ramified) diagram commutative: non-singular X EI r covering of Y. as a subsheaf of the constant sheaf fit..... = det(j*%.2).There is a canonical choice for CO. we define the geometric logirrithmic d(P) = rp 1 :~+2q’ discriminant . deg SW. which will describe two approaches to prove the boundedness of heights of sections of the family x X + Y.. 0 deV&)... we Id q’ be its genus. in n. by The above notation and definitions will be in force throughout $2 and $3.). = det & ‘which gives us our imbedding.. is exact and fl:.Y.... We define the geometric canonical height by the formula 1 h(P) = ly’.. which we denote by K = Kx. we obtain a natural isomorphism %. Then W. Finally...

Theorem 2. The bound on the right-hand side is remarkable in that the factor of d(P) does not depend on the genus of the generic fiber. we have k. 52. Before we go into the ideas of the proof. We apply this to the case of our surface X. Chapter II. and similarly for @. we recall a basic fact from clcmentary algebraic geometry [Ha 771. Proposition 7. then the above construction is meant to be applied to its desingularization.(P) 5 (2 + c)d(P) + O. and let s: Z-)X be a finite morphism.. To give a morphism Z + P(R) of a scheme Z into P(&‘) over X.(I). See also [Vo gob]. . a line sheaf 5? on Z. Vojta’s proof is then based on the following construction of Grauert. If Z is a singular curve with a finite morphism into X. Then we obtain making the following diagram the corresponding mapping t. Given e z 0.s*Qi in 0. as usual in the theory of surfaces.VI. Vojta’s conjecture is that this factor can be replaced by I + E.: Z +P(@) commutative. for all algebraic points P of C d@ined by our morphism mer k(Y). We have abbreviated 0: = Q:. is equivalent to giving a morphism s: Z -+X. Let C be the generic curve <f a family K. where Sym 8 is the symmetric algebra.12.1 (Vojta [Vo 9Oc]). Let Z be a non-singular curve. one can form the projective bundle p: P(8) +X defined by P(f) = pr&ym(O). The map f. We continue with the notation of $1.. which is a graded algebra. Given a vector sheaf 8 over a base scheme X. GRAUERT’S VOJTA’S CONSTRUCTION INEQUALITY AND In the function field case. Vojta was able to give a remarkably good estimate for the height of algebraic points as follows. We get a homomorphism where 2 is the image of . and a surjective homomorphism of sheaves s*& + 55’ on Z.

1 follows from (*) and Proposition 2.(D)).2. Vojta then proceeds in two steps. and let w E H”(X. In the first place. Let X. Q:(D)).(l) is the hyperplane line sheaf on PBY Note that P(@) is a variety of dimension 3. one must dig deeper.2.p*W. The next theorem provides a key finiteness result Theorem 2. which when represented by a pair (U. n.(l)) + c deg.. Let deg. We may then think of w as a rational differential l-form on X. Let D be a divisor on X.1.2 for those points P and corresponding section s = sp such that t. Theorem 2. Then: (I) There are i@nitely many irreducible Pfqfhan dioisors with respect to o if and only if w is of the form w = q d+.. letting 8 = Cl: we haue (Z. = 2g(Z) . Let for the moment X be an arbitrary projective non-singular variety in characteristic 0. To prove Theorem 2. in a way which leads to solutions of certain algebraic differential equations. Z. of Z and the generic fiber of z 0 p. and as usual. fu) on an open subset U of X has the property that u. where 0. We describe the main step. Fix a rational number E > 0. There exists ~1constant c and an @‘ectiue divisor D on P(Cll) such that for all irreducible curues Z on P(@) not contained in D. cm 921 We let csscntially PC? = P(C&).(Y) is not contained in D.(Y) is contained in D. O. For those points P such that t.. for some rational functions ‘P and $. taken in the projective The sheaf 9 is the inverse image bundle. co and D be as ahwe. .3 (Jouanolou [Jo 781)..y) 5 (2 + E)(Z. The above is Grauert’s construction.I48 is THE GEOMETRIC CASE OF MORDELL’S CONJECTURE the differential of S. where H” denotes the global sections. one has the inequality (*I deg t:@&(l) 5 deg 0. We define a pfaffian divisor W with respect to w to be a divisor. Z = intersection Then: number Proposition 2. A y ” E H’(U.

Let X br a non-singular projective surface.) This method depends on bounding the self intersection ((I&). even conjecturally. NS(X).5). Indeed. regular. Then the set of irreducible cwws Z on X which lift via the Grauert construction to curues contained in D is a union of finitely many algebraic families. Corollary 11. Y in mere generality than the basic situation described at the beginning of $1.e. it does not give a bound for the degrees (or heights) of solutions. Then an . X is integral. is base change By a theorem of Zariski. irreducible (i. using the Grauert construction.1. is known today in number fields. which we must now define. WC say that X is regular semistable over Y if the following conditions are satisfied: SS 1.2 conclude the proof of Theorem 2. Let o be a discrete valuation ring with quotient field F and residue class field k. like other proofs in the subject at the moment. so Corollary 2. a:)] &weri group + p + 1. Let Y = spec(o).4. then their number 149 is bounded If there we . for which no equivalent. For this general definition.4 and Proposition 2. 53. Neat bounds are obtained under an additional condition besides the basic situation described in $ §31 (2) PARSHIN’S METHOD WITH ((!&) many such diuisors. One limitation of the above proof.jinite/y by dim[HO(X. Intersection numbers and degrees for elements of a family are constant in the family. and let X + Y be a flat proper morphism.1 is an immediate consequence of the following corollary. The scheme geometrically from F). VI. Q:(D))/u where p is the rank of the N&on A HO(X. Then Theorem 2. we temporarily use X. Corollary 2.. which we assume perfect. remains and the generic irrrducible under fiber X. is that it involves horizontal differentiation. PARSHIN’S METHOD WITH (W:. Although the proof bounds the number of solutions (to a dilTerentia1 equation). Let D be an @ctioe diuisor on P(Qk). of relative dimension I. an algebraic family of curves in X lifts to an algebraic family of awes in P(Qi). namely semistability. it follows that the special fiber is geometrically connected ([Ha 771 Chapter 111. We define the geemetric fiber above the closed point y E Y to be the base extension of the fibcr to the algebraic closure of the residue class field k(y).

The geometric fiber above the closed point is reduced und has only ordinary double points. or Y is a curve over a field k. and it is also preserved under unramified extensions of flip.irrcduciblc component of the fiber X. then we say that X + Y is semistable if it is semistable over the local ring of every point of spec(R). we denote by X” the corresponding stable model. If Y = spa(R) where R is a Dedekind ring. and he proved that given an abelian variety over a Dcdckind ring. Semistability is preserved by making a base change from Y to spew where my is the completion. Let C be a curve over F. there always exists a finite extension over which the abelian variety is semistable. WC do not assume that X is regular. There is a hijection between stable and semistable models. rational curves with self intersection -2 can be blown down. = C. See also Artin-Winters [ArW 711. except for the following two modifications: In SS 1. Deligne-Mumford [DelM 691 proved the corresponding result for cuwes. Grothendieck defined the notion of semistability for abelian varieties. Note that C is semistable if and only if its Jacobian over F is semistable. If X/Y is a semistable family of curves of genus g > 2 and y is a closed point of Y. then Z meets the other components qf the geometric jihrr in ut leust two points. we also have the notion of stability. resulting in a semistahlc model. The singularities of a stable model can be resolved be a sequence of blow ups in a canonical fashion. splits into a finite conjugate geometric components in the geometric fiber. If Z is a non-singular irreducible component of the geometric . Using Grothendieck’s theorem. In addition to semistability. SS 3. number of SS 2. The definition is the same as for semistability. a non-singular rational component of a geometric fiber must meet the other components of the geometric fiber in at least three points. then we let . resulting in a stable model. Conversely. In SS 3. If X is semistable.fiber. We say that C is semistable if there is a regular semistable X + Y such that X. and Z has genus 0. so X may have singularities. Suppose now that Y is either spec(o) for a discrete valuation ring or Y is a complete non-singular curve over an algebraically closed field of characteristic 0.

has bad reduction. Parshin and Arakelov work not only with (W:. For the rest of this section. for (O&r) or related objects Arakclov proved [Am 711: will be called canonical class CC 1. Miyaoka.2 + 4 ~ 6. Let F = k(Y). Bogomolov.WY .Wq A variant was obtained cc 2. Then (@:. I find Vojta’s proof of CC 3 in [Vo 887 considerably easier to follow than previous references. Yau.g.[VI.? = number of double points on the geometric One has the upper bound of Arakelov Proposition 3.3 [Am 711: fiber over y. = &. Parshin. by Vojta [Vo 881. unless otherwise specified: We Ief X + Y be a semistable family of curves of genus g 2 2 ouer a complete non-singular cww Y of genus y.1.2 + s). 151 of double points on the geometric 6 = I?. [Am 713 or [Szp 811.) fiber over y.. $31 6. = number PARSHIN’S MET”“” WlTH (W$.. (co:. Let s he the number of points of Y where X. arising from the work Ven. . Let y0 be the dimension f$ the F/k-trace r$ the Jacobian (f a generic fiber X. We now return to the basic assumptions of $2..) . Y If u is the valuation corresponding to the closed point y of Y. Y + (28 ~ 2) max(2y ~ 2. in order to get neat bounds for various objects..O). 6: 5 3y . e. of van de which results from another inequality. then we let a.. See also Parshin [Par 891. we also write 6. 5 (29 .) 5 3 cd..g. who gives cw:.. but with the added hypothesis of semistability. and Arakelov: cc 3. all owr an alyehraically closed field k of characteristic 0.r) C 6k .. Inequalities inequalities. On the other hand. If Xx/Y is a stable family and y is a closed point of Y. = 1 c$.

= deg det x$&. The inequalities H 1 from Szpiro and H 3 from Esnault Viehweg are not only effective but completely explicit.. Parshin observed that the canonical class inequalities inequalities. which and with the line sheaf det K$&. from quite another H 3. Szpiro’s proof is completely algebraic (valid even in characteristic p).. in [Vo 881 and [Vo 90bJ .6 H 2. k. one has: in [EsV 901. is known today. CC 6. by a result of Xiao Gang [Xi 871.1)’ s+ 1+ F + .I.. imply height notably: 8+ 4 9> deg GJx.... whereas Esnault-Viehweg approach the problem from the direction of semipositive sheaves and their proof uses certain aspects of complex analysis for which no immediate substitute over number fields. We have the relation cc 4.. namely: cc 5. are of the same order of magnitude.. deg Q-Ly. (O:... Vojta’s inequality H 2 has such a factor of d(P) linear in 4. see also Cornalba~Harris [CorH 881..(P) 5 8.) and deg n*W. (a variation sheaf of rank 9.(P) 5 2(2y .2 + s).. h. h. rather than exponential. Several variations of these have been obtained. but it is not clear from Vojta’s proof how effective is the term O(l)..152 THE CEOMETRlC CASE OF MORDELL’S CONJECTURE is a vector [VI. Furthermore. even . $31 but also with the direct image x*W.I.... of No&her’s formula) (CO. Note that H 3 improves H 1 in that at least the factor involving y is quadratic in y.I)‘(d(P) + s) direction.. obtains S 8s .3”+‘(g .. Vojta argues by using a refinement of the Parshin construc- .. H 1. We define the degree deg @.) + d = 12 deg n$l. Arakelov 65 ( For n*O. > in [Szp 813.(P) 5 __ 3 88 d(P) + O(1) On the other hand.~.

..(U) . It is based on horizontal dit%xentiation. .(l) is not made explicit. or possibly h. On the other hand. MANIN’S METHOD WITH CONNECTIONS The method of this section historically gave the first proof for the Mordell conjecture in the function field case [Man 631. a fundamental problem is to determine simultaneously constants b. As b..(P) 5 b. De Rham cohomology We usuully write Cl. the number ho increases. We first give the definition for the de Rham cohomology group H&(X/Y). relying on Coleman’s account [Co1 901.. $4.. So the final word on all these inequalities is not yet at hand. Indeed. and the problems is to determine the best possible set of pairs (h. Let F = k(Y). From this point of view. b. and Cl:.WI. d(P) + h.. b. From the point of view of keeping b. By a I-cocycle y = {(w~. namely 4dP) 5 (2 + 4 d(P) + O. Of ccwse. It is already a problem to get an inequality linear in y with explicit O(1). decreasesfrom a function of the genus to I + E. independently of the genus. 641 MANTN’SMETHO” WITH CONNECTICINS 153 tion on the canonical class inequality CC 2. But geometrically. the Vojta conjecture does not supercede other inequalities involving functions of the genus as coefficients. I am much indebted to Coleman for his useful suggestions. Vojta’s inequality which we gave in $2. an absolute constant. Let Y be an uffine non-singular cum mm an algebraically closed Jield k of characteristic 0.(l). so that the inequality h. it is still giving rise to many investigations./~. and is even ineXxtive as things stand today. and Cl: instead of Cl:. Let be CIproper. smooth schemeover Y. again the term O.(d(P) + s) + b.~)} WL mean a family of elements 0” E Q:. Vi.. We shall now describe this method. whence the problem about O(1). holds for all P E X(q).) in R*. and to this day no analogue is known for it in the number field case. the question then arises whether a linear function of the genus is best possible for b. is even better in so far as the coefficient 2 + I: of d(P) occurs.

For simplicity. finite and smooth over Y We define the relative cohomology group but with the additional restriction on the vanish on Z. Z) in the same way. map satisfying the Leibniz rule V(w) = du 0 w + uV(w) By a connection v:MHR.=O.f. Since X is smooth over Y..=f. we have an exact sequence . Let Z under x..(Y) 0 f&(X/Y).“.x(Y)@A4 on A4 we mean a k-linear f?r u E k[Y] and o E M. indexed by pairs of open sets in the covering satisfying the following conditions: (1) (2) (3) dw." = f" .indexed by the open sets of an open covering.andf. In the applications. We have fo. and a family of functions fo..v +f. so wI. is closed..onUnVnW. so wu is a l-form over U. we let t E k[ Y] be a non-constant function.. functions always be achieved after localizing.(U n V). such that for each U there is a By a coboundary.wy = dfu. We shall give the definition in such a way that it applies later to another situation. The factor group of I-cocycles module the subgroup of coboundaries is the de Rham group H&(X/Y)..” E O. We view H&(X/Y) as a k[Y]-module. we mean a cocycle function .(X/Y. We shall define the Gauss-Manin connection V: fL%XIY) + Q:... A4 will consist of modules of dilTerentia1 forms of various sorts.(U) satisfying 0" = dfu and f". which won’t at&t the be a Zariski closed such that n(Z) = Y.. H&. = 0. This can theorem. and we assume that H’R: = k[Y] dt." and . V we have wu . Connections Let A4 be a k[Y]-module. E O. fu.fl that they subset of X. relative to Y. For all pairs U.f".

One uses the sheaves: %z ai. s2 be sections of x and let Z be the Zariski closure of the of sI and s2. We also use the exact sequence instead We follows.(U) such that Then {(vu.. and and there exists vu t Q.. maps to 0 in the map Qi + R&. g”. on Z. /“. .X maps to 0 in the arrow &-ov”-dfu.. Let images of ES 1. apply these connections to the case when Z is constructed as S. we can lift wu to a form ~6 E Q:(U).u if w lies in HkR(X/Y. 0 + cl:. and 0: respectively.} we can find a cocycle {(au. Next suppose Z is a Zariski closed subset of X. The collection (do$} is a family of local sections of the sheaf n:..“)} representing w such that by the surjective map in the above exact sequence. which occurs as the middle term of the exact sequence ES 2. c&” n: + n: + cl:. so we have the relative group H&(X/Y.. E.. For a sufficiently tine open covering {U. Z). on Z forms vanishing instead of 0.)} is a cocycle. we can achieve that the .w. After localizing on Y. E C&(U n V) such that Hence for each pair U. We may then define the relative Gauss-Manin connection Exactly the same construction using functions which vanish defines V. smooth and finite over Y. which represents VW by definition.. dw.. = sheaf of functions = sheaf of differential vanishing on Z... g. + 0. Z) previously defined. Furthermore. $41 MANIN’S METHOD WITH C”NNECTl”NS 155 Let w E H&(X/Y).. V there exists Q$ + a&.

. Z) making ES 3. and the association which sends rp to this class is an injective homomorphism 0 + k[ Y] -t H&(X/ . We fix a derivation a: k[Y] --t k[Y] such that Dql. we could take d = d/dt. Let 9 be the algebra of differential with . Taking the vertical maps (connections) the exact and commutative diagram connections. We have two natural injections giving rise to a commuta- .156 THE GEclMETRlC CASE OF MORDELL’S CDNJECTURE cm $41 images of s1 and s2 do not intersect. operators which are finite sums = k[Y]a. f""sl=cP Then the families {(&. Z). To define the map of k[Y] into the de Rham group. For each point of X pick an open neighborhood U and function fu E &y(U) such that and fu 0 s2 = 0. let 9 E k[ Y]. For E k[Y]. Horizontal diiferentiation which Z) of of into account.. Z) H&Af/Y) -0 the following diagram exact and commutative.f. We shall define an injection WI 4 HAMY. we may speak ES 3 as an exact sequence A connection allows us to differentiate horizontally in a manner we now describe. and that Z is smooth.)} define a cocycle. so we have the Gauss-Manin connection in this situation.. 0 k[Y] f H&(:/Y. whose class lies in H&(X/Y.

~‘. that is PF = ker@ to be the kernel of the first 6 Ho@. we obtain homomorphisms 9 0 H’R:.(X/Y) and PF 9 @3H’R:.(X/Y) wHv(a)w.:0 H’R:. KS. + H&.157 tive diagram: Namely. 8) E H&(X/y. s2). such sets U. Let c( E H&(X/Y. be- defined as follows. and similarly without Z). that for to o the There a global all open class of are also differential form o gives rise to a cocycle {(o. mod H”R:. Z). + H&(X/Y. By linearity and by the natural injection of H’R$.~ choice s = (sI. so The image of D in k[Y] depends we indicated s in the notation. map the Z. -+ 4hWY). homomorphisms (depending on a connection V): V(a): f&(X/Y) VA% f&(X/Y. Then PF has a natural image in H&(X/Y. We define the Picard-Fuchs group pairing. . independent of Z. + H&(X/Y)). We define the Kodaira-Spencer + H&. and it follows from ES 3 that the image of PF lies in k[Y].. Z).. Z). We associate this cocycle in the de Rham group. Z) + H&WY. we get the operators V. Z). 0)}. V we have uu = LU and fLI.. say. into the de Rham groups... so we obtain a homomorphism PF + k[Y] which we denote by on the original DH. by We shall see later the significance of the Kodaira-Spencer map. They from the duality tween derivations and differentials..” = 0.(a)a = (Vzcz..

only Theorem 4.. On the other hand. and the association then applies to all choices of sections into k(Y). Theorem 4. The association (D. Let (8.sryuence ES 3 (us a sequence of connections) splits $ and only if there exists un integer m > 0 such that ms E rB(k).S is u bilinear map PF x A. + Wk).For each choice of sections (sI. Hence we shall view the function fD.) of We may further describe an imbedding of the group in a finite product F x F x . in the context of an abelian scheme when the sections form a group.(F) + k(Y) The set of sections s such that JD.. See also Manin’s letter in Izvestia Akad. For the application to Mardell’s conjecture in the function field case. Theorem 4.. giving rise to the pair (s.. For this we use the horizontal differentiation and the functions SD. be a section and we let s2 be the zero section. and we use s to denote a section.. Abelian varieties (sl. Using this ..2 (Theorem of the kernel). = 0 for all D E PF is precisely &UT. x) be the F/k-trace the generic fiber A.. Manin claimed to have proved the theorem of the kernel in [Man 631. Manin’s work gave rise to further work by Deligne. Coleman found a gap in Manin’s proof [Col 901. Nauk. but a quarter of a century later. s2) as a homomorphism of PF WC apply ES 3 to the case when X is a family of abelian w&ties over Y. Let s be a section. In that case. x F. 1990. 0). from now on as an element of the function field k(Y).1 was needed. s2) we had to localize on Y to insure the smoothness of Z. The . S)++f”. we let s. A(F)/(rB(k) + A(F).. Let A/Y be an abcliun scheme. and could be proved..1.

X is stably split. Theorem 4. the KodairamSpencer for i = 1..D. The next proposition gives a seemingly weaker criterion Kodaira-Spencer map to be 0.. Suppose K: X + Y be a proper that: smooth family for the of curues of There exist sections of arbitrarily large height.?fD. E H’R. and hence B lifts to a derivation is stably split.. The Kodaira-Spencer map is 0.4. We say that the family is stably split if there exists a finite morphism Y’+ Y such that the base change X. Then words. = k[Y] §41 MANIN’S METHOD WlT” CONNECTl”NS 159 work Chai was able to prove the theorem of the kernel completely [Chai 901.)“. we get: Corollary 4. Lrt genus (a) (b) 2 2.TH(fD gives an imbeddiny c$ A. we shall use a differential criterion for the family to split. such that if V is the Guuss-Manin connection.... Proposition 4.. the Kodaira-Spencer so the family mup is 0. WC return to the derivation a such that Dcr. There exist a finite . that is there exists a curve X..s. the factor group of A..5. x Y’ is birationally equivalent to X. We pass on to a smooth family of curves X + Y of genus 2 2. 2 map is 0 on a 2-dimensional in (H”%. . ..s) mod A(F). o2 E H’Q$. and to the Kodaira-Spencer map in the present context. In other space. ... over k such that X.(F) by its torsion group and rB(k) is finitely generated. is split. Following Coleman.. Since by the Lang-N&on theorem...(F) number of d&xntiul operators D. then ooer k[ Y] V(J)w. linearly independent There exist forms w.. The following conditions are equivalent: which lifts d to a deriva- (1) (2) (3) There exists a deriuation dX E (H”C&)’ tion in the dual of Ho@... + sB(k) into k(Y)“......3. such that the association .

. E PF. Proposition 3. When one cannot apply the situation of Proposition 4. so that the kernels of the Kodaira-Spencer maps have the same dimension for all n.finite dimensional vector subspace of the . namely: proposition.. = G&W’).5. v(a)‘w co2 E H”C&.6. Given o E H”R1 B. are compatible. be the Jacobian of C.4. is used to bound heights. + P. which is the kernel of 9 8 H’R’ + H&. 7. Let V br (1 . w./J. Then the set of sections s for which thertp exists a function f E V such that f o s = 0 has bounded height. Then infinitely many points lie in the same coset of J(F)/&(F) where J is the Jacobian of X. Then we can pick the curves C. E J(F) such that all the points P in that coset can be written in the form P=mQ+P.5.(p). wz respectively.. over F. = this process to obtain a By Proposition 4. They by the definition of PF... As we have seen we get an associated function . Let . and it follows that H%’ B.I.160 Manin’s subsequent THE CEOMETRlC CASE OF MORDELL’S CONJECTURE key lemma used both for the above step.Y + WW’Q:. we are reduced to the case when the dimension of the kernel of the Kodaira-Spencer map is < 1. In the present case.. = J. We restrict the covering J + J given by xumx Co to obtain a covering C.. q2 be the pull backs of w. one then uses a tower of coverings similar to those used in the theory of integral points which we shall encounter in Chapter IX. We pull back this relation to the curve C. Then there is a point P. if we put then 0.-. to the curve X. CVL A41 and the Lemma 4. Then the Kodaira-Spencer maps for C....finction field k(X) mer k. Let B. Let q. the Kodaira-~Spencer maps are inject&. Namely given an integer m 2 2. suppose that there are infinitely many rational points in X. + w* = 0.. We then iterate tower of curves C” -) c. such that + V(d)m. and J.y there exist w.

that is r’ = subgroup of J(Fa) consisting of all points x such that nx E r for some positive integer n prime to the characteristic p. when passing to coverings Y’ of Y. Assume that J is ordinary (that is J(F”) has pg points of order p). Let D be a non-trivial derivation of L. Let C be a projective non-singular curue of genus g > 2 defined ouer.. Suppose C is imbedded in its Jacobian J over F. and from this calculation deduces that the set of sections have bounded height..1 ([Vol 901). Following Raynaud’s work dealing with my conjecture on the intersection of a curve with the division group of a finitely generated group in the Jacobian. Recall that we say that C is stably split if there exists a curve Co defined over k such that for some finite extension E of F we have C E z Co. Lemma 5.. Assume that C is not stably split. Let F be a function field over a constant field k. It would be valuable to have a simplification of this part of Manin’s paper. Let r be a finitely generated subgroup of J(F”) and let r’ be its prime-to-p division group. one might reconsider this part with an eye to seeing if. especially an analysis which would give a sharp bound for the height of sections. Let X be a projectiue non-singular curve owr . One would like to see directly that a bound on the heights of these functions gives a bound for the sections. CHARACTERISTIC p AND VOLOCH’S THEOREM After Manin’s proof Samuel gave a proof of the Mordell conjecture in characteristic p when the curve is defined over a function field and when it cannot be defined over the constant field [Sam 661. and involves the following lemma. Voloch [Vol 901 gave a 2. We suppose that k has characteristic p > 0. Let r be a finitely generated subgroup of J(Fa) and let r’ be its prime-to-p division group. $ proof of this more general property. Furthermore.&id F of characteristic p > 0. one gets anywhere close to the bound for heights conjectured by Vojta as mentioned in $2. VI. Let L = F’ be the .. The proof is of independent interest.Manin calculates the function j..the . Theorem 5. Let C be a projective non-singular curve defined over F.2.separable closure of F. which we assume algebraically closed for simplicity. in other words C becomes isomorphic to a constant curve over E. Then C n r’ is finite. under certain circumstances which we shall now make precise..function .

then first without loss of generality one can replace C by a curve which cannot be defined over LO. By a suitable mixed characteristic version of Hilbert irreducibility ([La 83a]. If C cannot be dejined over L’ then C n pJ(L) is finite. [La 721. This is true for elliptic curves. . Then X may be defkinrd over Lp (i. X is isomorphic to a curve lifted to L . Chapter 9. cf. Indeed. namely: Lemma 5. and the curve is always a “Tate curve”. Theorem 5. provided one could prove that there is some prime p (presumably infinitely many) such that the Jacobian of the reduced curve is ordinary. since the . If C is as in Theorem 5. taking a non-stably split curve defined over a function field in characteristic 0.e. Voloch’s proof suggests a possibility for a new proof of Manin’s theorem in characteristic 0.4 used in Manin’s proof.3.3) one obtains a reduction to Voloch’s theorem. Supposr D lifts to X (meaning that D lfts to a derivution of’ the Junction field L(X) mapping every local ring oJ a closed point into itself).j?om LO). This lemma is the analogue of Theorem 4. one may reduce the curve mod p after taking a suitable model.L. $2. See also my Elliptic Functions.1.1 then follows rapidly. Then Voloch proves a first-order analogue of Raynaud’s results in number fields.j-invariant is transcendental over Q. Chapter 15. Corollary 6.

In this chapter. so I review some analytic terminology in $1. We shall be concerned with metrized line sheaves and vector sheaves. $2. but to extract some basic definitions to show how that theory is relevant to diophantine applications today. before globalizing over number fields. 53. Here I shall assume some elementary definitions (of d. which combines the algebraic side of algebraic geometry. the second is Vojta’s use of the higher dimensional theory to prove the existence of sections for certain sheaves on the product of a curve with itself. to carry out his vast extension of the older methods of diophantine approximation on curves of higher genus. and that this intersection number was actually defined on the rational equivalence classes. the complex analytic side. the point is not to give a general summary of Arakelov theory. complex differential geometry. is Parshin’s proposed bound for (0)&r). I do cover the two main existing possibilities: the first. thus providing the beginning for the ultimate transposition of all algebraic geometry to this case. This amounted to the corresponding Riemann surfaces and their differential geometric properties once the number field gets imbedded into the complex numbers. . the Chern form). which will however be given later in the more differential geometric context of Chapter VIII. This is a huge program. analogous to the one of Chapter VI. in a completely open ended unification of mathematics as far as one can see. Arakelov showed how one could define a global intersection number for two arithmetic curves on an arithmetic surface.CHAPTER VII Arakelov Theory In 1974 Arakelov indicated a way to complete a family of curves over the ring of integers of a number field by including the fibers at infinity. conjectural. partial differential equations and Laplace operators with needed estimates on the eigenvalues. d’.

1)-form on X normakzed such that cp = 1.: X . then there exists a C” function c1 on U such that for all P $ supp(D) we have . because the difference of two such functions harmonic on the complement of D. Let cp be a real (1. g.supp(D) + R satisfying the following conditions: for D with respect to cp GR 1. $1. constant.q = O. On any variety I/ over C with a Cartier divisor D one defines a Weil function associated with D to be a function 2. to is D so sx A Green’s function always exists. Condition GR 1 independently defines a quite general notion. A function satisfying these two conditions is uniquely determined up an additive constant. + W). it is also harmonic on D.supp(D) -+ R which is continuous. ADMISSIBLE ARAKELOV THEORY CVK 411 C METRICS OVER Since metrics are given at infinity. we here consider a complete nonsingular curve X over the complex numbers C.EtD(P) = -loglf(P)12 GR 2. We suppose that the genus of X is g 2 1. = (deg D)cp outside the support of D. Finally we require: GR 3. If D is represented by a rational function f on an open set U.164 VII. and such that if D is represented by a rational function f on an open set U.: V . and for this section we identify X with the Riemann surface X(C). then there exists a continuous function c1 . and being sufficiently smooth on by the continuity of the partial derivatives. ddcg. By a Green’s function we mean a function g. s X Let D be a divisor on X.

.CVIL 911 ADMISSIBLE METRICS OVER c 165 on U such that for all P $ supp(D) we have MP) = -b4f(P)l + W). and is C” outside the diagonal.“. of X x X.f') for P # Q.. is the section defined by 1 in the function field. Among all forms cp. Q) = -logI Up. Thus a Green’s function is 2 times a Weil function. Let 3’ be a line sheaf on X with a metric p. Q) = g(Q.. the Green’s function is a Weil function on X x X with respect to the diagonal. Q. We call this the canonical metric (with respect to cp).I. We denote by [D] the line sheaf Ox(D) with the unique admissible metric p such that if 1. Although we shall not go specifically into these applications. or gAQ) = g(P> Q). of first kind with respect to the hermitian product . there is one which gives rise to special structures about which more explicit theorems have been proved.. As such. and the quotient of two such metrics is constant. Such a metric exists.(Q) = g(R Q) It is a fact that g is symmetric. we define that form because of its importance.(A).. We say that this metric is q-admissible if cl(p) equals cp. We shall also write g.} be an orthonormal basis of the space of differentials Let {R. Then A is a divisor on X x X. then Let A be the diagonal the line sheaf O. and can be viewed as a function on X x X minus the diagonal. with There is a unique metric p on O(A) such that 0. in the sense that 0. Let D be a divisor on X.. cp.

natural isomorphism fi:. is geometrically irreducible.. or canonical form. and let Y = spec(o. ARAKELOV INTERSECTIONS Let F be a number field with ring of integers oF. 52. Furthermore. and if we give Q $(P) the tensor product metric of the canonical metrics as described above.(C) with X. whence we get a metric on Q. Then the residue gives an isomorphism cl:. VII.. be the sheaf of differential forms on X.. we mean an integral scheme of dimension 2 together with a projective flat morphism 7c: x + Y. We often identify X..$. let P E X. We let so the generic fiber is be the disjoint union of the curves over C defined by all imbeddings of F into C.mIp -+ c9 and this isomorphism is an isometry if we give C the metric of the ordinary absolute value. denotes the restriction of d to X. Then there is a where A is the diagonal. via this isomorphism. By an arithmetic surface. This metric is admissible (with respect to the canonical Arakelov form) and will be called the canonical metric. a projective curve over the field F... = W)“lA..) as before. The norm on I coming from the metric will be denoted by .166 We define the Arakelov ARAKELOV THEORY CW 921 volume form. to be Let Ri.(P) = Q:. if &. then Eb. has been equipped with a hermitian metric h invariant under complex conjugation. 0 O. whose generic fiber X. be the set of archimedean absolute values on F. By a hermitian vector sheaf on X we mean a vector sheaf d such that. Let S. and let IZ be the set of imbeddings (T: F -+ C.

be the ideal of cOx (= U. t). We define Let c. and let (s. s X.. and only points in the intersection of (s) and (t) make a non-zero contribution to this sum. with admissible metrics.. On the other hand. 0 The collection of norms {h. We define the Arakelov-Picard group (or arithmetic Picard group) Pit. the intersection number at infinity to be + . ~2’ have non-zero sections s. is a finite abelian group. p “) and (A.(C).&%l~l’&.(X) to be the group of metrized line sheaves. Let x be a closed point of X. we also write of a metrized vector sheaf. As we have done above. we often omit the specific mention of the norms p y.[VII..(X) + R. Then a basic theorem is that: The intersection number (9. Then &. with the group operation given by the tensor product.. and we define the finite part of the intersection number by The sum on the right-hand side is finite. t). Let (9. ~$2 at the point x.) generated by s and t after a choice of local trivialization of 2.) on each component 6 will be denoted by h” if we need to refer to &’ in the notation. Instead of hermitian vector sheaf. pA) be two metrized line sheaves on an arithmetic surface X..(X) x Pit. p A from the notation.J(s. . t respectively. 921 We let ARAKELOV INTERSECTIONS 167 Us) = n Us). up to metric isomorphisms.(p “) be the first Chern form of the metric pM on A. Suppose that 9.(C) We define the intersection number to be which is independent of the choice of s and t.(P”?. A) linear form extends uniquely to a symmetric bi- Pit.(C) splits into a finite number of points with pi E X. and that the divisor (s) of s has no common component with the divisor (t). suppose that the divisor (s) on X.

then we let EP be the Zariski closure of P in X.c log/s/. A summary of the results.).(l).(1)&C on P. let us consider a divisor D on X. and the universal quotient Cf. p) = log(2 : soF) .168 ARAKELOV THEORY lwk 621 We can also define the notion of a hermitian vector sheaf on Y = spec(o. one can then try to translate intersection theory on surfaces into the present context.). On the constant vector sheaf O. and thus we get a hermitian metric on the line sheaf 0. Having the intersection number. $5 we define its (Arakelov) degree deg(g. If P is an algebraic point on the generic fiber. such a sheaf corresponds to a finite module over oF without torsion.(q). This was done for the adjunction formula. Without the hermitian structure. $1. going beyond and including Noether’s formula (also due to Faltings) will be found in SoulC [So 891. The expression on the right is independent of the choice of section. which is just Ox(D) with the metrics at infinity defined in $1. Chapter II. and one can then show that if hp(x) denotes the height of the point x in projective space defined in Chapter II.13. To any point x E P”(F) there corresponds a morphism s.‘i we have the standard product metric arising from the ordinary absolute value on C. Theorem 8. and we let [EP] be the cor- . a projective module over oF. s:0e(l) is a metrized line sheaf on spec(o. Riemann-Roth (Faltings). By pull back. then 1 deg(s.* O. On Pl we have the line sheaf O. If (2. i. by taking the quotient of the standard metric. Then we have the line sheaf CD]. so P E X.( 1)).) -+ Pi = P.: spec(0. and the Hodge index theorem (Faltings-Hriljac). invariant under complex conjugation. Ll for any non-zero section s of 2. M4 = [F: Q] More generally. [Ha 771.e. The hermitian structure simply gives the module a hermitian positive definite scalar product at each imbedding CJof F into C. Proofs will be found in [La 881. p) is a hermitian line sheaf on Y then as in Chapter IV.

1.. Fix a divisor D on X. $1 is valid in the present context. is the possibility they give of being translated to the number field case via Arakelov theory. via the scalar product of forms on a Riemann surface. as a line sheaf without metrics. if j: X --t S is an imbedding of X into a smooth scheme S over Y. X. constant field played a role in this discussion. and also as The discussion in Chapter VI. with an admissible metric. The height can then be described Arakelov intersections as in the next proposition. Then the association 1 in to is a height function ciated with DF. is just where K is a canonical divisor on X. where f before is the sheaf of ideals of Co. we have P E X(Q”). $1. and let DF be its restriction the generic jiber.YIXo f&/c!. just as we did in the function field case of Chapter VI. Thus Oxlv is a metrized line sheaf. Indeed. A theorem of Faltings asserts that in the semi- .. Note that the restriction Wx. = and we shall assume throughout that Wxlr has the admissible metric which we defined in 92. 1 hK(P) = [F(P) : Q] By Proposition for 2. then we have the conormal sheaf as before. Furthermore. $2 via canonical class inequalities and Parshin’s construction to obtain height inequalities..&.. Proposition 2.det j*sZ& 0 det %?.. to a complex fiber X0 is simply no @X.1. the restriction of Oxjy to the generic fiber X. (CKI *C&l) + O(1) One great advantage of the method of Chapter VI. defining 0 x/r . in the class of heights mod O(1) on X(Q”) asso- One can define the canonical sheaf 021 ARAKELOV INTERSECTIONS 169 responding metrized line sheaf.

One may also view the constants a2. I suggested that this may happen only if there is complex multiplication.) in this case. Parshin Do there exist positive numbers aO.(% . going beyond the original finiteness proof. In the number field case.170 ARAKELOV THEORY CVK VI has raised the following stable case. and found a non-zero numerical value for ((!. as variable. the family becomes birationally equivalent to a product over a finite extension of the base Y. Bost. implies height inequalities of the same type as H2 of Chapter VI. would depend on E. for every E > 0. and generic fiber of genus g 2 2 the following inequality holds: ((f&) 5 a2 1st Y log #k(y) + a. In addition.. and all semistable families X/Y with Y = spec(o. 2 0. a2 absolute constants.e. if @.I&.. For Parshin’s own discussion of these questions see [Par 891. az) for which the inequality is true gets raised here in even stronger form than in the function field case.) Parshin showed how such an inequality would imply bounds for the height of rational points. such that for all number fields F. For instance.. if (ci)&. In the present context of Arakelov theory.2)CF : Q] d(F) + a. we always have (6&) question. a2. a. a. The problem about the set of (a. i. a. effectively computable.). could be used to prove the Mordell ConjectureeFaltings theorem. Vojta conjectures that aI can be taken as 1 + E. and under some additional hypothesis (for instance if X. that is. As usual. this is presumably a rare occurrence. with al.) = 0. this happens only if the family is stably split. Vojta’s arguments apply both to the function field case and to the number field case.I$. we also mention the extent to which Vojta has gone toward proving his conjecture bounding the . has good reduction everywhere).. by refining the Parshin construction. What is the geometric shape of this set in R3? Szpiro has also raised the question when is ((II&) = 0.[F: Q]. a proof of such a diophantine result using Arakelov theory techniques may ultimately lead to effective upper bounds for the heights of solutions. (The sum is taken over all closed points of Y. and a0 depending on g. Szpiro gave another proof for Raynaud’s theorem that the intersection of the curve with the group of torsion points in the Jacobian is finite [Szp 841. Vojta shows in [Vo SS] how the proposed Parshin inequality. a. Mestre and Moret-Bailly [BMMJ have done some computations in the case of one of the irreducible factors of the Fermat curve. in which case a. and even a weaker form of it. In the function field case. 52.) > 0 (strict positivity!).

This arithmetic discriminant bears to the (logarithmic) discriminant d(P) the same relation as the arithmetic genus of the singular curve in algebraic geometry bears to the genus of a desingularized curve. For each imbedding cr: F -+ C we get a variety x.CVK 031 HIGHER DIMENSIONAL ARAKELOV THEORY 171 discriminant height of algebraic points. is a variety. We often view X over spec(Z).=x x.(P) 5 (1 + E) d. for instance.2. VII. We let again Y = spec(o. because the arithmetic discriminant is too large compared to the discriminant. projective and flat over Y. algebraic point P to be Vojta defines the arithmetic 1 of an d’(P) = [F(P) : Q] (CEPI. and such that the generic fiber X. who define the arithmetic analogue of intersections for cycles of all dimensions.) where F is a number field. Although nothing like such a bound was known up to now on curves. The existence of such sections was used in a spectacular way by Vojta in his new proof of Faltings’ theorem.Y we mean a regular integral scheme X. C&l + Qf. Let 71:X + Y be an arithmetic surface as above. By an arithmetic variety n:X-.Y).c. . $3. it is still too weak to prove even a weak form of the abc conjecture. Fix an integer n and E > 0. it is usually large compared to d(P). But using it. but I want to deal with one aspect of the theory having to do with the existence of sections for a vector sheaf or a line sheaf. and also the usual objects entering into the HirzebruchGrothendieck Riemann-Roth formula. It would take too long here to give all the definitions. HIGHER DIMENSIONAL ARAKELOV THEORY This higher dimensional theory was developed by Gillet-Soul6 [GiS 881. Vojta proves [Vo 90d]: Theorem 2. since Y itself is over spec(Z). Then for all points P E X(Q”) of degree 5 n we have h.(P) + O(1). and it is worth while to see more precisely how the higher dimension enters into that picture. In particular.

).. Via go and h. &0). We can form the derived functor (cohomology functor) W%. . there is an orthogonal direct sum decomposition AO*q(&o) = Im a @ Im a* @ HO. . Therefore we get a scalar product on A’v~(&~. 84. metric ga.. q) with coefficients in &. with corresponding we let p0 = 13:/d!. CqJ.q(X. on X. which is a coherent sheaf on Y. Note the role of torsion in this definition. Let /\o~~(&?~) be the sheaf of differential forms of type (0. Similarly. . See [GrH 781. on go. c rz HO. the elements si . For each Q we suppose given: a hermitian metric h.. Chapter 0. We also have 781 R%*6 0. E Rqn. so of absolute dimension d+ 1. there is a hermitian scalar product on K+r(&‘O).“(&... choose elements sl. and we denote its module of global sections by Rqrr. . Recall that if M is a finitely generated module over Z.. Let & be a vector sheaf on X.d. . a Kahler form o. and then one defines the Euler characteristic of the sheaf as the alternating sum of the Euler characteristics of its cohomology. under the imbedding 0 . Furthermore. This definition can be applied to the cohomology groups over Z of some sheaf.s. Then Rqn.. we define the Euler characteristic X(M) = -log Vol(M. This is a finitely generated module over oF.b which are linearly independent over oF and are maximal such.. .s.&./M) + log #(M.) be the vector space of C” sections.) by the formula (*I <II?(r](x). given a volume on MR = R @ M.q(X. For each cr: F -+ C we get the vector sheaf b. on X0.172 ARAKELOV THEORY cm 031 We let X be of relative dimension d over Y.. .sj) is finite. and as shown in [GrH p. 4)X”(C) = ?‘(X))&(X s The Dolbeault operator has an adjoint a* relative to this product. .b/(Co. for each 0. $6. and let A’.

(X).2 but then we would be missing the “torsion” at The right-hand side is independent of the choice of basis. .. qa Then we define the arithmetic ~~~(8) = i Euler characteristic of Gillet-Soule + c $z(&O). the zeta function has an analytic continuation as a meromorphic function on C..1. We define the L2-degree of Rqn.(X) = @ Q 0 CH%. the critical comments in [Weng 913). 5 I. is a positive real number. . They also define an arithmetic Chern character chAr: metrized vector sheaves + Q 0 CH.q has eigenvalues 0 < ..q(&O). (. therefore introduce. &‘O).d 0 : Co.(s) 1 v. Let H. but we do not reproduce this definition (cf. d to be q=o (. Then det H. = which converges absolutely for Re(s) > d. g. R%.l)qcG..(Yx. with a hermitian metric hxlr corresponding to the chosen Klhler form o (at each imbedding a). and the associated zeta function L..6.q(o). In a purely geometric context. We define the analvtic torsion +%b)= c (. The operator..6 = -f 1 log det H. and hence the zeta function.q = aa* + a*a which we must viewed as an operator on lm a @ Im a* in the orthogonal decomposition of AO. we would define the Euler characteristic to be qio1)’ Rq~. be matrix representing the scalar product (*) with respect to this basis.da.6 by the formula deg. + log(Rqrr.1)4 deg. .cm 931 HIGHER DIMENSIONAL ARAKELOV THEORY 173 form a basis of HoVq(X. hxjy). Let Ao. where Q 0 CH. depend on d and the metrics h.d We have the relative tangent sheaf 5&.. Rqn. Then Aa.(X) . Gillet-Soul& define an arithmetic Todd class Td. By a basic theorem of Seeley [See 671... holomorphic at s = 0.

on the product of two arithmetic surfaces Xl and X. &‘h@ yP@“) for all positive integers arithmetic intersection theory. Then GillettSoule have announced that the usual Hirzebruch formula is valid. Weng [Weng 913 has pointed out that the formula is not correct. namely: Statement 3. .1 (GillettSouli: [GiS 891). However. and he has given a modification of the definition of the arithmetic Todd genus for which the formula is conjecturally valid. b]. Gillet-Soult’s theorem is proved under the assumption that the metric on 2 is positive.&%h) = de ChdGJ~ TdAr(Ti. the first term on the right-hand side dominates.Yr hXjy). over Y. In the applications described below. one needs the with an error term formula only for IO Z@“. c’?~ Sp@‘“) 2 ndfl &Wt+‘) 0 for n -+ co.174 ARAKELOV THEORY cw 931 is the direct sum of the arithmetic Chow groups. 1f (yt+i) > 0 and cl@. and only asymptotically O(nd log n) for n + co. powers of a line sheaf.. However. some of which applied to non-positive metrics. Let us define hO(X. h) be a hermitian vector sheaf on X. Then in particular. We let r be the rank of b. In this case this formula is proved. Theorem 3. Consequently. Vojta works in the case of an arithmetic variety 7~:X + Y such that X has a birational morphism X -+ Xl xy X. Let rc: X -+ Y be an arithmetic variety of relative dimension d.. and Vojta used metrics which are not necessarily positive.2 (Gillet-SoulC for all C. Then ~. and gives the existence of sections for n large. We take the trivial vector sheaf 6 and some line sheaf 2 with metrics. the maximal power is defined as a real number. We go back to Statement . b) such that IsIL2 5 l}. tensored with Q. notably the analysis of Bismut-Vasserot [BiV 88a.) > 0 Theorem 3. [GiS 88d]). E. and Vojta had to go back to the ingredients which went into its proof. Let (3.2 as stated was of no use. then hO(X. By Gillet-Soult’s we also have hO(X. say.Wd log 4 In particular. The simplest context is that of a line sheaf 8 with hermitian metric pa for each e. (J$+‘) n. p) be a metrized line sheaf.) = log #{sections s E H’(X. Let (8. Specifically. One of the main applications of such an expression is to provide sections for.

if (T3) > 0. which we state in the special case used in [Vo 90a].Wd log 4 for njc. and I wanted to show one example of how estimates at infinity are needed in the Arakelov geometry to obtain number theoretic results. and can thus be absorbed into the error term.. For i = 1. R’z. The estimates showing how the analytic torsion can be absorbed into the error term are particularly striking.1. 2 let pi be a volume form on Xi. Let the Kiihler form on M be p:pI + pI. but without the whence the first term on the right dominates assumption that the metric is positive. Lemma 3. §31 HIGHER DIMENSIONAL ARAKELOV THEORY 175 3. Let pi: M + Xi be the projection.Y to get = . Then G. Let &. be a metrized vector sheaf on M... let ZP be a metrized line sheaf. The two lemmas are purely analytic. 2 are O(n2 log n). The bound in Lemma 3.0.IWI. Let 9 be a metrized line sheaf on M.(Y3) + O(n2 log n) Bn + c MC@“) d and we want to use this formula to show the existence of sections for large n. Lemma 3. see Chapter IX. Then C.(O)2 .n be the zeta function of the Laplacian on A0*q(6 @ Y@*). x X..ul. normalized to have volume 1. The problem is then with the analytic torsion.* ?1*2@” = g(Z3) + O(n2 log n).l)i deg. $6. and one then sees how deg. and let [q. which is proved asymptotically i$ (. This requires certain upper and lower bounds on &(O). .4 amounts to finding a lower bound for the first eigenvalue of the Laplacian. x {P2) the restriction of 9 to this jiber is a positive metrized fine sheaf whose metric is admissible with respect to PI. The arguments of ordinary algebraic geometry show that the terms with R’ for i = 1. Let M = X. be a product of two compact connected Riemann surfaces..3 Let M be a compact Kiihler mangold of dimension d. Assume that for each jiber X.(O)= O(n2loi34 for n-+03.4. and one has to show that these terms also can get absorbed into the error term.

but also there is an equally deep seated analogy . we shall get quantitative diophantine criteria by inequalities at one absolute value. we see a pattern emerging. if one considers holomorphic families of varieties. by geometric conditions (the Mordellicity of the complement of the special set). in first approximation and qualitatively. or complex analytic invariants. in Chapter IX. For instance. I conjectured that a projective variety X defined over a subfield of C finitely generated over the rationals is Mordellic if and only if every holomorphic map of C into X(C) is constant. which is a local condition. and also by conditions at one archimedean absolute value. that certain global diophantine properties of a variety are controlled conjecturally. the problem of determining whether there exist only finitely many sections can be studied from a complex geometric point of view. First. It is known in many cases that certain projective varieties have this holomorphic property. Thus one obtains complex analytic criteria for a variety to be Mordellic. to deal with quantitative estimates. In addition. By now. But it also turns out (conjecturally at the moment) that the property of being Mordellic for a projective variety can be characterized in terms of purely complex differential geometric invariants. it is not known that they are Mordellic. but except for curves of genus 2 2 or subvarieties of abelian varieties which do not contain translations of abelian subvarieties of dimension > 0 (Faltings’ theorems) or varieties derived from those by products or unramified coverings or quotients. Vojta has taught us that not only was there a classical analogy between algebraic numbers and algebraic functions.CHAPTER VI I I Diophantine Problems Complex Geometry and Complex differential geometry intervenes in diophantine problems through several factors. Similarly.

but with complex spaces. via Nevanlinna theory. For any positive number Y we let D(r) be the open disc of radius r. A fairly complete exposition with proofs is given in [La 871. b] +X such that r(a) = x and y(b) = y. and is intended only as a brief guide. We let D be the unit disc in C. The inf could also be taken over piecewise C’ curves.D. In . DEFINITIONS OF HYPERBOLICITY We shall work not only with complex manifolds. Just as an algebraic space is defined locally by a finite number of polynomial equations in affine space.: D + D(r) gives an metric holomorphic isomorphism between D and D(r). we touch on all these themes. Note that for z = 0. Hence the account in this chapter will be rapid. or as we shall also say.z 1 . a complex space is defined locally by a finite number of holomorphic equations.D can be identified with C itself. centered at the origin. a complex hermitian manifold. I4 hyp. In this chapter.r. the hyperbolic metric is the same as the euclidean metric. The hyperbolic metric on D(r) is defined by I@L. $1.D can be identified with a complex number. y) = inf b I&(t)1 dt Y sa where the inf is taken over all C’ curves y: [a. defined on a tangent vector o by the formula where IL is the euclidean norm on C. We have the hyperbolic metric on T.lz/r12’ = Thus multiplication by r m. y with respect to this metric by d(x. We can define the distance between two points x. VIII. If z E D. the tangent plane T. Let X be a complex manifold with a hermitian metric.CVIIL §11 DEFINITIONS OF HYPERBOLICITY 177 between algebraic numbers and holomorphic functions. and a tangent vector u E 7.









particular, we let d,,, denote the hyperbolic distance on the disc D or D,, to distinguish it from the euclidean distance d,,, . Next, let X be a connected complex splace. Let x, y E X. We consider sequences of holomorphic maps
J:D-,X, i=
1 3 . . . . m,

and points pi, qi E D such that f,(p,)

= x, iLhJ


Y, and

= .L+l (Pi+1 1.

In other words, we join x to y by what we call a Kobayasbi chain of discs. We add the hyperbolic distances between pi and qi, and take the inf over all such choices of fi, pi, qi to define the Kobayasbi semidistance
dx(-T Y) = inf igl dhyp(Pi2 Yi).

Then d, satisfies the properties of a distance, except that d,(x, y) may be 0 if x # y, so we call d, a semidistance. If X = D then d,,, = dD, in other words, the Kobayashi semidistance is the hyperbolic distance. If X = C with the euclidean metric, then d,(x, y) = 0 for all x, y E C. Let f: X --+ Y be a holomorphic map of complex spaces. Then f is distance decreasing for the Kobayashi semidistance, that is

d&-(x), .0x’)) 5 4(x, x’)


x, x’ E X.

The Kobayashi semidistance is continuous for the topology of X. We define X to be Kobayasbi hyperbolic if the semidistance d, is a distance, that is, x # y in X implies d,(x, y) > 0. Hyperbolic will always mean Kobayashi hyperbolic. All other types of hyperbolic properties which we encounter will be subjected to a prefix to distinguish them. Directly from the definition, we note that to be hyperbolic is a biholomorphic invariant. Furthermore, if X, Y are hyperbolic, so is X x Y. A complex subspace of a hyperbolic space is hyperbolic. Discs and polydiscs in C” are hyperbolic. A bounded domain in C” is hyperbolic, since it is an open subset of a product of polydiscs. A quotient of a bounded domain by a discrete group of automorphisms without fixed points is also hyperbolic. More generally, let X’ -+ X be an unramified covering. Then X is hyperbolic if and only if X’ is hyperbolic. There is another notion of hyperbolicity which will be relevant. We define a complex space X to be Brody hyperbolic if every holomorphic map f: C +X is constant. It is trivial that:









implies Brody hyperbolic.

The converse holds under various compactness conditions,

for instance:
Then X is

Theorem 1.1 (Brody). Let X be a compact complex space. Kobayashi hyperbolic if and only if X is Brody hyperbolic.

For proofs of this and related properties, see [La 871. We now return to the considerations of Chapter hyperbolicity. I conjectured [La 741 and [La 861:

I, $3 in light


Conjecture 1.2. The following conditions are equivalent for a projective variety X, deJned over a sub$eld of the complex numbers finitely generated over the rationals. X(C) is hyperbolic; X is Mordellic; Every subvariety of X is pseudo canonical.

Either one of the second or third condition would show that the property of being hyperbolic is algebraic. See also Chapter I, §3, Conjecture 3.6. In particular, we have the subsidiary conjecture: Let X be a projective variety de$ned over a $eld F finitely generated over Q. If a: F + C is one imbedding of F into the complex numbers, if X, denotes the resulting complex variety, and X,,(C) is hyperbolic, then for every imbedding a: F + C the complex space X,(C) is hyperbolic. Just by itself, this constitutes an unsolved problem today, independently of any connections with diophantine properties, or the algebraic geometric condition of being pseudo canonical. In addition, we recall the algebraic special set defined in Chapter I, $3, which we now write as Sp&X) because we introduce the holomorphic special set SphO,(X) to be the Zariski closure of the union of all images of non-constant holomorphic maps f: C -+ X. But I conjectured [La 861:
Conjecture 1.3. The algebraic and holomorphic special sets are equal.

Thus the conjecture that X is pseudo canonical if and only if the special set is a proper subset now applies also to the holomorphic special set. A question also arises as to the extent it is necessary to take the Zariski closure in the above definition. The answer is known for abelian varieties, see Theorem 1.10. Previously, there was a weaker conjecture of Green-Griffiths, implicit in [GrG SO].







L-VIII, 411

Let X be a pseudo-canonical projective variety over C. Let f: C --f X be holomorphic. Then the image of f is contained in a proper Zariski closed subset.

For a result in this direction, see Lu-Yau [LY 901. Note that the property of a variety expressed in the Green-Griffiths conjecture, i.e. that every holomorphic map of C into the variety is not Zariski dense, is not equivalent to X being pseudo canonical. The example X = C x P’ where C is a curve of genus 2 2 shows that there may be a surface covered by holomorphic images of C, without the surface being pseudo canonical. The special set in this case is the whole surface. Nevertheless, the image of every non-constant holomorphic map C + X is contained in one of the fibers of the projection C x P’ -+ C. One can also pseudofy the notion of hyperbolicity. We say that X is pseudo-Kobayasbi hyperbolic if there exists a proper algebraic subset Y such that if x, x’ E X and d,(x, x’) = 0 then x = x’ or x, x’ E Y. We say that X is pseudo-Brody hyperbolic if the holomorphic special set is a proper subset. I conjectured:
Conjecture 1.4. The following conditions are equivalent for a projective variety X defined over a subfield of the complex numbers finitely generated over Q: X X X X is is is is pseudo-Kobayashi hyperbolic; pseudo-Brody hyperbolic; pseudo canonical; pseudo Mordellic. the set Y mentioned above can be taken to be the special

Furthermore set.

Even the equivalence of the first two conditions is not known today. In parallel with the conjecture that the complement of the special set is Mordellic, I also conjecture that the complement of the special set is hyperbolic.
Example 1.5 (Hyperbolic hypersurfaces and complete intersections).

Brody proved a conjecture of Kobayashi that the property of being hyperbolic is open, say in the following sense. Let

be a proper holomorphic map of complex spaces. If f -‘(y,) is hyperbolic for some point y, E Y, then f-'(y) is hyperbolic for all y in some open neighborhood of y,. However, the property is not closed. An

cvm 011




181 of hyper-

example was first given by Brody-Green, namely the family surfaces x; + ... + x”3 + (tx,x,)d’2 + (tX,X2)d’2 = 0.

They proved that these varieties are hyperbolic for d even 2 50, and all but a finite number of t # 0. But for t = 0 the variety is a Fermat hypersurface which contains lines, and so is not hyperbolic. Kobayashi has raised the question whether the generic hypersurface of degree d in P”, with d 2 II + 2 is hyperbolic [Kob 701, and similarly for the generic complete intersection of hypersurfaces of degrees d,, . . . ,d, if

4 +

... + d, 2 n + 2.

Since a non-singular hypersurface is simply connected for n 2 3, one sees that hyperbolic spaces include a lot more than those which have bounded domains as universal covering spaces. Suppose that X is a projective non-singular variety over C. Then we have the canonical class K, and also the cotangent bundle T,” . The class K, is the divisor class associated with the maximal exterior power max /i TX” . Kobayashi proved [Kob 753:
Theorem 1.6. If T,’ is ample, then X is hyperbolic.

Kobayashi-Ochiai [KoO 751 conjectured that if X is hyperbolic then the canonical class K, is pseudo-ample, but I would make the stronger conjecture:
Conjecture 1.7.

If X is hyperbolic then K, is ample.

The converse of this last statement is not always true. A Fermat hypersurface of high degree has ample canonical class, but contains complex lines, so is not hyperbolic. In any case we have (with a conjecture in the middle):
TX” ample Subvarieties


X hyperbolic

=> K, ample

=> K, pseudo ample.

of abelian varieties

My conjecture that a subvariety of an abelian variety is Mordellic unless it contains the translation of an abelian subvariety of dimension > 0 led me to conjecture its hyperbolic analogue, which was proved by Mark Green [Gr, 781, namely:







CVIK 011

Theorem 1.8. Let X be a closed complex subspace of a complex torus. Then X is hyperbolic if and only if X does not contain a translated complex subtorus # 0.

In addition, the study of complex lines in an abelian variety from the point of view of transcendental numbers led me to conjecture the following statement, proved in [Ax 721:
Theorem 1.9. Let A over C, and let X parameter subgroup, a translation of g(C) be be i.e. or an abelian variety imbedded in projective space a hyperplane section. Let g: C + A be a one a holomorphic homomorphism. Then X contains the intersection of X and g(C) is not empty.

For subvarieties of abelian varieties over the complex numbers, the UenooKawamata fibrations of Chapter I, §6 have the stronger property to take into account holomorphic maps of C into X.

f: C + X be a non-constant

1.10. Let X be a subvariety of an abelian variety over C. Let holomorphic map. Then the image of f is contained in the translate of an abelian subvariety, contained in X.

Several people contributed to this theorem. Bloch in 1926 was the first to make the conjecture that if X is not the translation of an Abelian subvariety then a holomorphic map of C into X is degenerate, in the sense of being contained in a proper algebraic subset. Ochiai [Och 771 made major progress toward this conjecture. Then simultaneously Green-Griffiths and Kawamata proved Bloch’s conjecture. Specifically, Green-Griffiths proved [GrG 801, $3, Theorem I’:
Let X be a closed complex subspace of a complex torus A. If X is not the translate of a subtorus of A, then the image of a non-constant holomorphic map f: C --, X lies in the translate of a proper complex subtorus in X.

On the other hand, Kawamata [Ka 801 proved not only Theorem 1.10, but also further results which combined with Ochiai’s criterion yielded the full fibration theorem recalled in Chapter I, $6. In particular:
Theorem 1.11. For a subvariety of an abelian variety, the algebraic and holomorphic special sets are equal, and one does not have to take the Zariski closure to dejine them.

However, note that there may exist countably many translates of abelian subvarieties which are not contained in the fibers of the Ueno-Kamawata

cvm §ll





fibration. They may occur as sections, but their images are contained in the union of those fibers. For the analogous structure theorem concerning semiabelian varieties, see Noguchi [No 81a].
Non-compact spaces

The above examples concern compact complex manifolds, There are also results concerning non-compact manifolds or spaces, but some new subtleties arise. Green has given an example of a Zariski open subset of a projective variety which is Brody hyperbolic but not Kobayashi hyperbolic. However, the possibility remains open that for an affine complex variety, the two are equivalent. For a discussion of this and connections with the diophantine properties of integral points, see [La 871. The most natural way of obtaining non-compact spaces is to take away some proper algebraic subset in a projective variety over C. In that line we have Borel’s theorem. Recall that hyperplanes in P” are said to be in general position if any n + 1 of them or fewer are linearly independent. Also if we pick n + 2 hyperplanes such that they are given by the equations xi = 0 for j = 0, . . . ,n, and x,, + ... + x, = 0,

where xj is the j-th projective coordinate, then for each subset I of (0, . . . , n} which consists of at least two elements and not more than n elements, we let the corresponding diagonal hyperplane be
D, = solutions

of the equation

1 xi = 0.

Then we have one form of
Theorem 1.12 (Borel’s theorem). Let Y be the complement of n + 2 hyperplanes of P”(C) in general position. Then every holomorphic map C + Y is either constant, or its image is contained in the diagonals.

Borel’s theorem was complemented

by Green and Fujimoto

as follows:

Theorem 1.13. Let Y be the complement of q hyperplanes of P”(C) in general position. Assume q 2 2n + 1. Then Y is Brody hyperbolic, that is every holomorphic map C -+ Y is constant.

Not much is known about the complement of hypersurfaces rather than the complement of hyperplanes. For a fuller discussion, examples, and

varying C” over x. 4). and let L be a holomorphic line bundle over X. then the metric p is given by where h. . represented by a function s”: U -+ C in the trivialization. = dzi. CHERN FORM AND CURVATURE Let X be a complex manifold. A hermitian metric is given by a positive definite hermitian product on each fiber. is a C” function. A dzip and similarly for fiJ.. The Chern form is of type (1.z. we have c. Green..(p)jU = -dd’loglsl. denoted by cl(p).. 4 are independent of the choice of holomorphic coordinates. We let d be the usual exterior derivative. A . maps real functions to real functions. d = 8 + a. . as where dz. [La 871 in addition to [Kob 701.184 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY cvm 421 bibliography of results of Fujimoto. then we say that the form is of type (p. Noguchi. = dd’log h. We say that a (1. . have We dd’= J-’ aa. and s is a section of L over U. 93.. We shall return to these questions from the affine diophantine point of view of integral points in Chapter IX. $2.: U -+ R.e. i. The integers p. If U is an (ordinary) open set over which L admits a trivialization. VIII. especially IX. 271 The Chern form of a metric p is the unique form. We let Then d’ is a real operator. cf. such that on an open set U as above. 1). 1)-form o = C hij(z)~dzi A ~ . . If a form is expressed in terms of complex coordinates zi.

If X is hyperbolic. We call G the Griffitbs function. Ric(f*o) = Gf *co with G 2 B. A (1. Let K. that is curvature. By the n-th power Ric(Y)” we mean the exterior n-th power. volume form. then X is hyperbolic. The converse is a major question of Kobayashi [Ko 701: Let X be a compact projective complex variety. 1) form o will be called strongly hyperbolic if it is positive and if there exists a constant B > 0 such that. for all holomorphic imbeddings f: D +X the Griffiths function of f*o is 2 B. where n = dim X. If h > 0 everywhere then we say that Y is a A volume form Y as above determines its Ricci form Ric(Y) = dd” log h(z) in terms of the coordinates z. We call Q(Z) is hermitian = d -J. Since Y is a volume form. In dimension 1. cl(~) = Ric(Y). 1)-form on X? does . there exist a strongly hyperbolic (1. It is easy to show (by what is called the Ahlfors-Schwarz lemma) that if a strongly hyperbolic form exists. and by definition. A metric p is called positive. Let Y be a form of type (n. there is a unique function G on X such that iRic(Y)” = GY. A volume form Y as above determines a metric K on K. A 2-form commutes with all forms. We can write Y locally in terms of complex coordinates Y(z) = h(z)@(z) with a C” function h. n). (via the local functions h).lr’ dzi A dyi the euclidean form. = Amax 7” be the canonical bundle.CVIK 021 CHERN FORM AND CURVATURE 185 positive if cl(p) is a compact complex if it has a holomorpositive is positive and we write o > 0 if the matrix h = (hij) definite for all values of z. G is minus the Gauss by definition. Kodaira’s imbedding theorem states that manifold admits a projective imbedding if and only phic line bundle with a positive metric. We write G = G(Y) to denote the dependence on Y.

using equidimensional holomorphic mappings f: C” + X (of dimension n) or j-:D”+X.. Let (X. namely measure hyperbolicity. 1)-form on them. 1)-form. and length functions which generalize the notion of hermitian metric. The existence of a hyperbolic (1. Proposition 2. To keep as sharp a focus here as possible. I was careful not to take as definition of hyperbolicity this property involving (1. We call the pair (X. 1)-forms may exist as substitutes. The above properties are the ones which are most important for us. then the matrix (hij) defines a hermitian metric. 1)-form gives a measure of hyperbolicity for the Kobayashi distance. w) be a hermitian manifold. = an.. It is also possible to define analogous objects for (n. We let d. however.186 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY CVIII. o) a hermitian manifold. 1)-forms. Furthermore and .&d. . However. The first inequality is the infinitesimal version of the second. n)-forms. [La 871 and the survey article [La 861. Assume that there exists a constant B > 0 such that for every complex submanifold Y (not necessarily closed) of dimension 1 we have G(ol Y) 2 B.1212)2 27r The factor of 2 in the numerator is placed there so that Ric(o.) so G(mD) = 1 (negative curvature -I). Bf *co 5 a. when I conjectured the equivalence of hyperbolicity and Mordellicity. Thus we pee that the existence of the hyperbolic (1. Then X is hyperbolic. For example on the unit disc D we have the hyperbolic form 2 J-ldzr\dz wD = (1 . be the hermitian distance obtained from this metric. which will be found in [Kob 701. meaning the manifold endowed with this metric. VI Given the problematic status of this question. gives an easy way of estimating from below the Kobayashi distance between points as follows. and to define a weaker notion than hyperbolicity. weaker objects than (1. If 0 = C hij~dzi A dzj. in the direction of jet metrics as in Green-Grifhths [GG 803.1.sdd.. we have omitted other properties. It is not known if the Brody-Green hypersurfaces have a hyperbolic (1.

Suppose f: X --+ Y as above is stably X. is a non-singular curve of genus 2 the family X over Y is stably non-split if for every finite base Y’ -+ Y. The the theorem then reads: Theorem 3. Parshin gave another proof for the function field case of Mordell’s conjecture. &J ‘%(Y. of the discus- we obtain an exact sequence of fundamental Yo) - 1. The points y. X0 = X .=Y-Uaandy. . groups a.. has bad reduction U = union of a finite number of discs centered at the points of S in some chart.. We need a definition. Y.(F) is finite. --m(*) 1 %(Z.f-‘(U) Z = fiber f-‘(y.cvm 631 PARSHIN’S HYPERBOLIC METHOD 187 This notion is less important for us than hyperbolicity.). y E Y(C) such that X. Let use a mixture of hyperbolicity and topology as fol- S = finite set of points at Y.. obtained by base change function field F’ = C(Y’) cannot be defined over C.1.? and x. let F = C(Y). x0) n. such that 2. actually in a slightly weaker form. From the smooth fibration j-:x. and let f:X+Y be a projective morphism from a non-singular surface X the generic fiber X. and we refer to the expositions in [Kob 703 and [La 871 for a more complete treatment.~Y~. to Y. and x0 as above are fixed for the remainder sion.).. $3. Then Parshin’s arguments lows.(X. the curve X. -+ Y. if-i(y. We say that extension of the from F to the weaker form of non-split. PARSHIN’S HYPERBOLIC METHOD In [Pa 861. VIII. Let Y be a complete non-singular curve over C.

There is only a jinite to the same class [a. the conjugacy class [a. by intersection theory. $31 The goal is to prove that there is only a finite number of sections of J If a section s goes through x0. and there is enough relative compactness involved in the definition of Y. In the first place. s’ are sections such that [s] = [s’] mod NS. The set of conjugacy classes of splittings of (*) coming from sections off is jinite.. The proof of the first proposition consists of general considerations of intersection theory and homology.]. Then in the first place.3. Then the path y induces an isomorphism of the new exact sequence with the previous one. by . Then s defines a splitting of a sequence like (*). and connect x1 with x0 by some path y on Z. f-‘(U)). Proposition number of sections s giving rise 3. the image of [s] in Hz(X) determines the section s up to a finite number of possibilities because a theorem of Arakelov implies that ([s]‘) < 0 under the hypothesis that X is stably non-split.3.].188 DIOPHANTTNE PROBLEMS AND COMPLEX GEOMETRY [VIII. We write [s] for [s(Y)]. X. of the sequence (*). The finiteness of rational points follows from two statements: Proposition 3. Hyperbolicity comes in the proof of Proposition 3.(X. by general criteria of hyperbolicity.] defines the image of [s] in H.) and Z.(X).2. is hyperbolic and the fibers are hyperbolic. is hyperbolic. We give the basic ideas. Hence [s] does not lie in an algebraic family of dimension > 0. We consider the exact and commutative diagram: N%(X) - NS(X) where NS. the space X. let x1 be the point of intersection of s(Y. and therefore s induces a conjugacy class of splittings [cr. Furthermore. then s defines a splitting ~1. if s. whence in the homology groups according to the arrows in the above diagram. we let [C] denote its image in NS(X). then [s] = [s’] except possibly for a finite number of sections. but with new base points x1 and y.(X) is the subgroup of the N&on-Severi group generated by the classes of components of the bad fibers over points of S. Finally.2 follows. = f(xi). essentially because Y. If not. and Proposition 3. If C is a curve in X.









taking out U and f-‘(U). By the distance decreasing property of holomorphic maps, one also sees that for any section s, s(Y,) is totally geodesic, meaning that for any points x, x’ E s( Y,) we have


x’) = d&(X, x’).

We then choose loops yi, . . . ,yn representing generators of xi (Y, , yO). Then representatives for a class TV,corresponding to a section s are given in the form j = 1, . . ..n. YYjY -l, where we write fi to denote the loop yj considered on s(Y,), and y is a path in the fiber connecting x0 with some point x1 E s( Y,) n f-‘(y,). Since s(Y,) is totally geodesic, it follows that the lengths of the loops 8 are bounded in the Kobayashi metric of X,, and the same is true for the set of loops ~7~y-i by the compactness of the fiber. Also by a compactness argument, the set of elements of rcl(XO, x0) represented by loops of bounded length is finite, whence Proposition 3.3 follows. By similar arguments, mixing the properties of the fundamental group and hyperbolicity, Parshin gave another proof for Raynaud’s theorem concerning subvarieties of abelian varieties in the function field case, which we mentioned as Theorem 6.7 of Chapter I. Also by similar arguments, Parshin proved the function field case of my conjecture concerning integral points on affine subsets of abelian varieties under the restriction that the hyperplane at infinity does not contain the translation of an abelian subvariety of dimension 2 1. We shall mention this again in the context of integral points in Chapter IX.






Consider again as in $3 our standard situation

of a proper morphism

where Y is a complete non-singular curve of genus 2 2 over C, and X is a non-singular surface, so we get a fibering. Let y, E Y be a point where X, has good reduction, F = C(Y), and let U be an open neighborhood of y, which is a disc in a chart. Then U is hyperbolic, and if we take U small enough, then X, has good reduction for all y E U, so f-'(U) is hyperbolic. Let s: Y + X be a section. Then the restriction sU of s to U is a holomorphic map s”: u-f-'(u),







cvm §41

which is Kobayashi-distance decreasing. If we let Y0 be the open set of points where X, has good reduction, then

s: Y, -+ f-'(Y,)

= x,

is also distance decreasing, and the set of sections restricted to Ye is equicontinuous. If Y = Ye then the set of sections is compact by Ascoli’s theorem. Since the degrees of sections in some projective imbedding depends continuously on the sections, it follows that these degrees are bounded, whence the heights of sections are bounded, and we have another proof of the Mordell conjecture in the function field case. Since in fact there usually are points of Y above which the fiber is degenerate, we must look more closely at that possibility, and how the Kobayashi distance degenerates in the neighborhood of such a fiber. So let X be a complex manifold (for simplicity) and let X, be a relatively compact open subset (for the ordinary topology). Following Kobayashi, we shall say that X0 is hyperbolically imbedded in X if there exists a positive (1, 1) form w on X (or equivalently a hermitian metric) and a constant C > 0 such that
dxO >= Cd,.

There are other definitions (Cf. [La 87)) which are equivalent to this one, but we have picked the most convenient one. In particular, if X, is hyperbolically imbedded in X, then X, is hyperbolic, both Brody and Kobayashi. Apply the above notion to our standard situation of a fibering f: X + Y, in the neighborhood of a point y, E Y where X, has bad reduction. To make the argument using the compactness of the set of sections go through, Noguchi proved the following result [No 851.
Theorem 4.1. Given a complete non-singular curve Y over C and a complete non-singular curve X, over the function field F = C(Y), of genus 2 2, there exists a proper morphism

from a non-singular surface X onto Y whose generic fiber is X,, and such that if Y, is the subset of Y over which f is smooth, then f -'( Yo) is hyperbolically imbedded in X. For such a model X the set of sections of f is compact, that is, every sequence of sections has a subsequence which converges uniformly on compact subsets of Y.

Thus the condition of being hyperbolically imbedded insures that in the neighborhood of a bad point of Y, the set of sections is still locally

CVIII, 941







compact, and the same argument as before works to get the conclusion that the projective degrees (or heights) of sections are bounded. Noguchi also has a higher dimensional version, assuming the hyperbolic imbedding of the open set where f is smooth in its compactification [No 851, [No 871, but Noguchi has shown in general that there does not exist a good compactification in the higher dimensional case, namely he has shown that the following statement is not true in general:
Let Y be a complete non-singular curve over C, let F = C(Y), and let X, be a non-singular projective variety, such that over some point y, where X, has good reduction, the jiber X,,, is hyperbolic; or alternatively, assume X, algebraically hyperbolic. Then there is a projective morphism j-:X-Y from some non-singular variety X onto Y, such that the generic fiber is X,, and such that, if Y, is the open subset of Y over which f is smooth, then f -‘(Y,) is hyperbolically imbedded in X.

Once one has such a family f: X + Y, then the compactness of the space of sections follows. We describe the situation somewhat more generally. Let M be a complex manifold and D an effective divisor on M. We say that D has normal crossings if in the neighborhood of each point there exist complex coordinates z r, . . . ,z, such that in that neighborhood, there exists a positive integer r with 1 5 r 5 m such that the divisor is defined by the equation Zl.. . z, = 0.
Theorem 4.2. Let: X, c X be a relatively compact, hyperbolically subspace; M be a complex man$old of dimension m; D a divisor with normal crossings on M. imbedded complex

Let f,: M - D + X, be a sequence of holomorphic maps, which converge uniformly on compact subsets of M - D to a holomorphic map f:M-D-+X,. Then there exist holomorphic extensions f, and f from M into X, and the sequence of extensions f, converges uniformly to f on every compact subset of M itself.

The existence of the extensions 7” and f is due to KwackkKobayashiKiernan, and the fact that the sequence of extensions f, converges uni-









formly to 7 is due to Noguchi [No 871. For proofs, Cf. also [La 871, Chapter II, Theorems 5.2 and 5.4. Also compare the preceding and following theorem with the Kobayashi-Ochiai Theorem 3.7 of Chapter I. From Noguchi [No 8lb] we have the following higher dimensional version of Theorem 3.1.
Theorem 4.3. Let F be a jiinction jield over an algebraically closed field k of characteristic 0. Let X be a projective non-singular variety deJined over F. Assume that the cotangent bundle of X is ample. Then:

(1) the set X(F) is not Zariski

dense; or there is a variety X0 over k which is isomorphic X,(F) - X,,(k) is jnite.

to X over F, and

Note that the condition for T”(X) to be ample is the condition under which Kobayashi proved that X, is hyperbolic for any imbedding 0 of F into C [Kob 751.




This theory gives a quantitative version for the qualitative property concerning the existence or non-existence of a non-constant holomorphic map of C into a complex non-singular variety. Let X be a projective variety over C. Let D be a Cartier divisor on X. By a Weil function for D we mean a function 1,: X - supp(D) -+ R which is continuous, and is such that if D is represented by Ed on a Zariski open set V, then there exists a continuous function a: V + R such that for all P $ supp(D) we have

MJ-7 = -loglcp(P)l + a(P).
The difference of two Weil functions is the restriction to X - supp(D) of a continuous function on X, and so is bounded. Thus two Weil functions differ by O(1). If L is a line bundle which has a meromorphic section s such that the divisor of s is (s) = D, and p is a metric on L, then we can take

MP) = -l%lS(P)l,.
Let f:C-*X





be a holomorphic map. Suppose f(C) is not contained in D. This is equivalent to the fact that f meets D discretely, i.e. in any disc D(r) there are only a finite number of points a E D(r) such that f(a) E D. Given a Weil function AD, we define the proximity function

mf,dr) = D) Qr,2n = s


For a real number larly, let

c1> 0 we let as usual log+(a) = max(O, log a). SimiA,+ = max(O, A,).

If D is effective, we could also use A,+ instead of AD in the definition of T-,D. The association D H &, mod O(1) is a homomorphism, and hence so is the association Darns,, mod O(1). Given a E C, let D be represented by the pair (U, cp) on an open set U containing f(a). We define vf(al D) = ord,(qo 0 f) (2 0 if D is effective),

N&(r) NAT, = c = D)
We call N/,,

a E D(r) a#0

the counting function, which counts how many times f hits

D in the disc of radius r. We define the height $,, mod O(1) to be

Following Vojta, note that the definition of the height is entirely analogous to the definition in the algebraic case that we met in Chapter II. Indeed, for each 8, r we have something like an absolute value defined on the field of meromorphic functions on D(r) by

lldle,r = ld~e’%
and for a E D(r) we have the absolute value defined by

-~~~lldl,,, = o,,,(g) = b-d, d log f I I


a # 0.

As Vojta observed, the Jensen formula from elementary complex analysis is the analogue of the product formula (written additively). Of course,








here, the places corresponding to 0 vary continuously, so instead of a sum we have to use an integral. Suppose X = P”(C) is projective space, and D is a hyperplane. Let

be a holomorphic map. We can always represent f by entire functions

without common zero, by using the Weierstrass factorization theorem. Then it is easy to see that the height is given by the Cartan-Nevanlinna expression
T,,,(r) =

2x 0


max 1




Thus the height is entirely similar to the height defined previously for algebraic points in P”, in the number theoretic case. It is a fundamental fact that:
The height Tr,, D mod O(1). depends only on the rational equivalence class

This is sometimes called the first main theorem, although it is simple to prove. Then the height can be characterized by the following conditions.
Let f: C -+ X be a holomorphic map into a projective variety X. each Cartier divisor D on X one can associate a function To

Tf,,: R,, + R
well defined mod O(l), depending only on the rational equivalence class
of D, and uniquely determined by the following properties.

H 1. The map D H TJ,, is a homomorphismmod O(1). H 2.


E is very ample and $: X + P” is an imbedding into projective space, such that E = f -‘(hyperplane), then

where TtiO/ denotes the Cartan-Nevanlinna height of a map into projectioe space.

expression for


In addition,

the height satisfies the further properties:

IfW”)l {b} C k.Vojta’s Nevanlinna Theory Dictionary from [Vo 873 Roth’s Theorem f: C -9 C. b) = h(b) + O(1) SecondMain Theorem i$ m(ai. 4 5 2W . b) 5 Wb) + OOogh(b)) m(a. r) The Dictionary in the One-Dimensional Case. 4 = 1 1% & IwI<r First Main Theorem N(a. . infinite b VES Ilbll”.Nl(r) + O(r log T(r)) // Defect 6(a) = lim inf ~ Conjectured refinement of Roth i$ m(ai.k 44 S Roth’s theorem 2 Artin-Whaples Product Formula logjf(re”)(dtl s 0 2x 1 Wlbll. non-constant . ” =0 + N(co. r) + m(a. [k : Q] 0 2n dfl log+If(re”)[+ N(m. [k ” II Property of heights N(a.b) h(b) Defect Relation Jc W 5 2 Jensen’s formula 2n loglc.+s log+ II. b) + m(a.+a2 T(r) 6(a) = limbinfm(a.‘.r) .N(0. r) . = 1 : Q] . VES or4 f log i Characteristicfunction T(r) = or&f.l = a. r) = T(r) + O(1) -(a3b)=&~slogil~&~~u c N(a’ b. v$s 1% No Logarithmic height h(b) = LClog+llbll. r) 2a 0 I Proximity function m(a~r)=~oz’log’~f(rei~)-a~~ Counting function N(a.

What would be the second main theorem of Nevanlinna theory concerns such divisors. defining the ramification divisor of J We can then define Nf. .196 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY WK VI H 3. . .. = T$o/ + o(1) where TtiOf is the Cartan-Nevanlinna height. Let D be a divisor on X with simple normal crossings. For any Cartier divisor D and ample E we have TS. We define the ramification order of f at the point to be e . do not all vanish at 0. Then where H Tf. .... is finctorial in (X. having the following property.z. D). In the neighborhood of a point of C. .. There exists a proper Zariski closed subset Z. where D’ is a divisor on Y.. but is only a conjecture today. Write f(w) = w”(a4. In other words. D)H T. We say that D has simple normal crossingsif we can write D=cDj as a sum of irreducible components Dj which are non-singular and have normal crossings (defined at the end of the preceding section). . then T. let D be a divisor on X.. then T. Let X be a projective non-singular variety over C and let f: C + X be holomorphic. . We suppose that f(0) = (0. .o. 2 -O(l). Finally.g.w + O(1). Let f:C+X .. . .LJ WJ-..1 Let X be a projective non-singular variety over C.O). let w be a complex coordinate. . 1f D is eective and f(C) Q D. .I = H 4. and D = $-‘(D’). H 6. A(W)) so Zi = Wegi(W) such that g 1. be complex coordinates in a neighborhood of the image of this point in X.Ram in a way similar to Ns. Let rl/: X + Pm be a morphism.. and supposeD = $-l(H) is a hyperplane. and let z 1. H 5. The association (f. Thus for each point of C we have assigned an integer 2 0. = Teo/.1... if $: X + Y is a morphism of varieties. Conjecture 5.

morphic map suppose we deal with a non-constant f:C+P" into projective space. = (the ramification counting function). .2. .Ram b-1 5 O(log r + log+ T. 0) = c (1 E D(r) a#0 NJ. Note that on projective space P”. . The zeros of the Wronskian define a discrete set of points on C. .D(r) + h&9 + ~mw(r) 5 o(log r + log+ T/. j = 0. There exists a finite union of hyperplanes Z. Let D = c Dj be a divisor on P” such that the irreducible components are hyperplanes in general position. E(r)) for r + co outside a set of finite Lebesgue measure.-. Then mf. . holo- As to the ramification.(r)) for r -+ 00 outside a set of finite Lebesgue measure. and such that NJ. represented by (fo. . without common zeros..A9 + f.(W) I I RamW log r. which are the ramification points off.. Theorem 5. We have the Wronskian W = W(f.cvm VI NEVANLINNA THEORY 197 Let K be the canonical be a holomorphic map such that f(C) Q Z. and we now turn to this case.f.C -rP" represented by entire functions f(C) Q Z. . where the functions fi are entire without common zeros. .D(r) + T.n.) = det with i.(W) log f + ord. having the following property. . ..). the canonical class is given by K = -(n + l)H where H is a hyperplane. Only a linear case of this conjecture is known today when X has dimension greater than 1. We define: N&r. . .f.. . the inequality holds: mf.f.). class and let E be an ample divisor. . Then we have the main theorem in this case.. For every holomorphic map f = (fo... ord.

4. Given w. assuming f(0) # a. That means we have to pick the Weil functions to have such properties.w’12 ‘Iw7 w”‘2 = (1 + ]wl2)(1 + lw’12)’ and similarly if a or a’ = co. [Nev 251. We shall do so in a variation given later as Theorem 5. which constitute the original Nevanlinna case. In higher dimension. see also [La 87). w’ E C we define )w . or pseudo ample. the situation today is in flux. [Car 331. Let us now deal with the error term. Siu has made an attempt to get the ramification term and the main theorem by using meromorphic connections [Siu 871. sufficed [Vo 89~1. Let us return to the general case. Then define more precisely q-b r) = s 2n 0 hWlre’“). Vojta showed that the existence of the finite union of hyperplanes Z. In particular. We need to normalize the height so that it has certain smoothness properties. Then the error term on the right-hand side is of a lower order of growth than the left-hand side TJ. one could rephrase the estimate in terms of maps of discs into X. 4 2 + hAIf( 4. For the case when D has components of higher degree. [Nev 701. Cartan proved the theorem under the assumption that f(C) is not contained in any hyperplane [Car 291. even if every holomorphic map C +X is constant. and suppose that the canonical class is ample. let $ be a positive (weakly) increasing function of a real variable such that s * 1 mdu e = boW . that is the term on the righthand side of the inequality in the main theorem. cx). but results are very partial. and then define This value is independent of a. In analogy with number theory. and his error term is not very good. Let us deal with maps into P’. or for varieties other than P”. Thus the main theorem implies the conjecture that if the variety X is pseudo canonical. But the Nevanlinna type inequality in the main theorem would give a quantitative estimate. then there exists a proper Zariski closed subset Z of X such that the image of every non-constant holomorphic map of C into X is contained in Z.198 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY cvm VI Nevanlinna proved the theorem when n = 1.K if the map f is not algebraic.

and for r. ajll i#j Then and b. b].(F) be the smallest number 2 1 such that b. s2k-l) 1 -2T/ + 1 +-(aj. 00. and f ‘(0) # 0. r ~rF’(r) function NEVANLINNA THEORY 199 For any positive increasing function F of class C’ such that is positive increasing. (Tf) we haoe -2T. $.(F) be the smallest number 2 1 such that F(r. . f(0) # 0.. and Tf (dating back to Ahlfors) in the relative case. Let Then for r 2 r. The relative case with the precise error term is due to Wong. II/.a4 be distinct points of P’.t log ~~(0).-. b.($). which I suggested.CVIIL VI is finite. 2 b.$ log ~(0) + 1.(r) S 3Wqq2. We let r. case). This formulation results from the work of Ahlfors [Ah 413. c > 0 we define the error W. and Wong [Wo 891. Let s = $ min llai. It is important to note the difference between the appearance of Tf in the error term in the absolute case.=p... A more structural description will be given in the higher dimensional context of Theorem 5. r) = log F(r) + log @W) + loi3 +kWrMW))). [La 90a.R.rF’(r) Theorem 5. c. . b17 $. and we let b. The absolute case with the precise error term is due to Lang. .3 (Absolute 2 e for rzl. Let a. For suitable function II/. .) 2 1. r) .5 below. one sees that the error term on the right is of .(q) b.. = 1242 + q3 log 4 and b = ilog b.+) outside a set of measure 5 2b.R.(r) 5 iW. r) + b where B. 4 + ~f. and for all + &-. Suppose that (Relative case). aj for all j. except for the use of the general Khintchine type function $.. Lang [La 881.

and one can thus get a measure of hyperbolicity. then one gets the same inequality except that the term corresponding to the canonical class is 0. If instead of P’ we take maps of C into a curve of genus 1. [? I would define the type of a meromorphic such that the error term has the form function f to be a function $ 1% wf) + O(1). .(r) + O. and the inequality gives a contradiction to the existence of such a map. But one can restrict attention to a map of a disc into the Riemann surface. theory (see Chapter IX. Ric f*q = dd” log yf. a complex torus of dimension 1. The problem is to determine best possible types for the classical functions. l)-form on Y then we The integral converges if df(0) # 0. Specifically. If one considers a map into a curve of genus 12. and let f: D(R) -+ Y be a non-constant holomorphic define the height for I < R by map. $2) I raised two ques- (a) Is this the best possible error term for “almost all” meromorphic functions. 8.200 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY CVIK PI the form (1 + 4 1% T. then the canonical class is ample. We write f*? = YfQ Recall that where @ = gdz A dz. that is. We need first a differential geometric definition of the height which often gives greater insight into its behavior. in a suitable sense of “almost all”? (b) What is the best possible error term for each one of the classical functions such as M. I. If q is a (1. We give one example of such a result.(l) In analogy with number tions in analysis: for every E > 0. let Y be a complex manifold (not assumed compact!). J. following Ahlfors-Shimizu.

II/. We need to make some definitions. Also no assumption is made on a compactification or normal crossings.(r) = N. 4 . one has to use a formulation involving a map from a disc. .(r. Let Y be a complex mani$old with a positive (1. every holomorphic map of C into Y is constant because Y is hyperbolic. Let X be a projective non-singular on X.[VIII. We shall now state a higher dimensional version because it exhibits still another feature of the error term.(T/. such a version was given by Carlson-Griffiths [CaG 721.~a&) 5 tStTf... not a map from C into Y. [Cher + !. l)-from r~.z. with the improvement on the error term stemming from [La 901. Theorem 5. $51 NEVANLINNA THEORY 201 One can define the order of vanishing at a given point of D(R) for the derivative of f. Furthermore. 0).).(q.. We define the differential forms w(z) = dd’ log(lzl12 and a(z) = d’ logllzlj2 A con-l. Then for r < R we have B?i. and that the map f is defined on a b. . The following theorem stems from Griffiths-King [GrK 733 and Vojta [Vo 871.. = b. and [La 90a]. in the context of differential geometry and Griffiths functions. Let f: D(R) -+ Y be a holomorphic map.J outside a set of measure 5 2b. It gives a quantitative measure of hyperbolicity.. Wong [Wo 891...4.. Note that the theorem is formulated for a manifold which is not necessarily compact. there exists a holomorphic function A on D(R). Let dim X = n and let variety over C. Suppose there is a constant B > 0 such that Bf *q 5 Ric f *q. Assume df(0) # 0. and a positive C” function h such that so we can define the ramification counting function N/. Theorem 5.-. whence a ramification divisor. The improvement on the error term then went through [La 88b]. Let D be a divisor f:C"+X .). On C” we consider the euclidean norm llzll of a point z = (zl.2.7. Let b.) log 1/1(O) for r 2 r. Actually.($). so to get a non-empty estimate in the main inequality. With a Nevanlinna type error term.

.e. cl(~) = Ric R. We can write f*CI = IAI%O. and essentially by definition. is independent of p. a positive (n.. Then A = 0 defines the ramification divisor of f.. be the Weil function given by MP) = -wwl. be a line bundle having a meromorphic section s whose divisor is D. denoted by Z.. in the sense that f is locally a holomorphic isomorphism at some point. and define the Let S. The height TfSK = Tf.202 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY cvm §51 be a holomorphic map which is non-degenerate. m. be the sphere of radius I centered at the origin. If r] = cl(p) then we write Tf.K associated with the canonical class K.. mod O(1). where J-1 @ = n 271 dzi A dzi and A is a holomorphic function on C”. instead of T/Al.(r)l%lS~f(W =SW + s(b”fb cl(p) = -dd” For simplicity we have assumed f(0) $ D and d!(O) is an isomorphism. Then T. Define the height where B(r) is the ball of radius r. Let L.Ricn is one choice of height Tf.. Let Sz be a volume form on X (i. n)-form). while h is C” and > 0. and let 1. Recall the Chern form logls& Let q be a (1.. Then R defines a metric K on the canonical line bundle L. proximity function = I.. Define the counting function We say that D has simple normal crossings if D = 1 Dj is a formal sum of non-singular irreducible divisors.. 1)-form on X. and locally at each point of X there . Let p be a hermitian metric on L.

When n = 1.. but the . The above theorem stems from the work of Ahlfors.~) outside a set of measure 5 2b.3 in higher dimension runs as follows.i positive (1. we suppose that r-F(r) and IH rZnmlF’(r) are positive increasing functions of I. taken with multiplicity 1. Suppose that D has complexity k. I want to emphasize the exponent 1 + k/n. The maximal value of k which can occur will be called the complexity of D.(F~. . ..n. No such good result is known in the number theoretic case. via the choice of sections sj and the metrics pi. .(pj) 5 q ‘yr = the function such that f*CI = ‘/s@. which applies for all k = 0.($). crossings. .(F) is the same as for n = 1. . c. $2 = volume form on X. the property of D having simple normal crossings is equivalent to the property that D consists of distinct points.?“-‘F’(r) 2 e for all Y 2 1. r) = log F(r) + log $(F(r)) + log ~(cr*“-‘F(r)~(F(r))). We let b. Theorem 5. We let: D = 1 Dj be a divisor on X with simple normal pj = metric on L.(r)F s aa =s’ lit B(.cvm 051 NEVANLINNA THEORY 203 exist complex coordinates zr. SZ)which can be given explicitly. For a function u define the height transform F. in higher dimension n. z. q. Then the analogue of Theorem 5. Suppose that f(0) 4 D and 0 $ Ramf. and some constant B = B(D. The case k = 0 is when there is no divisor.Es. The definition of r.(F) be the smallest number 2 1 such that b. q = . . . D is defined by zr . . such that in a neighborhood of this point. 1)-form such that R 5 q”/n! and also c.) o . and with such good error term is due to Lang.5.zk = 0 with some k 5 n.. and we define the error function S(F. Wong and Lang as in the one-dimensional case. Then for all I 2 P-. Finally. . I). Thus the value distribution of the map f would be determined in the error term on the right-hand side by the local behavior of the divisor at its singular points..

. [Cher 911). Ram(l) above 0. It should be emphasized that Theorems 5. Let p: Y+C” be a possibly ramified covering. This is the reason for our having stated the theorem in higher dimension. Theorem 5. For all the objects (a. @.l)! [ Y : C”]) outside a set of measure 5 2b. term N. in a very simple way. Extending the proof of Theorem 5. w.4 and 5..) defined on C”. 1)-form on X such that f: Y + X be a non-degenerate the set of points y E Y such that p(y) = 0. Ric sI(~) + Nf. following GriffithssKing [GrK 721. and assume that Y is normal. Let [Y: C”] be the degree of the covering.... put a subscript Y to denote their pull back to Y by p. as the case may be. Ram(r) N.. The height can be normalized right away as in number theory. Stoll [St 811. Let holomorphic map. In any case. Cherry gets a similar error term. As before. we note that the 1.204 DIOPHANTINE PROBLEMS AND COMPLEX GEOMETRY WK §51 analytic theory suggests what may be the ultimate answer in that case.. obtains factors and a dependence on the degree which do not properly exhibit the conjectured structure. my = p*m. CTy p*o. as follows. 5 CY: cnl . William Cherry showed that the degree occurs only as a factor.5 can be formulated and proved for normal coverings Y of C or C”. since the structure of 1 + k/n did not appear in dimension 0. to make everything look better. S(Tf. rs&o) 1% Y.($). dividing by the degree. For instance coy = p*o. f unram$ed Then ?./CY : C”l> $94 . We suppose that q is a closed positive (1.5..6 ([Cher 901. Then the degree [Y: C] or [Y: C“] appears in the error term. inverse image of B(t) under p. Ram occurs with coefficient ./(n . is defined as for maps of C” into X.(Y) for all r 2 rl (q. except that the integral over B(r) is replaced by an integral over Y(t). Assume p. etc. = We denote by Y(0) YJ by f*n = ypD. When there is a divisor D. we define The height T/.

as Vojta observed. but in the context of non-compact varieties. Vojta actually integrated the theories of rational points and integral points by subsuming them under a general formalism transposed from Nevanlinna theory. We shall run systematically through their various aspects. Thus Vojta provided an entirely new approach and proof for . and to have a logarithmic singularity on the divisor. We shall state Vojta’s most general conjectures. hyperbolic conditions. and we shall indicate his proof of Faltings’ theorem along lines which had been used before only in the context of diophantine approximations and integral points. Proofs for the foundational results of $1 can be found in [La 831. Whereas previously we have concentrated on diophantine questions involving rational points. and they play a role analogous to the proximity function in Nevanlinna theory. They are normalized to be continuous outside the support of the divisor. He showed for the first time how one can globalize and sheafify this approach to obtain results on rational points in the case of genus > 1. There are many uses for those functions. and essentially Green’s functions in one form or another at the infinite u. and measure the distance of a point to the divisor in some fashion. They give the natural tool to express results in diophantine approximations. Again we meet curves. those local functions are associated with a divisor. notably affine varieties. Integral Points and Diophantine Approximations The height can be decomposed as a sum of local functions for each absolute value II. These functions are intersection multiplicities at the finite v.CHAPTER IX Weil Functions. so they tend to infinity on the divisor. Whereas a height is associated with a divisor class. References for other proofs will be given as the need arises. subvarieties of abelian varieties. we now come to integral points and conditions under which their heights are bounded and there is only a finite number of such points.

We describe briefly these results. Faltings then succeeded in applying this method to prove my two conjectures: finiteness of integral points on affine open subsets of abelian varieties. but sufficient results following a method originated by Baker are known already to yield important diophantine consequences. give two examples of a general principle whereby diophantine properties of varieties result from their behavior at the completion of the ground field at one absolute value. we could determine an exceptional set. INTEGRAL POINTS Mordell’s conjecture. and finiteness of rational points on a subvariety of an abelian variety which does not contain translations of abelian subvarieties of dimension > 0. In addition. we consider quantitative diophantine properties. which does not pass through the Shafarevich conjecture and accompanying l-adic representations. at least equal to 2. relying on the Thue-Siegel-Schneider-RothSchmidt method. One of these culminated with the Masser-Wustholz theorem. Vojta used the higher dimension because even though one starts with a curve. the other relying on diophantine approximations on toruses. diophantine approximations. analysis and PDE. Nevanlinna theory. and classical diophantine problems about rational and integral points. namely bounds on the height. and we shall see in $7 how certain inequalities obtained at one imbedding of the ground field into C give rise to bounds on the heights of rational points. and the preceding chapter. but even higher to get more precise results. The optimal conjectures are still far from being proved. and I have only extracted those aspects of . this special set does not depend on the imbedding of the ground field into the complex numbers and has an algebraic characterization.3. to be described in Theorem 7. in the present cases taken to be archimedean. Thus we behold the grand unification of algebraic geometry.206 WEIL FUNCTIONS. This chapter. Actually. In the present chapter. In Chapter VIII. we saw that under one imbedding. Conjecturally. The conjecture that the complement of this exceptional set is Mordellic shows how a qualitative diophantine property is determined by the behavior of the variety at one archimedean place. namely the Zariski closure of the union of all non-constant images of C into the variety by holomorphic maps. especially Baker’s method and its extensions. there are two major aspects of diophantine approximation methods: the one as above. By the way. Some of the methods of diophantine approximation arose originally in the theory of transcendental numbers and algebraic independence. I have not gone into this subject. Following Vojta’s extension of diophantine approximation methods. he applies that theory to the product of the curve with itself a certain number of times. Vojta by casting his approach in the context of Arakelov theory also shows how to use the recently developed higher dimensional theory of Gillet-Soul& and Bismut-Vasserot for a specific application. the non-archimedean analogue of this property remains to be worked out. These inequalities concern lower bounds for linear combinations of logarithms (ordinary or abelian) with integer or algebraic coefficients.

supp(D) we have A. (Note: N&on [Ne 651 in his exposition and extension of Weil’s work called Weil functions quasi functions. By a Weil function associated with D we mean a function I. = AD. Let D be a Cartier divisor on X. The continuity of a is v-continuity. . which we assume extended to the algebraic closure F”. defined over F. and I found his terminology misleading. Let ~1:V(F”) + R be a real valued function. and which some readers might find useful: [La 60b]. Let V be a Zariski open subset of X. v(x) = -loglxlV. Let B be a subset of V(Fa).x.“1X(Fa) .(P) = v 0 q(P) + a(P). As usual. [La 711 and [La 741. . .) and a constant y > 0 such that for all x E B we have max lxilV 5 y. If D is represented by a pair (U. WEIL FUNCTIONS AND HEIGHTS Let F be a jield with a proper absolute value v. is a homomorphism mod O(1). [La 65b].supp(D) -+ R having the following property.) The association DH1. Over three decades I wrote several surveys where I went further into the theory of transcendental numbers in connection with diophantine analysis. . IX. We say that a is locally bounded from above if a is bounded from above on every bounded subset of V(F”). But there is nothing “quasi” about these functions. We say that B is affine-bounded if there exists a coordinatized affine open subset U of V with coordinates (x1. We define locally bounded from below and locally bounded in a similar way. not Zariski topology continuity. $1. We let X be a projective variety dejined over F. We define a to be bounded from above in the usual sense.CK 011 WEIL FUNCTIONS AND HEIGHTS 207 diophantine approximations which are directly relevant (as far as one can see today) to diophantine questions. We note that X assumed projective implies that X(F”) is bounded. cp) then there exists a locally bounded continuous function a: U(Fa) -+ R such that for all points P E U . . We say that B is bounded if it is contained in the finite union of affine bounded subsets.

is not locally compact.B).) is necessarily locally bounded. associated with f *D. over some constant field which is not finite.(P) 2 --y for all P E X(F*).. or merely that X(F”) is bounded. the completion of the algebraic closure of F. . Assume that X is projective. and this Weil function then defines a Weil function on X itself. One way to construct a Weil function is as follows. Thus two Weil functions differ by O(1).. . . when F is a function field of one variable.208 WEIL FUNCTIONS. Let D be a divisor on X and let . and such that the supports of E. Ej effective for all j. Assume that X is regular. Assume that f(X’) is not contained in the support of D. Weil functions behave functorially in the following sense. In the function field case. Let R be a discrete valuation o. we also want to consider Weil functions on X(F. be the completion as usual. Let D be an effective Cartier divisor.. Let 9’ be a metrized line sheaf with a meromorphic section s whose divisor is D. Weil functions preserve positivity in the following sense...“(P) = -logIv)I” is a Weil function.m) be Cartier divisors. and that the generic fiber is a complete non-singular variety X. = inf lDj + O(1). . Let j-:x+x be a morphism defined over F. . If 1. Then we may have defined a Weil function on the base change of X to C. However. Thus we made a general definition which applies to all cases. . of is a Weil function on X’. F may be a number field. in which case F. however.. INTEGRAL POINTS cm $11 Let F.. Then there exists a constant y > 0 such that I.Al. Example for P E X(F”) . Furthermore. say. is continuous and bounded on X(Fa). have no point in common. with D. then a continuous function on X(F. Then 1. .supp(D) and valuation 2. and Z. In practice. then F. Then the function b. Let Y = spec(R) and let ring with quotient field F be a flat morphism. i Example 1. . let Dj = D + Ej (j = 1. These could of course be viewed as a compatible family of Weil functions on the set of points of X in finite extensions of F.E. and C. and if we have a Weil function when X is a variety over F. Let D be a Cartier divisor on X. Then I. is locally compact. are Weil functions associated with the same divisor then their difference I.

N&on models for abelian varieties. Example 3. where H is a hermitian form called the Riemann form. where A is a lattice in C”. defined by A. To each divisor D on .cm HI WEIL FUNCTIONS AND HEIGHTS 209 DF be its restriction to the generic fiber.(z)1 + . 1 ~~12. there is associated p-‘(D). This example applies to. let D be represented by a rational function 40 on a Zariski open neighborhood of P in X. and in fact stemmed originally from. Let X be a non-singular complete curve over the complex numbers. Our normalizations are such that for the ordinary absolute value u. is the Zariski closure of P in X..4(C) a normalized theta function FD on C” whose divisor is is uniquely determined up to a constant factor.5 C”. Let g. Then o(rp(P)) is independent of the choice of q. my Introduction to Algebraic and Abelian Functions for the basic properties. For each point P E X(P) and P not lying in DF. Example 4.) Then the function I. this function is normalized in such a way that it satisfies an additional property under translation by . z) is a Weil function whose divisor is D.. and which definition. Let A be an abelian variety over the complex numbers. By malized theta function is a meromorphic function on condition F(z + u) = F(z) exp rcH(z. U) + :H(u. (Cf. where E. In fact. H(z.(z) = -log IF. A(C). We call this choice of Weil function the one arising from intersection theory. the function is a Weil function associated with D. so we have an analytic isomorphism p: C/A --. and K(u) is real valued. a norC” satisfying the U) + 2nflK(u) for z. because v(cp(P)) may be viewed as an intersection number of D and E. and the function is a Weil function associated with D. be the Green’s function associated with an effective divisor D.

4 + 44. Then the choice of A. Then we can pick a Weil function I.. INTEGRAL POINTS cm (ill h&) = w .. is the translation of D by a. Let D be a Curtier divisor on X. Let 9 be a line sheaf with a section s whose divisor (s) is D. the function II. For the remaining finite number. and suppose that X is a projective variety defined over a finite extension of the base field F. Then this set of Weil functions can be used in Theorem 1. For each v select a norm on the u-adic extension such that for UEF. is called a N&on function. defined by the intersection theory. Then there exists a choice of Weil functions .. and c(a) is a constant depending on a and D but independent of z. we may choose any Weil function. To do this. for each u. if we put 1 h. We want to take the sum. namely.: F&&‘) [F : F] ” for P E X(F).(P) = ____ 1 CF. is the generic fiber of a morphism X+Y where Y = spec(o) such that X is regular. Assume that the set of absolute values sutisjies the product formula. See the last section of [Ne 651.I. Say D is effective. P # supp(D). satisfying the product formula.. proper and flat over Y as in Example 2.. Itl. to take their sum and obtain the height associated . Fix a divisor on X. Suppose in addition that X. then hl is a height associated with the divisor The choice of Weil functions can be made following Example 1..1. = 1 for all but a finite number of v. t&Z l4” = blvltlv and such that for each t.. Then for all the discrete valuations of o we have a Weil function as in Example 2. class sf D. together with a finite number of other valuations. for each v such that. WEIL FUNCTIONS. since in general it involves a resolution of singularities. Suppose that the set of absolute values satisfying the product formula comes from the discrete valuations of a Dedekind ring. The regularity assumption is a crucial one.1. where D. As such. as follows. it is necessary to make the choice such that certain uniformity conditions are satisfied.210 a E A. One can do this a priori to get: Theorem 1. as in Example 1 will work for Theorem 1. Suppose that instead of one absolute value u we have a proper set of absolute values.1.

. As Vojta pointed out. Vojta then conjectured diophantine inequalities in the number theoretic case. We now return to an arbitrary fixed proper absolute value u. If we replace F by a finite extension F’ and S. and look into the possibility of normalizing Weil functions more precisely than up to bounded functions.. and from known results in Nevanlinna theory. how close a point can be to. and the definition of mD. by its extension S. Define the proximity function mDdP) = [F : F] vss.(P) 1 = CF : F. . variety defined over F”. on F’.S(P) for P E X(Fa). Suppose X is defined over F. If F contains a set of archimedean absolute values S.v(P) for P E X(F) and P $ supp(D). say. Having chosen the Weil functions A.. then cp determines a Weil function Am = -loglcpml”. We want them normalized up to an additive constant.S(p) = hD(P) mD.. we do not expect to achieve more. we introduce the proximity function. but in the present context. we may also write the counting function in the form N. suitably to give a decomposition of the height into a sum of these Weil functions over all u with suitable multiplicities. Let D be a divisor on X. With respect to the finite S we define the analogue of the counting function in Nevanlinna theory to be ND. To measure this closeness. So we let r be the group of constant functions on X(F”).: FJb. So we can legitimately omit F from the notation on the left-hand side. & CFv: F”%dp) for points P lying in X(F). then the right-hand side is unchanged. One basic idea of diophantine approximation theory is to determine. ~ 1 c IF. .s(P) applies to algebraic points P over F. The arbitrary choice at a finite number of u simply contributes to the term O(l). this proximity function is the analogue of the proximity function in Nevanlinna theory. We shall deal with these in 94. and for each finite extension F let S. We work under the standard situation of a field F with a proper set of absolute values Let X be a projective non-singular satisfying the product formula..IIK 911 WEIL FUNCTIONS AND HEIGHTS 211 with the divisor class. we assume that S 3 S.. Observe that if rp is a rational function on X. be the extension of S to F. in some sense. a divisor D.. without further normalizations. Let S be a finite set of absolute values of F.

the Weil functions as in Theorem 1. (1) (2) (3) The association D I-+ AD is a homomorphism mod I. and uniquely determined up to an additive constant by the following properties. = il. Again. be Then there exists a constant Zf D homomorphism mod I-. = T. then 2. Zf D = (cp) is principal. (1) (2) (3) The association D H .3. Zf f: X’ + X is a morphism deJined over F. Theorem 1.212 WEIL FUNCTIONS.(D). satisfying the following conditions. + constant. and D is algebraically equivalent to 0 on X. is a = (cp) is principal.2. so the Weil function defined above is determined up to an additive constant.2 are called N&on functions. Let A be an abelian variety dejined over F. = I. then Let a E A(F”).v such that Functions normalized as in Theorem They satisfy the additional property: (4) Let 1. AD = 1. relations between divi- . On arbitrary varieties. associated with D. We then obtain a bilinear pairing between divisors (algebraically equivalent to 0 on an arbitrary variety) and O-cycles of degree 0. Theorem 1.1. Having normalized the N&on functions up to additive constants. Then A.. To each divisor D algebraically equivalent to 0 on X one can associate a Weil function unique mod constant functions. Let T. then the divisor of a rational function determines the function up to a non-zero multiplicative constant. we can get rid of these constants if we evaluate these functions by additivity on O-cycles of degree 0 on A or X as the case may be. 0f mod r. To each divisor D on A there exists a Weil function I... such that f *D is dejined. Let X be a projective non-singular variety defined over F.. The normalizations of Weil functions as in the next two theorems are due to N&on [Ne 651. = 1. one cannot get such a general characterization without a further assumption. This pairing is called the N&on pairing or N&on symbol. mod I-. INTEGRAL POINTS CIK 011 If X is complete. then A. As with heights. and put D. f: B + A be a homomorphism of abelian varieties over F. y. 0f mod I-. translation by a.3 are called N&on functions..

Given E. I I More generally. q have the same order of magnitude.1 4 q2+E p/q in lowest form. Following [La 66a].BI .. $2 extend mutatis mutandis to algebraic families of N&on functions. let c1 be any real irrational number.2W) 5 log +(h(P)) states that Lebesgue for all rational B E Q. See [La 831. . the theorems concerning algebraic families of heights which we gave in Chapter III. we can use the height and rewrite the inequality as -log u . If a fraction p/q is close to ~1. then p. Chapter 12. THE THEOREMS OF ROTH AND SCHMIDT Let tl be an algebraic number. although possibly some additional restrictions on the function . In addition. 52. I -2logq~Elogq for all but a finite number of fractions p/q. so instead of log q in the above inequality.2h(p/q) 5 E log q. with q > 0. We refer to [Ne 651 or [La 831 for a list of such relations. [La 66~1 we define a type for c1 to be a positive increasing function I) such that -lwl~ .CIX? 921 THE THEOREMS OF ROTH AND SCHMIDT 213 sors are reflected in relations between the N&on functions or the N&on symbol. IX. The inequality can be rewritten -log I a-. A theorem of Khintchine almost all numbers c1have type rj if A basic question is whether Khintchine’s principle applies to algebraic numbers. one has the inequality Roth’s theorem states: I I a-P for all but a finite number of fractions z.

Let R. 55. it occurred to me to transpose my conjecture from number theory to Nevanlinna theory. and for the classical numbers. However. It becomes a problem to determine the type for each algebraic number. but not empty.. In the sixties. as follows. See also [La 831. Let S. The point is that one needs to solve linear equations and one needs to bound the solutions as a function of the height of the coefficients.. For instance. This form of Riemann-Roth guarantees the existence of many functions having sufficiently large absolute values at those VES. be a finite set of absolute values of F containing all the archimedean absolute values if any. thus giving rise to the error terms which have been stated in Chapter VIII. which is much better than the “probability” type and goes beyond Khintchine’s principle: the sum c l/q+(q) diverges. Riemann-Roth is precisely the tool which accomplishes this for us. at any rate. Roth’s theorem can be formulated as saying that an algebraic number has type 5 qE for every E > 0. INTEGRAL POINTS cm 921 might be needed. is a ring. [Ad 671 that e has type c 1% 4 W) = logloi% 4 with a suitable constant C. . there is no example of an algebraic number about which one knows that it is or is not of type (log q)k for some number k > 1. Let F be a finite extension of F. which all have bounded type (trivial exercise). denote the subset of elements x E F such that 14” 5 1 for all u E S. In light of Vojta’s analogy of Nevanlinna theory and the theory of heights. in terms of a function $ analogous to the Khintchine function.. Let F be a field with a family of proper absolute values satisfying the product formula. Under these hypotheses.214 WEIL FUNCTIONS. See also Bryuno [Bry 641 and [RDM 621. I conjectured that this could be improved to having a type along the Khintchine line. essentially a weak form of a Riemann-Roth theorem. In [La 60a] I pointed out that Roth’s theorem could be axiomatized to fit the general pattern of height theory. Then R. except for quadratic numbers. which is true in the number field case and the function field case. say with (log q)l+‘. = R. called the ring of S-integers. it follows from Adams’ work [Ad 661. Then the Roth-type inequality could be written -log I m-i I -2logqS(l +&)loglogq for all but a finite number of fractions p/q. one can formulate Roth’s theorem as follows. One needs to assume an additional property. Let R.

and assume v extended F(cl.O. . . lla. 921 THE THEOREMS OF ROTH AND SCHMIDT 215 to For each v E S let a. As Vojta remarked. .1. Let L be a finite set of in Q”. Let N be a positive integer.i(x)llv B SizeCPN-” for all but a finite number of x E og. are said to be in general position if M 5 N and the forms are linearly independent. in a way which makes the formal .S.M E L with M 2 N. A higher dimensional version of Roth’s theorem was proved by Schmidt [Schm 703. F and S. linear forms in N + 1 variables with coeficients “Qs . F and S. . . .x.WB) i MP) Thus the analogy between algebraic numbers and algebraic functions held also in this case.CIX. 0g max(l. and Vojta improved the statement of Schmidt’s theorem to the following. . N+l Also linear forms L 0. v0S. it is also useful to rewrite these inequalities in terms of heights.) define the size size(x) = max /[xi 11”. and let S be a finite set of absolute values containing the archimedean set S. be algebraic over F.i with xi E o~. see 96. To go further..~ be the ring of integers of F. .L. There exists a finite union Z of proper linear subspaces of QaCN+l) having the following property. Given a number field F.PM . Given E.LV. The finite set of exceptional points lying outside 2 still depends on E.. the elements /3 E F satisfying the condition &F&l have bounded height. In Schmidt’s version. the exceptional set 2 depends on E.Q II Lv. [Schm 80). or M > N and any N + 1 of them are linearly independent. localized at all primes not in S. and for each v E S given linear forms L. . .i’ lying outside 2. For an account of the method of proof. let F be a number field. Let o~. However.). For x = (x0 .~ so XEOF. but Vojta succeeded in eliminating this dependence [Vo 89~1. from Thue-Siegel to Vojta. The above statement reflects the way inequalities have been written on affine space. however. .. and following Vojta. . . we have for every E > 0 Theorem 2. this theorem was analogous to Cartan’s theorem in Nevanlinna theory. a finite set of absolute values S containing S.

Theorem 2. For a discussion of conjectures concerning similar inequalities when the points P are allowed to vary over all algebraic points. Vojta discusses which functions can reasonably occur. having to do with differential fields which provide still another context besides the number field case. For this purpose. a term must be added on the right-hand side involving the discriminant. We let R.216 WEIL FUNCTIONS. for instance f(N) = N.1 can then be formulated with Weil functions as follows. .. . ...&‘) . Theorem 2. [OS 851 for diophantine approximations and Nevanlinna theory. or holomorphic Nevanlinna case. we have i$ mHi. . .5 1 for v $ S. function field case.x. of the form f(N) d(P) for some function f(N). a set of absolute values S. . Under such less restrictive conditions. which would result from his general conjectures. Let H. which we state in 54. IX. be the subring of F consisting of those elements x E F such that 1x1. and let S be a finite set of these absolute values containing all the archimedean ones if such exists.. then we can define a Wed function associated with H by the formula &r. if L is a linear form on projective space P” and H is the hyperplane defined by L = 0. Since already in Roth’s theorem one does not know how to improve the type from qE to a power of log q or better. INTEGRAL POINTS Let F be a field with a proper set of absolute values.W + l)W’) 5 W’) except for a finite number of points in PN(F) outside Z.HM be hyperplanes in general position in PN over Q”. the good error term with the complexity of the divisor suggests the ultimate answer in this higher dimensional case.“(P) = -1% IWN” max I lxilu for any point P E PN(F) with coordinates P = (x.) and xi E F. a fortiori no such result is known for the higher dimensional Schmidt case. Finally it is appropriate to mention here the direction given by Osgood [OS 811. There exists a set Z equal to a finite union of hyperplanes having the following property. . $3.2. INTEGRAL POINTS cw 931 analogy with Nevanlinna theory clearer. Given a number field F. . But as we remarked in the Nevanlinna case. and E > 0. . see [Vo 89~1.

We call (x 1. such that b. ..) a set of affine coordinates on K Let R be a subring of some field containing F. all P E I.e. .x.“(P) 5 G for all u 6 S... or if the genus of X is 0. Then F[V) is finitely generated. . P E I. Let F be a Jield finitely generated over Q and let R be a subring jinitely generated over Z.CK §31 INTEGRAL POINTS 217 Let V be an affine variety defined over F.(P) E RF. . We let F[V] be the global ring of functions on JJ. Theorem 3. or R-integral.) is a set of affine coordinates integralizing the points in I. the theorem was extended to the more general rings in .u = log max(L lxl(p)l. If R = R.]. he defines the notion of (S. we can write F[V] = FIX1. D)integralizable set of rational points by using the condition stated above in terms of the Weil functions.(P) E R for all i and all P E I.. such that bx.1. Let T/ be the complement of the hyperplane at infinity in projective space P”.). for each u 4 S. When F is a number field and R is the ring of integers. .. . but there are at least three points in the complement of V in X. i. A set of points in V(F) is S-integral if and only if there exists a set of affine coordinates such that the values xi(P) have bounded denominators for all i and all P E I. then every set of R-integralizable points on V is jinite. . Mp)I.x. Zf the genus of X is 2 1.x. The function field of V is the quotient field F(I/) = F(x. then we also say that I is S-integralizable.. the theorem is due to Siegel [Sie 291. so with non-affine open sets V. applicable in this more general case.s for all i. We say that I is R-integralizable.). for instance certain moduli varieties. if (x1. . By bounded denominator. if there exists a set of affine coordinates such that x.. . This follows immediately from the definitions. . . .X. we mean that there exists b # 0 in R. and let D be this hyperplane. . After the work of Mahler [Mah 333 for curves of genus 1 over Q. A subset Z of Y(F) is S-integral if and only if there exists a Weil function A. we can take the Weil function to be AD... One basic theorem about integral points concerns curves.. For instance.. . or S-integral. For this purpose.. Let V be an afine curve defined over F and let X be its projective completion. D)-integrality or (S. In [Vo 871 Vojta works with possibly non-ample effective divisors D. . and let I be a subset of points of 1/ rational over the quotient field of R.

) zs 1% max(l. and we let cpE F(X) be a non-constant function. so we have heights. Let us consider first a curve defined over a finite extension of a field F with a proper set of absolute values satisfying the product formula.218 WEIL FUNCTIONS. we consider the set of points in X(F) such that q(P) E R. which will define integrability for us. together with a lifting procedure using coverings of the curve. bounds on the heights of integral points may be of a different type than bounds for the height of rational points.2. and cp E F(X) not constant. proposition shows what it implies for curves of higher genus. INTEGRAL POINTS CIX? §31 [La 60a]. Note that when we have the factor E on the right-hand side. projective imbedding. That is. Then given set of points P E X(F) such that The next E > 0 the logIcpO’)I. Suppose that X has genus 2 1.) 2 tcrh(P) . Then the set of points P E X(F) which are not among the poles of q. However.. Ilcp(P)II. Let r be the largest of the multiplicities of the poles of cp. Proposition 3. Proposition 3. We let X be the complete non-singular curve. since the estimate applies to each term. The method is of sufficient interest so we shall describe it briefly in the form given in [La 60a]. Let K be a number > 2 and C a number > 0. The first proposition applies to a curve of any genus 2 0. We state both steps as propositions. so quantitative forms of the theorem are of interest independently of Faltings’ theorem. In light of Faltings’ theorem. only the cases of genus 1 and genus 0 are relevant in the qualitative statement we have given. even in higher genus. and are such that & have bounded height. The inequa- . Siegel’s method uses Roth’s theorem (in whatever weaker form it was available at the time). and the decomposition of heights into a sum of Weil functions. the sum becomes irrelevant. Let X be a projective non-singular curve dejined over F. This is accomplished by putting together a geometric formulation of Roth’s theorem.. We may call these the q-integral points.3. defined over F. 2 W’) has bounded height. and we want to show that they have bounded height.. Let S be a Jinite set of absolute values.C to the given (The height h is taken with respect The above proposition is merely a version of Roth’s theorem.

Proposition 3. Then there is a point POE J(F) such that all the points P E I in that coset can be written in the form P=mQ+P. applied to the function cp 0 f. As a first example.CK §31 INTEGRAL POINTS 219 lity for curves of genus > 0 is reduced to the inequality for curves in general by means of the method described in the next proposition..4 is unramified. using the functorial properties of the height. a2 E R (where R is finitely generated over Z) and ul. We restrict the covering J -+ J given by x~mx + P. as functions on X’(Fa). the zeros of cp and of cp of have the same multiplicities. and assumethat J(F)/mJ(F) is finite.4. we have &jL. and a projective imbedding of X’ such that h of = m2h’ + o(h’) for h’ + co.3 follows at once. Then r/rm is finite. so Proposition 3. Indeed.) h(cp(f’)) W’) = 2 for some E > 0. in their projective imbeddings. Since the covering of Proposition 3. The unramified covering f: X’ + X is obtained from the Jacobian. Let I be an infinite set of points in X(F). such that f induces an injection of I’ into I.3 shows that the height of integral points is bounded. Let m be an integer > 0. Let J be the Jacobian of X over F. The heights h and h’ refer to the heights on X and X’ respectively.’ 0g max(L Ilcp(P)II. .4. an injinite set of rational points I’ c X’(F). unequal to the characteristic of F. and especially the quadraticity. consider the equation (called the unit equation) (*) alul + a2u2 = 1 with a.2 in combination with Proposition 3. suppose infinitely many points P E I lie in the same coset of J(F)/mJ(F). As an application to q-integral points P. We have emphasized the method of proof because variations and substantial extensions occur systematically in the theory. We use Proposition 3. to the curve to obtain Proposition 3. Thus we have a form of descent by coverings.. Then there exist an unramijied cooering f: X’ +X dejined over F.. u2 are to lie in a finitely generated multiplicative group r. Suppose that X has genus 2 1.4. Applying Proposition 3.

INTEGRAL POINTS CK 931 Writing lJi = Wi”bi with wi E P. this method is inefficient for the unit equation. Let A be an abelian variety dejned over a jinitely generated field over the rational numbers. 1 H(P)“’ and h. Folgerung 2.6 (Faltings [Fa 901). and results showing a much tighter structure are conjecturable.” = 1. Let Z be a subvariety of A over a number field F.Z such that dist. Before we state his inequality. Faltings proves his theorem by going through the higher dimensional analogue of Proposition 3. and for m 2 3 the new equation (**) has genus 2 1 so we can apply Theorem 3.w. However.w. I conjectured [La 60a] and Faltings proved [Fa 901: Theorem 3. as an application of his p-adic extension of the Thue-Siegel theorem (pre-Roth version) [Mah 331. Theorem 3.5 (Faltings). divisor E. In higher dimension.(P).5 follows from Vojta’s conjectures which will be mentioned in $4. Originally. (*) in P give rise to we see that infinitely many solutions of the equation infinitely many solutions of the equation (**I a. is the height with respect to any ample . Z) in a natural way.(P. Then every subset of R-integral points in V(F) is Jinite.” + a. note that for each absolute value v and subvariety 2 of A one can define a v-adic distance d. The case of genus 0 in Theorem 3.220 WEIL FUNCTIONS.3. there is only a finite number of rational points P E A(F) . Z) < __ where H(P) = exp h. Note also that Theorem 3. the equation x1 + x2 = x3 in relatively prime integers divisible by only a finite number of primes over Z was considered by Mahler. Let V be an afJine open subset of A.3 by taking similar ramified coverings. Given E > 0. It was considered as a “unit equation” (for units of a finitely generated ring) explicitly in [La 60a].b. as we shall do in $7.b.(P.1 to see that (*) has only finitely many solutions with Ui E P. and give a general framework for this type of finiteness. Fix an absolute value v on F.1 is reduced to the case of genus 2 1 and Proposition 3. Let R be a jnitely generated ring over Z contained in some finitely generated jield F.

and for rational points the term with the discriminant d(P) does not appear. And in higher dimension. In the direction of differential equations. Let V be an a&e curve over F. namely: Theorem 3.8.C~X 931 INTEGRAL POINTS 221 If Z is a divisor D. [OS 851. On abelian varieties I had conjectured actually a stronger version of such an inequality [La 741 which we shall discuss in §7. or of genus 0 but with at least three points at infinity in its projective completion. Faltings’ inequality fits the general type of Vojta inequality stated below in Conjecture 4. Then a set of R-integral points in V(F) has bounded height. Also in higher dimension. there are also the relative cases of finiteness. in terms of Weil functions and a finite number of absolute values S. sion 1 and higher dimension. .1. I conjectured the analogue: Let V be an afine open subset of an abelian variety defined over the function field F. of genus 2 1.D is contained in a finite union of linear subspaces of . . Let k be an algebraically characteristic 0. Let F be a number field and let H.Hq be hyperplanes in general position in P”. . This conjecture was proved by Parshin [Par 861 using his hyperbolicity method under the additional assumption that the hyperplane at infinity does not contain the translation of an abelian subvariety of dimension 1 1. and let F be a function field over k. In [La 60a] I proved: both in dimenclosed field of R be a finitely of some affine Theorem 3. .Z. we can rewrite this inequality in the equivalent form for all but a finite number of points P E A(F) . because the canonical class on an abelian variety is 0. Let S be a finite set of absolute values and let D=cHi. Then every set of R-integralizable points in V(F) has bounded height.7. see 96. For comments on the proof. and so is finite modulo the Ffk-trace. Let generated subring of F over k.. the number theoretic analogue of Bore13 theorem that the complement of 2n + 1 hyperplanes in general position in P” is hyperbolic was proved by Ru and Wong [RuW 901. Then for every integer 1 5 r 5 n the set of (S. so R is the affine ring variety over k. Of course. see also Osgood [OS 811. D)-integralizable points in P”(F) .

(b) A restriction was made for the algebraic points to have bounded degree.(P) + O. The use of these weights is related to Vojta’s extension of Schmidt’s theorem as stated in Theorem 2. even with a Khintchine-type function. In current applications. To what extent is there such independence in the more general case at hand? From my point of view.D is jinite.(l). and let K be the canonical class on X. Let D be a divisor with simple normal crossings on X. IX. provided that q > 2n . Let S be a jinite set of absolute values containing the archimedean ones. D)-integralizable points of P”(F) .222 WEIL FUNCTIONS. Vojta showed how the exceptional set does not depend on E. Let r be a given positive integer and E > 0. such that Conjecture for all points P E X(Q”) not lying in Z for which [F(P) : F] s r. In particular. The framework covers both rational points and integral points.1. The proof extends the proof of Schmidt’s subspace theorem. Several comments need to be made concerning the extent to which certain hypotheses are needed in this conjecture. 4. the error term should anyhow be of the form W’) + (1 + 4 log h. Chapter VIII. Vojta has raised the possibility that these hypotheses can be weakened as follows. $5. Let E be a pseudo-ampledivisor on X. Indeed. (a) In his improvement of Schmidt’s theorem. as in [Vo 871.r + 1. if q 2 2n + 1 the set of (S. the estimate of the conjecture suffices to imply numerous other conjectures concerning rational points in X(F). but the relation has not yet been made explicitly. INTEGRAL POINTS CIX 041 P”(F) of dimension r . We work over number fields. which are seen as special cases.1. There exists a proper Zariski closed subset Z of X (depending on the above data). VOJTA’S CONJECTURES These conjectures provide a general framework for a large number of previous results or conjectures.1. Let X be a projective non-singular variety over a number jield F. as for the error terms in Nevanlinna theory. by refining the approximation method using certain weights for the approximation functions. 54. the various proofs of implication rely on the con- .

Chapter VIII. Vojta shows in several cases how his theorem applies to a blow up of the divisor which does have normal crossings. The two parts are quite separate. Sometimes one wants to apply an estimate on heights with respect to a divisor which comes up naturally but does not satisfy that hypothesis. Let S be a jinite set of places. Corollary 4. The question Vojta raises in his inequalities is whether the more refined estimates for the canonical height also hold uniformly.1 for rational points on an abelian variety was already given in [La 741 p. See [Cher 901 and [Cher 911. 783 and [La 641. But the approach by considering linear .1 (so the corollary is itself conjectural). We note that concerning the abelian varieties lowing corollary Vojta proved that his conjecture implied my conjecture finiteness of integral points on affine open subsets of [Vo 871 Chapter 4. Let K be the canonical class and assume that K + D is pseudo ample. The fact that the degree occurs only as a factor without any extraneous constant term provides evidence that Vojta’s conjecture should follow a similar pattern and be valid uniformly for all algebraic points.2. and on the other hand. For instance. When I wrote up whatever results were known around 1960. to what extent does the term O. Theorem 5. The entire basic theory which makes the heights functorial with respect to relations among divisor classes goes through for X(P).6. A current result of William Cherry in the analogous case of Nevanlinna theory shows that the analogue of Vojta’s conjecture is true. Cf. A formulation for the inequality in Vojta’s Conjecture 4. The reader will note the persistent hypothesis that the divisor D in the statements of theorems and conjectures has simple normal crossings. Then an (S. $2. and let D be a divisor with simple normal crossings. D)-integralizable set of points in X(F) is not Zariski dense in X. Dealing with all algebraic points in X(P) makes this separation quite clear.CK §41 VOJTA’S CONJECTURES 223 struction of coverings which lift rational points to points of bounded degree. the estimate applied to the blow up gives the expected estimates on the heights. imply that certain sets of points have bounded height. Let X be a non-singular projective variety dejined over a number jield F.(l) depend on the degree of the points. 51. I did not consider algebraic points with the corresponding estimate of the logarithmic discriminant. In each case. He obtains this from the folof Conjecture 4. I was careful to separate the parts of the proofs which. show that certain sets of bounded height satisfy certain finiteness conditions. on the one hand. in the context of diophantine approximation on toruses. Thus the stronger property that the inequality should hold without the restriction of bounded degree would exhibit a phenomenon which has not been directly encountered in the applications. with essentially best possible error terms with a Khintchine type function.

Then there exists a proper Zariski closed subset Z of X depending on the previous data. and [La 831.224 WEIL FUNCTIONS. Vojta has shown that the case of Conjecture 4. Then for every pair of points P E X(Fa) and Q = f(P).4 is a generalization to the ramified case of a classical theorem of Chevalley-Weil which we give as a corollary.U’) + O(1). By using ramified coverings. X’ be projective non-singular varieties over a number jield F. and both contain the discriminant term on the right-hand side. which may be ramified.(l) for all points P E X(Fa) .m. INTEGRAL POINTS CIX 041 combinations diophantine O(log h. Then for all P E X(Fa) .. if dim X = 1 so X is a curve. Let S be a jinite set of absolute values. See 97. 4. Let f:X+X be a jinite surjective morphism. Let E be a pseudo-ampledivisor on X.supp(Z) for which f(P) E X’(F). such that mdp) + MP) 5 d(P) + h(P) + O. Let E> 0..3 and also Conjecture 4. Let f: X + X’ be an unram$ed jinite covering of projective non-singular varieties over a number field F. To show this consistency and other matters.supp(A) we have d(P) . Vojta compares the discriminant term in coverings as follows. Let D be a divisor with simple normal crossings on X..d(fU’)) 5 K. the relative discriminant of F(P) over F(Q) divides a $xed integer d. Vojta’s Theorem 4. . Hence the conjectures must be consistent with taking finite coverings. Let X. Let S be a jinite set of absolute values. [We 351.. Let A be the rami$cation divisor of f.4. and briefly NA..(P)) conjecture of Conjecture of logarithms (ordinary or abelian) and their properties of approximation led to the conjectured better error term rather than eh#). Similarly.5 (Chevalley-Wed). then the case of Conjecture 4. See [ChW 321. Let f: X + X’ be a generically Jinite surjectiue morphism of projective non-singular varieties over a number field F. Corollary 4.1 with .1 are applied to coverings. Conjecture 4.3. previously.. Theorem 4. implies the general case with a divisor D. We now consider a second Vojta having to do with coverings.3 when D = 0.s = h. We defined N.

where 43 is the upper half plane. Actually. Example. the Chevalley-Weil theorem was proved for normal varieties. Let f: X +X’ X(F) is jinite for all number jields F if and only if X’(F) is Jinite for all number jields F. in the unramified case. one obtains: be as in the previous corollary. See [Vo 873. and let X(N) be the modular curve over Q of level N. Conjecture 5.4. Since there is only a finite number of number fields of bounded degree and bounded discriminant.1. Since it is not known if Kobayashi hyperbolicity for V(C) is equivalent to Brody hyperbolicity for I’(C). Then Corollary 4. If V(C) is hyperbolically imbedded in a projective closure. Technical remark. Then there is a correspondence n X(2n) =‘\ P’ J such that the liftings of @” and X(2n) Cf. we have X(N)(C) z I(N)\@. Over the complex numbers. .CIX? VI D = 0 implies CONNECTION WITH HYPERBOLICITY 225 the general case with a divisor D. there is some problem today about being sure what form the transposition of my conjecture to the affine case takes for affine varieties. The non-singularity is a convenient assumption when there are singularities in the ramified case. I formulated one possibility as follows in [La 873.1. $5.6. over each other are unramified. then V has only a finite number of integral points in every jinitely generated ring over Z. Let @)nbe the Fermat curve of degree n. Kubert-Lang [KuL 753. IX. Proposi- tion 5. CONNECTION WITH HYPERBOLICITY We have already remarked that Parshin used a hyperbolic method to prove part of the function field conjectures on rational and integral points on subvarieties of abelian varieties. whereby a finiteness condition on integral points is equivalent to some hyperbolicity condition. Let V be an afJine variety over a number field contained in the complex numbers.

Theorem 5. every Note: “Components” in NS(X) and D are meant to be F-irreducible components. The question is what does “sufficiently large” mean. The problem is to what extent does one need additional restrictions near the boundary where the Kobayashi distance may degenerate (perhaps none as in Faltings’ Theorem 3. D)-integralizable set of points in X(F) is not Zariski dense. even degree 1 which means hyperplanes. see [La 871. or one can take out a divisor of high degree. Fujimoto.2. 1)-forms as follows. Let D be a divisor on X consisting of at least dimX+p+r+l components. Let S be a finite set of absolute values of F. Let D be a divisor with normal crossings on X and let V = X . that the diophantine condition implies that V(C) is Brody hyperbolic. the problem is to prove that such hyperbolic non-compact varieties have only a finite number of integral points. Vojta proved [Vo 871. $7).1. 1)-form w on V(C) which is . INTEGRAL POINTS CIJC051 Lacking the equivalence of the hyperbolicity conditions. Chapter 5. Green in this direction. not necessarily geometrically irreducible. Chapter VII. which is less developed than the compact case. Theorem 2. Furthermore. In higher dimension the complement of 2n + 1 hyperplanes in general position in P” is Kobayashi hyperbolic. where A’ is the Picard variety of X and let p be the rank of the N&on-Severi group NS(X). For our purposes here. if we take out three points from P’. there is no systematic theory giving conditions for hyperbolicity in the non-compact case. Assume that there exists a positive (1. under the strongest possible form of hyperbolicity. Vojta has given a quantitative form to estimates for the heights of integral points.3 (The (1. Then (S. 92. Let r be the rank of the Mordell-Weil group A’(F). Vojta’s improvement of Schmidt’s theorem also lies in this direction. A theorem of Griffiths [Gri 711 asserts that one can always take out a divisor such that the remaining variety has a bounded domain as covering space. by using hyperbolic (1. which is one of the strongest forms of hyperbolicity. For examples of Bloch.5). Let X be a projective non-singular variety over a number field F contained in C. then the remaining open set is Brody hyperbolic (Picard’s theorem). Classically. 1)-form conjecture [Vo 871. Roughly speaking. Today.226 WEIL FUNCTIONS. In that line. Conjecture 5. taking out a sufficiently large divisor from a projective variety leaves a hyperbolic variety.D. A general idea is that one can take out several irreducible divisors of low degree. Let X be a projective non-singular variety over a number field F. I would formulate the converse only more weakly.4. of which I expect that it has only a finite number of integral points as above.

over a number field F of degree 5 r. if f: D + V(C) is a non-constant holomorphic map. r. that there exists a constant B > 0 such that. acting as an exceptional set. Vojta applies the (1. Let S be a finite set of absolute values. we have 444 5 . Also the error term EhE should be replaced by O(log h. If the set I is contained in the rational then the conjecture implies that I is finite. he proves that Conjecture 5. positive integers n. The question whether the restriction that the points should have bounded degree applies as well to the present case. n. In this conjecture. We mention two of them. E) such that for every semistable principally polarized abelian variety A of dimension n and good reduction outside S. Remark 4. Remark 3. Also assume that w 2 cl(p) for some metric p on a line sheaf 9 on X. 1)-form conjecture to deduce several number theoretic applications. + E d(F) + C. then Ric f *w 2 Bf *co.4. points X(F). Given a jinite set of places S. Chapter 5. I expect it to be equally superfluous in the present arithmetic case. by applying the conjecture to the moduli space and its boundary divisor. and E > 0. D)-integralizable points of bounded degree over F. there is no Zariski closed subset Remark 2. Let I be a set of (S. Specifically. and in fact.3 implies the Shafaverich conjecture.CIX §51 CONNECTION WITH HYPERBOLlCITY 227 strongly hyperbolic. In light of the analogous result in Nevanlinna theory which motivated the conjecture. Then for all points P E I we have h. and implies: Corollary 5. Let E be pseudo ample on X.3 gives somewhat more uniformity. Remark 1. > ( .(P) 5 Ad(P) + ehE(P) + Q(1). First he proves in [Vo 871: Conjecture 5. 57. but for which the assumption that D has normal crossings turned out to be superfluous. there exists a number C = C(S. r.J or better. in [Vo 871.

See [Vo 90a]. qJS+E 4 Of course this was short of the conjectured result ultimately proved by Roth. Gelfond. and culminated with the recent work of Vojta who combined all the aspects of previous work into an Arakelov context. IX. and also for the height. .3 is applied in a rather delicate case. Dyson. [Vo 90b]. the logic starting from such numbers fil and fi2 is such that the proof is not effective. thus expanding enormously the domain of applicability of this method. In a relatively early version of determining the best approximations of algebraic numbers by rational numbers. and also such that the quotient of the heights h(PJh(&) is large. then such &. Note that the moduli space is not affine. Roth. q E Z. Viola.228 WEIL FUNCTIONS. both in the results achieved and its prospects. [Vo 90~). but it sufficed to prove the finiteness of integral points on curves as discussed in $3. as distinguished from the notion of integral points on affine varieties. If there are infinitely many solutions to the above inequality. $6. since we don’t have an effective starting point for the existence of &. However. One then shows that there exists a polynomial G(T. one had the Thue-SiegelDysonGelfond result: Given E > 0 and an algebraic number c1 of degree n over Q. q > 0) such that I I a-5 1 s------. Second. INTEGRAL POINTS CIX WI But this corollary is itself conjectural. similar to those in the function field case. there are only finitely many rational numbers p/q (p. so Conjecture 5. and similar to those which we recalled in Chapter VII. 1)-form conjecture implies a bound on (O&r) in Arakelov theory. but something can be said to give an idea of this program. . The method of proof used two approximations & = PI/q1 and & = p2/q2 such that /?i and & have large heights. &. FROM THUE-SIEGEL AND FALTINGS TO VOJTA A basic approximation method which started with Thue-Siegel went through a number of developments due to Schneider. Faltings boosted the method further by proving a higher dimensional result [Fa 901. 7”) with integer coefficients which are not too large. Schmidt. Vojta shows in [Vo 881 how the (1. Vojta’s current program is still in progress. D)-integralizability is used in a rather strong way. and the notion of (S. /& can be found. We begin by a few words concerning intermediate results before Roth’s theorem.

some derivative does not vanish at (h. . . Schneider [Schn 361 showed that instead of using two approximating numbers. by using methods of algebraic geometry having to do with the analysis of singularities.&. yielding the full Roth theorem. T. and a fairly large t. one can then estimate the height of 5 from above to be small because many derivatives of G vanish at (a. d. a) = 0 for i.) in m variables. . . satisfying certain linear conditions..[IX WI FROM THUE-SIEGEL TO VOJTA AND FALTINGS 229 such that G vanishes of high order at (a. not too large. . . &. Since 5 is a rational number. and constructing a polynomial G( TI. B2X that is t = D”D’2G(&. That approach was vastly expanded by Esnault-Viehweg [EV 841 to get a result in arbitrarily many variables. He showed how the appearance of the 2 + E is due to a combinatorial probabilistic estimate in the course of the proof having nothing to do with algebraic numbers. .y-G(&. by using arbitrarily many. j. this height has a lower bound depending only on the original degree of G and the heights of &. . D. that is Df’D$G(a.+“r<t 1 4 namely with suitable positive integers d. then one would obtain the desired approximation with q’+&. . . one shows that in fact. /$) # 0 with jr. Such a lemma was provided especially by Dyson and Gelfond. . Bm. on the other hand.) does not vanish..‘. However. as . say /?r. Viola [Vio 851 proved a non-vanishing result at the level of the Dyson-Gelfond lemma.. . Schneider did not see how to complete the second step in the proof where one needs that a suitable derivative D+ . however. was not well suited for extensions to algebraic geometric contexts. However. Solutions of linear equations and linear inequalities in the above construction can be found to yield a contradiction at this point. and to prove this second step by means of a classical Wronskian method which he saw how to adjust to yield the non-vanishing of the derivative. Although that method did not yield the q2+‘. LX). and so he obtained only a partial result. The Wronskian method. Esnault-Viehweg work on products of projective spaces. . it did yield a fruitful alternative approach to one of the key steps in the proof of that theorem. Roth’s achievement was to see very clearly through the entire situation. i.. . a). Then in a crucial lemma.

.‘. .. say. I specifically pointed out that the construction of the polynomial G(T. (b) By globalizing and sheafifying not only on the projective line but on curves of arbitrary genus over. and we consider the product C x C. . Furthermore. the product P’ x P1 of the projective line with itself. We suppose xi(Pi) = 0. among many other patterns.230 WEIL FUNCTIONS. See [Vo 89b] for the function field case. including partial differential equations. readers might prefer to read proofs which do use Riemann-Roth theorems. Let s be a section of a line sheaf 5Z on C x C. In Diophantine Geometry ([La 621. /I2 amounts to a choice of point on the product A’ x A’ of the affine line with itself. . For i = 1. In Roth’s theorem. and other tools from algebraic geometry. weaker versions of Roth’s theorem. their result does not yet give what is needed for Schmidt’s theorem. Vojta developed his method first in the function field case. not only my adjustments of Roth’s own proof but Vojta’s subsequent contributions which jazz things up even more: (a) By making the analogy with Nevanlinna theory. INTEGRAL POINTS CK WI they point out. Vojta needs a Hirzebruch-Grothendieck-Riemann-Roth along the most substantial lines developed in Arakelov theory by Gillet-Soule in relative dimension > 1. T’. the ring of integers of a number field. to show the connections not only with algebraic geometry. thus showing once again the effectiveness of the analogy. Using such Riemann-Roth theorem. the choice of &.i. In this case. see also [La 831) I axiomatized the Roth proof so that it applied also to the function field case. he was able to give an entirely new proof of Mordell’s conjecture-Faltings’ theorem. Mordell in his 1964 review of that book commented: “The reader might prefer to read [Roth’s proof] which requires only a knowledge of elementary algebra and then he need not be troubled with axioms which are very weak forms of the Riemann-Roth theorem. by following a pattern stemming from the previous. defined in a neighborhood of P by a formal power series f(Xl? X2) = C i. and then translated the method into the number field case. We describe Vojta’s method at greater length. x2) are coordinates on C x C at P.” On the other hand. or if you wish. We now consider a projective non-singular curve C of genus g 2 2 defined over a number field F.) depends on solving linear equations which amounted to a weak form of the RiemannRoch theorem.CO Uili2Xf1Xp. So different people prefer different things at different times. analysis. but with analysis and Arakelov theory. . We let P = (P. . PJ be a point in C(F) x C(F). Then (x1. and [Vo 90a] for the number field case. and diophantine approximations. . 2 let xi be a local coordinate on C in a neighborhood of Pi.

(5) Get a contradiction between this lower bound and an upper bound obtained by a suitably globalized version of Dyson’s lemma.)/h(P. We let n:X-+Y be a regular semi-stable family of curves over Y with generic jiber X. :+:<t.) with suitably large m has a section s. The difficulties which arose previously exist in even stronger form in the present globalized context. d. to be the largest real number t = t(s. As usual. P2 on C such that h(P. of Step 2 at the point P = (PI.) such that for all pairs (i1. (2) Show that there exists a thickening of D. The main steps of Vojta’s proof then run as follows. and that the tangents for both branches of the jibers (in the complete local ring) are also rational over the residue jield. motivated by Viola’s proof. find a divisor D. P2). In addition. = C. P..)..IX WI FROM THUE-SIEGEL TO VOJTA AND FALTINGS 231 Let d. 1 2 If D is an the index tions are the notion effective divisor on C x C. that is. The above definiindependent of the choice of local parameters. d. (4) Prove a lower bound for the index of the section s. then we define the index of D to be of a section s on 6&. such that the line sheaf O(mZ. we let Y = spec(o. The semistability and rationality assumptions can always be realized over a finite extension of a given base. on C x C which is ample if r is sufficiently large. d.) is sufficiently large. and globalize of index stemming from Dyson-Gelfond and Schneider-Roth.-(D) for which (s) = D. The first one to be surmounted was Vojta’s globalization of Dyson’s lemma [Vo 89a]. We define the index of s at P relative to d. also such that certain sphere packing conditions are satisfied. on the generic fiber and satisfies certain bounds at infinity. (1) For certain values of a parameter r. and h(P. i2) of natural numbers satisfying .) where or is the ring of integers of F.. . be positive integers. when r is close to h(P. d. we assume that all double points occurring on the geometric fibers of IK are rational over the residue class fields. to a divisor on X xY X. find an Arakelov divisor & on X xr X which restricts to D.) is sufficiently large.)/h(P. satisfying certain bounds at the infinite places. (3) Choose rational points Pi. .

. .P.... Hirzebruch-Grothendieck-Riemann-Roth sufficed. P2 to get a better measure of approximation corresponding to Roth’s theorem instead of ThueeSiegell Dyson’s theorem. Faltings uses a new method of algebraic geometry to show that some suitable derivative does not vanish in step (4). He succeeded in proving thereby that a subvariety of an abelian variety. A simplification of Vojta’s proof also eliminating the use of Arakelov .232 WEIL FUNCTIONS. Faltings [Fa 901 influenced by Vojta’s paper. Finally. not containing the translation of an abelian subvariety of dimension > 0. has only a finite number of rational points in any number field F. w h ere D is a divisor consisting of distinct points on C. Instead he uses a globalized version of a lemma of Siegel. Whereas in the function field case. the situation is in flux. Vojta complemented these results to fit his situation. . 43. Faltings works with a sequence of points PI.. .JP). and forward along the lines of Vojta’s conjectures stated in $4. . and finite unramified quotients of the above. and giving suitable bounds for the heights of the solutions in terms of the height of the coefficients. The second direction is to extend the result to include the proximity term m. so it is not clear what use will be made in the future of Vojta’s or Faltings versions of the general method globalizing the Roth-Schmidt theorems. which in Vojta’s case is applied to X xY X. This required estimating eigenvalues of Laplacians and applying such bounds to the geometric situation of metrized line sheaves. or any finitely generated field over the rationals. such that the ratios of successive heights is large. except of course for the product of curves. The existence of a section in the standard context of algebraic geometry (without metrics on line sheaves) is given by the Hirzebruch-Grothendieck-Riemann-Roth theorem. . restructured and extended the diophantine approximation method. instead of using something like Dyson’s lemma. On the other hand. finite unramified covers. Faltings eliminated the use of the Gillet-Soule Riemann-Roth theorem in higher dimension to obtain the desired section in step (2). the number field case required an Arakelov version since bounds at infinity have to be taken into account. At the moment of writing. See Chapter VI. he develops the use of points P. Vojta’s work is in progress in at least two directions: First. P. Faltings’ result was the first time that a variety of dimension > 1 was proved to be Mordellic. They could not be applied directly since certain positivity conditions were assumed previously on the metrics. In any case. solving integrally equations with integer coefficients (or algebraic integers)... so Vojta had to provide additional work to deal with certain metrics which are not positive. not just two points. having relative dimension 2. INTEGRAL POINTS IX %I A much more elaborate difficulty came from proving the existence of the section s. in step (2)..instead of PI. . Major results in this direction were obtained by Gillet-Soul& [GiS 881 and BismuttVasserot [BiV 891. Secondly.

say B. We have already seen analogies between these two cases. following [La 641 and [La 741.1. then there is some root a # 0 of the rational function q such that fi is close to c(. But as we have also seen. and so the group of complex points of an abelian variety. we gave a geometric formulation of Roth’s theorem. notably in the formulation of results or conjectures describing the intersection of subvarieties of semiabelian varieties with finitely generated subgroups. We let G = G. up to - . The worst possible case is that in which this multiplicity is m. DIOPHANTINE ON TORUSES APPROXIMATION We have used the word “torus” in two senses: one sense is that of complex torus. serviceable to study all integral a finitely generated subgroup of the multiplicative group G(F). We shall go more deeply into this question here. Then given E. IX. the other sense is that of linear torus. be the multiplicative group. Conjecture 7. A point P E G(F) is represented by an element of F. a group variety isomorphic to a product of multiplicative groups over an algebraically closed field. Iv(B)1 is approximately equal to 1~1 PI”. and let m be the maximum of the multiplicities of the zeros on G (so distinct from 0 or co). The multiplicity of tl in a factorization of cp is at most m. Let F be a number field and let l. we also want to consider the more special situation when we restrict our attention to a finitely generated subgroup of the multiplicative group. In that case. In Proposition 3. If /I is close to CI. Let r be the rank of r. 1 Iv(‘)1 < h(p)‘m+& We shall transform inequality (1). the height of points P in r which are bounded away from 0 and co and satisfy the inequality (1) is bounded. In addition.2. Let us start with the linear case. The function field F(G) is generated by a single function. 57. so let us start with a conjecture from [La 641.CK 971 DIOPHANTINE APPROXIMATION ON TORUSES 233 Gillet-Soule theory (but still using Riemann-Roth) was given by Bombieri [Born 901. then its distance from any other zero or pole of cp is bounded away from 0 (approximately by the distance of c( itself from another zero or pole). Let cp be a nonconstant rational function on G defined over F. and cp is just a rational function. this simplification also shows how to use Roth’s lemma instead of the version following Viola. that is. and if cp(p) is small in absolute value.

not all 0 such that 14151+ . Hence the exponent rm in the conjecture cannot be improved upon. Khintchine’s theorem already mentioned in §2. and l4il 5 4. it is easy to see that q >x< h. . . .1(l) we may assume that we have always the same [.. in dealing with solutions of inequality 7. . 1~1 /?I is of the same order of magnitude as /log c( . . . We have therefore transferred our diophantine approximation on the multiplicative group over a number field into an inhomogeneous approximation on the additive group. + cL5A<<&& 4 with q = maxIqil. the solutions of the inequality l&+1 + 4151 + ..41 log BI . . . . . and gives rise to r + 1 free choices of the coefficients ql. . + w&l <<+. Let &. . The period 2ni of the exponential function contributes one term to the sum on the left. Define 4 = 4(P) = maxI%l. WEIL FUNCTIONS. Since [ ranges over a finite set. INTEGRAL POINTS CIK §71 Hence our inequality amounts to 1 Iu .5.+1 of real numbers. A standard application of Dirichlet’s box principle shows that r cannot be replaced by a smaller exponent on the right-hand side. be free generators group. . modulo the group of roots of unity.B..234 a constant factor. . .BI << ho’+“’ This approximation can be transferred to the additive group via the of r modulo its torsion logarithm. Let u0 = log(cr[).qr+l. Then we may rewrite our inequality in the form I% . i. .. imply that for Lebesgue almost all sets 5 13 .4rlog Pr + %+124 < + where the log is one fixed value of the logarithm.e. given real numbers cl.. Then we can write for some torsion element [.& and an integer q > 0 there exist integers ql. . As a function on I-. Furthermore.*.q. .. .log PI. and suitable generalizations by Schmidt [Schm 601. In fact. .. 5.

. $71 DIOPHANTINE APPROXIMATION ON TORUSES 235 are finite in number. Gelfond [Gel 601 had obtained some inequalities for linear combinations of logs of two algebraic numbers. . imbedded in the complex numbers.} be a basis of I.[IX. Thus the conjecture essentially asserts that the logarithms of algebraic numbers behave like “almost all numbers”. Then given E.&. . Feldman made a key improvement in Baker’s method. but Baker made a breakthrough when he extended such inequalities to arbitrarily many numbers. . . .*. The rational function cp then becomes a meromorphic function @ of a variable z E C. For qi E Z not all zero. Dirichlet’s box principle shows that the conjecture gives the best possible exponent. Besides those of Baker himself. Let I. 1411% + .. An explicit dependence can be given. To avoid more complicated exponents. map we parametrize 1 < h(p)(‘/Z)“(‘+‘)+E’ points A(C) by an exponential the complex exp: C -+ A(C) whose kernel is a lattice. who reconsidered linear forms of logarithms in light of more recent insights using certain techniques of algebraic geometry. Let m be the maximum of the multiplicities of the zeros on A. in C. . c’ depend on the heights of & a finitely generated subgroup of A(F). Let {PI. The same sort of arguments as above apply to elliptic curves or abelian varieties. . . . one has the Baker-Feldman inequality. Conjecture 7. > $ Here c. The full conjecture stated above is still unproved. and let Uj = log Pj meaning that 4 = exp(uj). the period lattice. [Wu 881. for which we refer to other expositions. Let cp be a non-constant rational function on A dejined over F. there is only a jinite number of points P E r satisfying the inequality Id’)1 Indeed.log&l Pl except for a jinite number of (ql.qr).mod torsion. from [Ba 663 to [Ba 741. + q. At this time. see especially Wustholz [Wu 851. . Let r be the rank of r. let us consider an elliptic curve A defined over a number field F.

then linear map at the origin is algebraic.. + . w2} be a basis for the period lattice. w2 are linearly independent over the reals. then the logarithms log P approach a point u0 = log(P. are there infinitely many real values of t such that p(tw) are algebraic and linearly independent over the integers? The extent to which r can be further lowered depends on the existence of such values of t. + .. Actually.. The exponential map is a complex analytic homomorphism exp: Lie(A). Given a non-zero period w. + qr+1u1 + a+2021 Thus our conjecture asserts that from the point of view of diophantine approximations. If a sequence of points P approaches a zero of cp. If we identify Lie(A)... with C”. with F contained in C. But by the quadraticity of the N&on-Tate height. + qrur + 4r+l@l + %+2@2 is close to uO.P. not necessarily in A(F)..+l. just as for the multiplicative group. Similarly. . the periods wr. the logs of algebraic points on the elliptic curve behave again like “almost all numbers”. and the fact that on a finite dimensional vector space all norms are equivalent. is a torsion point. This would seem to indicate that the exponent r + 1 should be replaced by r. 971 Let us assume for the moment for concreteness that our exponential map is given explicitly by the Weierstrass functions (p.236 WEIL FUNCTIONS.. p’). Hence there exist integers q. + Q where Q E A(F). qr+2 such that q1u.. considerations apply to an abelian variety A over F.Q). Let {wi.2 amounts to the inequality 1 < I+1+E. Of course PO may be algebraic. we have h(P) >x< q2 where q = max 1 qil. + q. Let P = qlP. We write P = exp u we normalize the exponential map so that the tangent and u = log P . + au.. INTEGRAL POINTS [IX. + . + A(C) whose kernel is the period lattice A. 4 Therefore the inequality (2) in Conjecture given among the elliptic logarithms by I-&J + q1u. 7. This also raises the following question.

albeit non-trivial ones. which was already conjectured in [La 741. The exponent of q on the right-hand side is determined by the probabilistic model and Dirichlet’s box principle. according to the probabilistic model and Dirichlet’s box principle. Even for elliptic curves. We obtain similar conjectured diophantine inequalities for linear combinations of logarithms of algebraic points of A(F). Only worse estimates are known. Taking the log. I have tried however to give references which will help the reader get acquainted with currently known methods. Let A be an abelian variety defined over a number field F. In this case. and supposing Z = D is a divisor represented by a rational function cp on a Zariski open set.EK 971 DIOPHANTINE APPROXIMATION ON TORUSES 237 for u E C”. The log here is sometimes called the abelian logarithm. which get constantly improved. Such an inequality is in line with the error terms found for the analytic Nevanlinna theory in Chapter VIII. Z) < ~ h(P) where c is an exponent reflecting the complexity of the singularities of Z and the rank of the Mordell-Weil group. and even in the case of complex multiplication. 55. Let A[m] denote the subgroup of points of order m in A(Q”). the analogue of the Baker-Feldman inequality for elliptic logarithms is not known. for elliptic curves. as of today. Faltings’ inequality in Theorem 3.(P. using a basis for the Mordell-Weil group. The best known result at this time (getting close) is due to Hirata-Kohno [Hir 901. 783. The first results (other than in the case of complex multiplication) were due to Serre. I would like to bring up one more possible application of methods of diophantine approximations to abelian varieties. The problem is to give a lower bound for the degree CWCml): Fl. which personally I find more satisfactory to describe the theoretical framework behind the mass of partial results available today in the direction of diophantine inequalities for the height. I have emphasized the conjectures.6 should be replaced by the inequality 1 dist. let G denote the Galois . When D has simple normal crossings. p. From this point of view. this inequality would also read (3) -log Icp(P) I < c log W'). but we shall not give an account of these partial results. then the constant c will reflect both the complexity of the singularities and the rank of the Mordell-Weil group.

w))lrl(“) .(A) make G an open subgroup of the product n Aut IT. Then we identify C” = Lie(A).. We let m be the degree of this polarization. The Riemann form has a unique normalized theta function (up to a constant factor) associated with it (cf. However. On the other hand. We return to the considerations of Chapter IV. 7. H(w. G.. Let A be an abelian variety defined over a number field F of degree d over the rationals. let B be an abelian subvariety of A. Example 4). We define its degree with respect to the Riemann form (or associated polarization) by the formula (deg. d) max(1.3. The main result linking Theorem IV and the methods used in the theory of diophantine is the following [Maw 913. Then the I-adic representations of G on 17. then essentially a normalization of the projective degree in the projective imbedding. the bound so far obtained falls very short of Serre’s theorem even when m is a prime power. Denoting Lie(A) the Lie algebra of A. We let H be a non-degenerate Riemann form (positive definite) on C” with respect to A. called the associated polarization.)c.(s) is the restriction of the Riemann If the polarization associated with H is very ample. this degree is corresponding 6. so that I could report on their results in this book. INTEGRAL POINTS CIX. Chapter IX.(A).1 of Chapter approximation Theorem subvariety Im HAcB)l. h&. Then deg. 5 c. In [La 751 I showed how techniques of diophantine approximation used in transcendence theory could also be applied to give some lower bound for the degrees of F(A[p”]) for a fixed prime p and n -+ co. n. n A and let G. 86 having to do with the proof of Masser-Wustholz’s theorem.(A).. we let Lie(A).238 WEIL FUNCTIONS.) over F. 971 group of F(A. Lie(B)c n A and H.(m. Let w E Lie(A). The results of Masser-Wustholz [Maw 911 are partly concerned with bounds for the size of Galois invariant subgroups of Ator. and the zeros of this theta function define a divisor on A whose class modulo algebraic equivalence is a polarization. be its extension to the complex numbers. and will now be discussed. be the smallest abelian dejined over F such that Cw c Lie(G. with its exponential map defined earlier in this section. I am much indebted to Masser and Wustholz for making available preliminary copies of their manuscript as well as for their guidance. B)’ = (dim B)!‘ldet where A(B) = form to A(B). $1..

this is a linear relation in abelian logarithms. We apply Theorem 7. . so that we may write p(cc)or as a linear combination with integers bj.(n) involving n!.a.4. n. We select a basis of periods {q i. To show the analogy I shall state the simple essentially analogous version on the multiplicative group. Let a 1. We select one suitably reduced period q of A in A.1.qzn) of B also suitably reduced. . be the smallest abelian subvariety defined over F as given by the theorem. When A is not simple. . In the case of an elliptic curve. . There exists a positive integer Y and a factor B of B’ such that if we let G. Theorem 1.1 to the abelian variety A x B2” and to the single period Let G. Suppose given an isogeny a: A-B. . Roughly speaking Baker’s method then yields another relation with suitably smaller coefficients bounded in terms of various 971 DIOPHANTINE APPROXIMATION ON TORUSES 239 where c1 and k.n(A x B’). Suppose that A is simple for simplicity. respectively. = G. Let R be the Z-submodule of Z” consisting of all relations among log a1. although these are by no means the best possible conjectured values. For instance. but a much lower value is expected to hold. we consider the complex representation p of CI. I = projection of CL on A x B.3 is used to imply Theorem 6. . then I is the graph of the isogeny we are looking for in Theorem 6. are constants depending only on (m. d) and n The proofs allow the writing down of explicit values for c1 and k. the proofs yield a value for k. .1 of Chapter IV. the situation is considerably more complicated. . In particular. we would pick z in a fundamental domain. . In transcendental terms. be algebraic numbers of degree 5 d. I shall now indicate briefly how Theorem 7.

. Then there exists a basis of R over Z such that for all elements (b 1. the kernel of 1 1. IP + QI 5 IPI + IQI We let r. In the present context. Theorem 7. one must actually consider the subspace R. h(. [Wa SO].) E Z” such that b.6. who first applied it to the group of units in a number field.. that is the subgroup of elements P such that IPI = 0. Then lmil”’ 5 6-l 1 141.(n.e.P. . for a suitable norm on this group. . and is simpler than Baker’s method.. . Chapter V. Q E r. The similarity of this statement with Theorem 7. To make the analogy exact... We assume that there exists 6 > 0 such that if IPI # 0 then IPI 2 6. Let PI. and also the context of Chapter III. except for the presence of the algebraic subgroup G. d) max (1. We shall deal with a finitely generated abelian group I. I shall give here very simple proofs for relevant statements which already show what is involved to get minimal norms. and use an algebraic subgroup of G” where G is the multiplicative group. there exists a relation CmiPi E r. i. . [La 781. . We let r be the rank of T/T. INTEGRAL POINTS CIX? 971 . The method stems from Stark.. log ~1~+ a.n. = 0. . that is..5. .b. The question arises whether Baker’s method is really necessary for the Masser-Wustholz result.written additively.b.240 WEIL FUNCTIONS. . [La 83a]. $1 among others. ... 1 ) satisfies: for all P E r.* + b. InPI = lnllpl for P.. and Masser [Mass 881. . . Instead of the height of basis elements one must use the height of this subspace in the Grassmanian.) log a. we want to find a basis of a finitely generated abelian group with the smallest norm. essentially as in Schmidt’s papers. j+i . 52 and $4. . generated by the relations over Q. with integers mi not all 0 satisfying Assume n > r. . be generators of r. See for instance [La 601. Chapter IV. all vectors (b. IPI L 0 for n E Z. 47. . we have the estimate lbjl 5 c.) in this basis. We let 1 1 be a seminorm on I-. Lemma 7. . $5.3 is evident.i))“-’ i for j = 1. and assume r 2 1. log a.

s. IP.P. + m. . .= 1.(. there exist integers 4. .+. (IPII. = 0 is free of rank 1 over Z. .. .(P.. . . By Minkowski’s theorem..P. . . . . 5 is strict for (See [Schm 801. so we write the relation in the form NP = m. . . we may assume without loss of generality that 1 1 is a norm on f.+. Consider say P = P. We define cj = i?/rlPjl for j = 1. .+. .. * + IPJ).sjl < cj for j = 1. . Let N = -m.) Let nj = qmj .P. . Then (PI.. + .P..P. Theorem j=l . .P. and the Z-module of linear relations (m. ... . Then co .. . .P. say (Ply . inequality so that lnjl < Ncj Let Q = qP .. ..(. We are going to show that INI 5 co This will yield Lemma where co = (r/i?)‘IP.+. .m. 7.) has rank r. . + n.s..r 141 co.> is a maximal family of linearly independent elements among Pi . I use a variation from Waldschmidt [Wa 801.5 by using the inequality .r.I)“’ 5 $1 + . c.. .) = 1. ...m.. .) such that m. Then NQ = nIPI + . ... . . .. . After renumbering the generators. whence .P.CK §71 DIOPHANTINE APPROXIMATION ON TORUSES 241 Proof. Rather than use Dirichlet’s box principle as in [La 781 and [La 83a]. .sIPI .+.I.+l..Nsj 2C. p.+. + m. si.P.. + . such that and The Iqmj/N . . . . 33. . generated by a vector such that (m. Passing to the factor group F/T.

P:} of T/T.. Let nj. . and let R be the Z-module of relations M = (m. E r be linearly independent there exists a basis {Pi.P. then a point P in I is a non-zero element ~1. On an abelian variety over a number field. Then the elements Pi. but leads to I6-’ C j#i Ipil. . = 0..sI) is a non-zero multiple of (N. Then Proof.’ form the desired basis. . + . such that mod r. = (PI. . i ..j-1 satisfy with some Pj’ E r. Let I. . .k 5 N . Theorem a slightly less elegant inequality [lmil”‘] box principle as in [La 781 (corrected in 7. If I is a finitely generated multiplicative group of elements in a number field K. and qP = si P.5. INTEGRAL POINTS CIX. The most important for us are the multiplicative group and abelian varieties.m. Remark. Let PI. The standard arguments of algebra give the desired estimated. jpj = Nq. . I PI + *. .6) is a simpler technique. nj. . I reproduce these arguments.) and IN( j (q( 5 co. . P. + nj..l. . m.) mod I. .. . .. .(P)“” gen- where /i is the N&on-Tate height associated with an even ample divisor class.. . . . . .. .) as above. . P.m.242 WEIL FUNCTIONS. $71 Hence the (r + 1)-tuple Hence Q = 0. . . (49 s This proves Lemma 7. Let N be the index (I : I. .6. Then NT has finite index in I’..). .).’ Without loss of generality we may assume 0 5 nj. we let I be a finitely erated group of algebraic points. .and we define as in [La 641 IPI = h(cc) where h is the absolute height. .. . Without loss of generality we can assume r..j be the smallest positive integer such that there exist integers nj. Next we have an example having to do with estimating relations.P. and we let IPI = t.1. Chapter 5. . Examples.nj.. . by working in the factor group T/T. . + sJ’. . . Lemma 7. . Let r = (Pi. Using the Dirichlet [La 83a]. .

Theorem . For elliptic curves and abelian varieties see [Mass 841. . we have 44 1 c/CQ(d : 41. .P. that is. . By Proposition 7. Lehmer’s conjecture.5. 7. Taking into account the weaker results proved in this direction.P. Let h denote the absolute height on Q”. in the direction of a conjecture of Lehmer.6. for each j = r + 1.7. We put with mi E Z.4 is a special case of Proposition 7.. . n-tuples of integers satisfying ~m. denoted by (IMIJ = max Imil. Mj The proof is concluded by applying Lemma 7. There exists a constant c > 0 such that for all algebraic numbers o! not equal to 0 or to a root of unity.n there exists a relation between Pr. . .CIXY 971 DIOPHANTINE APPROXIMATION ON TORUSES 243 E I. and for the multiplicative group see the best known result in Dobrowolski [Do 793. .. and one needs minimal heights for the non-torsion points on abelian varieties over number fields. Proof. the sup norm on R. In the applications one needs minima for the heights of algebraic numbers of bounded degree not equal to roots of unity. There exists a Z-basis for R such that all elements M in the basis satisfy the bound IIMII 5 t-B* where B = 6-l i i=l /PiI. q satisfying the bound 11 II I” 5 B. Proposition 7. .7.

To me at the moment. with the problem of determining the structure of the variety under this opposite set of conditions.or co-canonical class) is ample. Sp 3 describing the special set of Chapter I. then one expects the polynomial equation to have non-trivial solutions in various contexts. we have dealt with cases when the main idea was to show the existence of few rational points. Other ways may be given via a zeta function. g3 is the whole variety. 53. when ordered by ascending height. but also the determination of conditions under which the Hasse principle . To some extent the present chapter may be viewed as dealing with the case when the special set defined in Chapter I. and one expects lots of rational points if there is one. when minus the canonical class (also called the anti.e. Then one can propose ways of counting them asymptotically. There is an opposite situation. i. very canonical and pseudo canonical can now be added the prefix anti. Conditions for the existence of a global rational point involve not only a global hypothesis such as the ampleness of the anti-canonical class. Artin’s conjectures (and those cases when they have been proved) remain to be looked at in this context. the rational image of a projective space or is generically fibered by linear group varieties as in Sp 2. Roughly speaking this situation prevailed when the canonical class is ample. one striking aspect of this direction is that it connects with an older idea of Artin that when the number of variables n is greater than the degree d (say of a homogeneous polynomial). The n > d condition appears today as precisely the condition which insures an ample co-canonical class. or conjecturally when a variety is not pseudo canonical. One possibility is to determine the extent to which a variety is unirational. How many solutions may be measured in various ways.CHAPTER X Existence of (Many) Rational Points In most of the book. To the three possibilities canonical.

The situation with Ci fields provided to my knowledge the first example whereby diophantine problems over a finite extension are reduced to a lower field by what became later known as restriction of scalars. We shall deal with some cases. except for those pertaining to linear algebraic groups.O) = 0. it has also been proved to be strongly Ci..t.. norm form be N(t 1.t. then there exists a global (non-singular) rational point. ... FORMS IN MANY VARIABLES Artin called a field F quasi-algebraically closed if every form (homogeneous polynomial) of degree d in n variables with n > d and coefficients in F has a non-trivial zero in F. A field satisfying the analogous properties for such polynomials will be called strongly quasi-algebraically closed. . In practice. . if we pick a basis {cli. [La 511. Indeed. let the . strongly Ci.’ + t&J.. . Such polynomials f always have the trivial zero f(0. supposing F’ separable over F for simplicity. is the same as quasi-algebraically closed. $1. . then a system of r polynomial equations in IZ unknowns over F’ is equivalent to a system of r[F’ : F] equations in n[F’ : F] unknowns over F itself. one may also consider polynomials without constant terms. X. in F. Aside from some reasons given at the end of $2. . + . .) = n(t& d + . a..+i. it seems to me that this direction would fit better in an encyclopedia volume dealing wholly with linear algebraic groups. Cf. . I am much indebted to Manin and Colliot-Thtlene for valuable suggestions concerning this chapter. Also in practice if F is Ci then the rational function field F(t) in one variable is C. Furthermore. .} for a finite extension F’ of F. . but results in this direction still appear to me fragmentary. . . + tmam with variables t.LX 911 FORMS IN MANY VARIABLES 245 holds: if there exist (non-singular) rational points locally in every completion. Instead of considering homogeneous polynomials. . whenever a field has been proved to be Ci. . Thus C. For various reasons I have chosen to end the book before going fully into that direction. resp. More generally. by writing a variable T over F’ as a linear combination T = tla. F is said to be Ci if every form of degree d in n variables with n > d’ has a non-trivial zero in F. If a field is Ci then every finite extension is Ci.

..“‘. Let t(l) = (t$“. then one applies the C. . In light of Wedderburn’s theorem that finite fields admit no non-trivial division algebras over them. we have the implication: If F is C.) is a form in m variables. . Tsen proved that a function field F in one variable over an algebraically closed constant field has no non-trivial division algebra of finite dimension over F. then H2(G. Let f(tl.] having only the trivial zero mod rc. a# 0 to show that every element of F is a norm from E. . Fa*) = 0 but not the converse. . .t.246 EXISTENCE OF (MANY) RATIONAL POINTS cx 011 where the product is taken over all conjugates cr of F’ over F. . There was no reason to believe the converse. and Ax found an example of a field F such that every finite extension of F is = (t’. The diophantine property of being C.. both globally and locally. with coefficients in F. t.t. . Let by adjoining certain Locally. which in Artin’s terminology states that such a field is quasi-algebraically closed.) be a form of degree m in m variables in RCt i. . . Artin noted that his method of proof implied something stronger.) or a power series over a finite constant field of characteristic p. with a E F. . Suppose F has a discrete valuation. So from this point of view. In the thirties. . [Ax 661. . with ring of integers R and prime element n. .. . and having only the trivial zero in F.Q’) be m independent sets of m independent variables. implies the non-existence of division algebras as above: if E is a finite extension of F and N(t. . whence the nonexistence of the division algebras.tgq. Then N(t i . the conditions d = n and d2 = n form natural boundaries for the property that if the number of variables is sufficiently large compared to the degree then there exists a non-trivial zero. Then g is a form of degree m in m2 variables with coefficients in F and having only the trivial zero. . In cohomological terms.“+. let k be a field in one variable F be the maximal . . . Let g(P).. This was proved by Chevalley [Che 351. Artin also conjectured that certain fields obtained roots of unity are C. property to the equation W 1. of degree m. . .) t’“‘) = f(P)) + 7cjp2’) + * * * + 7F1f(P).) is the norm form.t. Artin conjectured that a finite field is C. ... . . . . . A) = at. . . . . but F is not C. . p-adic field (finite extension of Q.

I proved the case of power series in [La 511.. Then I proved Artin’s conjecture that F is quasi-algebraically closed. but the conjecture for p-adic fields turned out to be false by examples of Terjanian [Ter 661. AxKochen proved [AxK 651: Theorem 1. then F is C. (Note that a more general conjecture over arbitrary fields has been shown to be false by Merkuriev. property holds.. [Ter 771.. and what is the dimension of the largest unirational subvariety. Looking back with today’s perspective. I realize that it is not known whether the algebraic set defined over such fields as in Theorem 1. This is proved for forms of prime degree. Looking at these old questions from this point of view would give them new life. the field of formal power series over an algebraically closed constant field is C. One may characterize F as the field obtained by adjoining to k all n-th roots of unity with n prime to 911 FORMS IN MANY VARIABLES 247 unramified extension of k. Let F be a field complete under a discrete valuation perfect residue class field. . . More generally [La 511: Theorem 1. (c) In particular. Then HeathBrown got it down to 10 for non-singular forms [H-B 831.] of degree d in n variables with n > d2 has a non-trivial zero in Q. .. conjecturally) by a homogeneous polynomial with n > d contains a rational curve. It is still expected that a cubic form in 10 variables over Q has a non-trivial zero in Q..2. For each positive integer d there exists a finite set of primes S(d). A special case of a conjecture of Kato-Kuzumaki [KatK 863 states that a modification of the C. Demjanov and Lewis proved that a cubic form in 10 variables over a p-adic field has a non-trivial zero [Lew 521. Artin conjectured that a p-adic field and the power series in one variable over a finite field are C. . On the other hand. T. asserting the existence of a O-cycle of degree 1 rational over a p-adic field if n > d2. and got the number of variables down to 16 [Dav 631. (b) If the residue class field is algebraically closed. The field of convergent power series over C (say) is C. Then: with (a) The maximal unramified extension of F is C. Global results are still fragmentary.1. and Hooley got it down to 9 if the form is non-singular and has a non-trivial zero in every p-adic field (see Hasse’s principle below) [Ho 881..1 (or global fields which are C. Davenport worked on the problem of the existence of non-trivial zeros of cubic forms over the rationals. ..) When d = 3. such that if p $ S(d) then every polynomial f E Q.[ Tt .

YTotally imaginary” means F has no imbedding in the real numbers. In addition. where one assumes the existence of a simple rational point in each completion F. Experience (starting with Birch [Bir 611) shows that it may be more useful to deal with the non-singular Hasse principle. and Patterson [Pat 851. if X has a rational point in every completion of F. a form of degree d in n variables has a non-trivial zero in F if n is sujiciently large with respect to d.. See Conjecture 4.4. The method also gives an estimate on the number of rational points of bounded height. The problem is to determine which X satisfy the Hasse principle. the non-singular Hasse principle of [Bir 611 worked with fewer variables than in [Bir 571. Over a totally imaginary number field F. If n is sujiciently large with respect to d and with respect to the dimension of the set of singular points. . what are the obstructions for its satisfaction. These theorems establish some special cases of Hasse’s principle. Peck [Pe 491 proved: Theorem 1. Let f be a polynomial of degree d in n variables with zero constant term and n > d over a number field F. Then f has a non-trivial zero in all but a finite number of v-adic completions F. Let F be a number field and let X be the hypersurface defined over F by a form of degree d in n variables. and Vaughan [Vau 811. and contains the previous results of Peck and himself as mentioned above. Thus Birch’s method applies for forms of even or odd degree.3.3. or an algebraic set X: If X(F. then the non-singular Hasse principle holds for X.) is not empty for each absolute value v on F. which states for a variety. then X(F) is not empty. see Davenport [Da 621.] For forms of odd degree. the circle method applies to complete intersections. over number fields. and if it is not satisfied. Greenleaf [Grlf 651 proved my conjecture: Theorem 1. In other words. then X has a rational point in F.5.. Roughly speaking. Schmidt [Schm 841. which is important in the context of $4. The above mentioned authors use refinements and variations of the Hardy-Littlewood circle method. Birch [Bir 611 showed Theorem 1.248 EXISTENCE OF(MANY) RATIONAL POINTS lx §ll For any degree. Danset [Dan 851. Adelic versions are given at a more sophisticated level by Lachaud [Lac 821. and one then concludes the existence of a simple global rational point in F. Birch eliminated the restriction that the field be totally imaginary [Bir 573. For an exposition of the circle method.

Up to technical details.g. Colliot-Thelene.6. the case of curves. For the proof of the remark. no hypothesis of geometric irreducibility is made. point in all Greenleaf reduced Theorem 1. namely: 871). e. Indeed. A variety over a number jield has a rational but a jinite number of completions. whence a p-adic point by variants of Hensel’s lemma. that is. The role of such irreducibility and of nonsingularity is not completely cleared up. and this rational Remark 1. The property point is a birational place induces a point on every other complete variety birationally alent to X over k. for an arbitrary number field k.Lx 911 FORMS IN MANY VARIABLES 249 Note that in all results of the above type. for a complete non-singular variety to have a invariant (cf. Theorem 1.8 is the analogue of Meyer’s classical result over Q that a quadratic form in 5 variables which has a non-trivial zero in all real completions has a global non-trivial zero (Hasse over number fields). For instance. again by using a refinement method. y) such that D&x. let X be a complete variety over a field k with a simple point P E X(k). that is. Then by Weil’s Riemann hypothesis in function fields. Let k be a totally imaginary number jield. cutting the variety with sufficiently general hyperplane sections we are reduced to the case of dimension 1. the curve has a nonsingular point mod p for all but a finite number of primes p.8 ([CTSSD . see also [CTSal 891. a variety is geometrically irreducible.5 to the above remark. Newton’s approximation method. having the same function field. I had already observed: Remark 1. y) # 0. Over a complete field.7. Sansuc and Swinnerton-Dyer proved a sharp result along the statements of this section. Then there exists a k-valued place of the function field k(X) lying above P. the existence of one simple rational point implies the existence of a whole neighborhood in the topology defined by the field. when I conjectured Greenleaf’s theorem [La 60b]. If for instance the variety is defined by one equation with a rational point (x. Then two quadratic forms over k with at least 9 variables have a non-trivial common zero over k. [La 541). The result of Theorem 1. there exists a value 7 close to y such that f($ 7) = 0. equiv- From a totally different method. the obstructions arising from the real places are the only ones. then for all values X close to x. Remember that for us. We shall go deeper into this phenomenon in $3.

and Artin-Tate [AT 683 for global facts. In this section I shall go into greater detail into the Brauer group and the Manin obstruction. We shall use the cohomology of groups.(F”)). X.. is the multiplicative group. as well as bibliographies. Then X. One of Kanevsky’s results is as follows. has a rational point in an abelian extension of F. For nonsingular cubic surfaces. is the multiplicative . Let F be a field of characteristic 0. I recommend Shatz [Shat 721 as a reference for proofs of basic or local facts. when applied to rational varieties leads to the above mentioned obstruction. Finally we mention Artin’s global conjecture that the field Q(r) obtained by adjoining all roots of unity to the rationals is Cr. and more generally non-singular projective rational varieties. 52. and to the extensive survey [CTKS 851. in particular over p-adic fields and number fields. THE AND BRAUER MANIN’S GROUP OF A VARIETY OBSTRUCTION All known counterexamples to Hasse’s principle for a variety X are accounted for by an obstruction defined by Manin [Man 701.250 EXISTENCE OF(MANY) RATIONAL POINTS PC @I Igusa has introduced a zeta function in connection with the problem of finding a zero for a form with many variables [Ig 781. the Brauer group Br(F) is WF) = H’(G. and G. the so-called BrauerGrothendieck group. Theorem 1. I refer to the appendix of Manin’s book [Ma 743. be a non-singular cubic surface over a number field F. Kanevsky [Kan 851 establishes a connection with the question whether the Manin obstruction to the Hasse principle is the only one (see his Theorem 3). for instance for cubic forms. where G. Second Edition 1986. G. For many results obtained in this direction and similar ones.9. is the only one. both for explaining theorems and for indicating the literature. to the survey by Colliot-Thtlene [C-T 863. We shall discuss the Manin obstruction in $2. to the survey by Manin-Tsfasman [MaT 861. but it has not so far led to the sharp conjectured results. By definition. Assume that for every jinite extension E of F the Manin obstruction to the Hasse principle for X. I am much indebted to Colliot-Thelene for his guidance in writing this section. Let X. gives a general obstruction which: when applied to abelian varieties leads to the Shafarevich-Tate group.(F”) group of the algebraic closure of F. where he shows that a generalization of the Brauer group.

“/F:‘)..(F. on the cohomology + H2(G”‘. one of them being Theorem 1. which is just the group of characters of G”’ with values in Q/Z. the order at the valuation gives a homomorphism ord. I = inertia group = Gal(F.Y)) From the short exact sequence 0 + Z + Q -+ Q/Z + 0 one gets a natural isomorphism (the inverse of the coboundary) H2(G”‘.. is complete under a discrete valuation be the maximal unramified extension of F. Q/Z).‘)) exact sequence of Galois cohomology inf H2(GF.(F. Thus the inflation gives an isomorphism. In addition. Z). H2(Z. G. G. Q/Z). we obtain a homomorphism which we denote by 8.( Hence we can view (2) also as a homomorphism of H’(G”‘..1 (a). : G. .): H2(G”‘.‘)) into EZl(G”‘. Let: unramified Thus we have a tower of fields: The inflation-restriction (1) 0 --f H’(G”‘. H2(ord. 921 Local remarks THE BRAUER GROUP OF A VARIETY 251 Suppose first that F = F. G. F/ u.“)) The term 0 on the right has various justifications.“)) --=-+ reads: = 0.) + Hl(G”‘.(F.(F”“‘) whence an induced homomorphism (2) + Z. Z) 5 Hl(G”‘. G. Composing the inverse isomorphism of the inflation-restriction sequence in (1) and the homomorphism of (2).(F.) be the Galois group of this maximal extension. Q/Z). G. G”’ = Gal(F//F.: Br(F.

We say that b is unramified if its image is unramified for all u. We define the Brauer group of X to be a certain subgroup of Br(F) as follows.252 EXISTENCE OF(MANY) RATIONAL POINTS lx §21 We define an element b E Br(F. The global case if d. inv. we can apply the local remarks to the completion F. In addition. For each discrete valuation u of k(X) given by a point of X of codimension 1. each rational point on X allows us to specialize an element of the unramified Brauer group of the function field to an element of the Brauer group of the constant field k. We recall some facts about Brauer groups over k.. in other Next let X be a variety. and maps Br(k.) to be unramified words the image of this element under 8. We let F = k(X) be its function field. for each rational point P E X(k) we have a specialization homomorphism Br(X) + Br(k) denoted by b H b(P). which is an isomorphism for v finite. We use a fact essentially from local class field theory that there is a natural injection. and call it the unramified Brauer group or Brauer-Grothendieck group. = C. We denote this group by Br”‘(F). If b E Br(F) then the image of b in Br(F. If b.) is unramified for all but a finite number of u. For X projective non-singular. then Br(X) and Br(Y) are the same subgroup of Br(F). Let X be a projective non-singular variety over a number field k. see Grothendieck [Grot 681. the group Br(X) is a birational invariant. We are now ready to define the Manin obstruction to the Hasse principle. we have a natural homomorphism Br(F) + Br(F. and defined over a field of characteristic 0. or equivalently a subvariety of codimension 1.b = 0. E Br(k). trivial if k. Corresponding to the inclusion F c F.). l/2} if u is real. is 0. then a classical theorem of Albert- .) on (0. For proofs of the above two properties.) 4 Q/Z. In other words.: Br(k. in the sense that if Y is another projective non-singular variety over k with the same function field F. The unramified elements of Br(F) form a subgroup which we denote by Br(X). the invariant for each absolute value u on k. non-singular in codimension 1. and which we call the Brauer group of X.

24). which has non-singular rational points in Q. then there exists a global rational point P E X(k). 941. Let X be a curve of genus 1 over a number field k. Then the Manin obstruction for X to the Hasse principle is the only one. then for all b E Br(X) we have T inv.) there exists b E Br(X) such that . Actually. The existence of elements b E Br(X) having the above property will be called the Manin obstruction to the Hasse principle.) is not empty for all v.1 (Manin [Man 743.x2)(x2 . ” Theorem 2. Theorem 41. consequence of Chapter VI. for all v. In particular: Assume that X(k. see [CTCS 803.) Z 0. 228. For proofs fitting the present context. Examples of Manin obstructions are given in Iskovskih [Isk 713. for instance the surface defined by the affine equation y2 + z2 = (3 . . but is easier for varieties which are birationally equivalent to projective space over the algebraic closure k”.) ” = 021 Brauer-Hasse-Noether THE BRAUER GROUP OF A VARIETY 253 states that C inv. b(P) = 0.(b. inv. If there exists a rational point P E X(k). the existence of the Manin obstruction for curves of genus 1 is hard to verify.W. Zf for all elements {PO} in n”X(k. b(P. in other words. but such that no projective desingularization of this surface has a rational point in Q. then X(k) is empty.2). We shall say that there is no Manin obstruction to the Hasse principle for X if there exists a family {Po> E n X(k) ” such that for all b E Br(X) we have 1 inv.) = 0. p. if there is no Manin obstruction. see [Shat 721 and [AT 681. For generalizations of this example and further discussions. Assume that LII(J(X)) is finite.

. one finds in [CKS 851 an example. Saito proves: Theorem 2. [CKS 851 gives an algorithm such that if Manin’s obstruction to the Hasse principle is the only one. Colliot-Thtlkne has proposed a more likely conjecture.254 EXISTENCE 0~ (MANY) RATIONAL POINTS lx 621 In addition. Indeed. be a complete non-singular curve over a number field F.. Assume that Br(X) is jinite. if that were the case. of a diagonal surface which has a p-adic point for all p but no global rational point. then for all non-singular complete intersections in P: of dimension >= 3. Note that the hypothesis on the finiteness of the Brauer group is essentially equivalent to the finiteness of the Shafarevich-Tate group of the . it is unlikely that the Manin obstruction is the only one for all projective non-singular varieties X. one can define the Brauer group Br(X) exactly as we did for a non-singular variety.. has a O-cycle of degree 1 rational over each completion F.. over a number field k. The constant field never played an essential role. Br(X)/Br(k) = 0. There have been some results of Saito [Sai 891 concerning curves. however. An actual example is lacking at this time.2. rational over a number Jield k. For such a scheme.Xq) = 0 of any degree. due to Cassels-Guy. and that the Manin obstruction for a O-cycle vanishes. and let X over spec(o. namely: 5x3 + 9y3 + 1oz3 + 12t3 = 0. Then there exists a O-cycle on X. and thus has a tendency not to have rational points at all. As Colliot-Th&ne has pointed out to me..) be a regular proper model of X. the Hasse principle would be true. rational over F and of degree 1. would then This is not expected. since for such X. then one can decide whether there is a rational point on a diagonal cubic surface over Q. because for instance the Hasse principle hold for a non-singular hypersurface f(x o. Thus X is a regular 2-dimensional scheme. namely: The Manin obstruction formulated for O-cycles of degree 1. Let X. of F. is the only obstruction to the existence of a global rational O-cycle of degree 1. But for high degree a general such hypersurface is conjectured to be hyperbolic. whereas the presence of local rational points is not expected to be so rare. that X.

fibered by tonics over the projective line. To shorten the notation we have written Pit X” instead of (Pit Xa)(ka). By methods of algebraic K-theory. G. as in the N&-on model of an abelian variety. A bibliography of several basic papers of Kneser and Harder is given by Sansuc [San 811. (An exceptional case need not be mentioned because of a more recent result of Chernoussov. 021 THE BRAUER GROUP OF A VARIETY 255 Jacobian of X. although on occasions we have seen that noncomplete varieties play an essential role. Hence modulo a constant part which is irrelevant for the Manin obstruction. the Manin obstruction for the existence of a O-cycle of degree 1 is the only one. we see that Ker(Br(X) -+ Br(Xa)) is the same as H’(G. 466): 0 + Pit(X) Ker(Br(X) -+ Pic(X”)‘k + Br(k) -+ + Br(Xa)) + Hi(G. the Brauer group.[Ix. then by a theorem of Tate ([AT 681. who develops these theories systematically. If k is a number field. To a large (if not exclusive) extent. then one gets a theorem of Manin (see Theorem 2. The theory of III (the Shafarevich-Tate group). Theorem 14) we have H3(Gk. and when the curve has genus 1. including the theory of the Manin obstruction.. It would be desirable to have a book giving systematically the general properties of the Brauer group. There is another interpretation of the Manin obstruction. Again we take X projective non-singular over k. I conclude this section by emphasizing the existence of this theory. his Corollary 8. For a discussion of this interpretation see [Man 70b] and [Man 741.(k”)) = 0. Pit X”). the theory of such O-cycles has developed in various directions. I cite one of his results. so X” is the extension of X to the algebraic closure of k.. I do not expand the book to include the diophantine theory of linear algebraic groups or group varieties in general.. However.) . Then there is a long exact sequence (see [CTS 871 p. extend to linear group varieties. For convenience we write Xa for the base change X. and the Manin obstruction can be interpreted in terms of the Picard group. initiated by Spencer Bloch. this book has been concerned with complete varieties.(ka)). Thus in recent years. the Manin obstruction.1)... Pit X”) + H3(G. Salberger [Sal 881 showed that for a non-singular surface (projective variety of dimension 2) over a number field. See for instance [Kn 691. as well as its applications to unirational varieties. Chapter 7. That the Hasse principle is valid for principal homogeneous spaces of semisimple simply connected linear group varieties is due principally to Kneser and Harder. Given the size of the book and mostly my incompetence.. G.7.

using some commutative diagrams. Then X satisJies Hasse’s principle.3.8 fits among them. Colliot-Thtlene and Sansuc’s idea is that one may replace the use of isogenies (which are principal homogeneous spaces under finite abelian groups) by the use of principal homogeneous spaces under tori. An example exhibiting non-trivial diophantine properties from the present point of view is the Cltelet surface V = V. The first examples date back to Chltelet. a Severi-Brauer variety over a field k is a variety which becomes isomorphic to a projective space over a finite extension of k. Then the Manin obstruction for and for X is the only obstruction to the Hasse principle for V and respectively. If X is a Severi-Brauer variety over a finite field. a V X Theorem 2. the terminology would have been more appropriate to call the varieties in question Severi-Chltelet. By definition. and let X be non-singular completion of V over F.4 (Chhtelet). the proof for V goes through the proof for the completion X. and in particular: Theorem 2. I refer to [C-T 883.256 EXISTENCE OF(MANY) RATIONAL POINTS ix VI F. The descent method of Chapter III. Actually. in light of Chdtelet’s basic results. For a survey of Chbtelet’s contributions in this direction. .. then X is isomorphic to projective space over this field. Example (The Chatelet surface). As Colliot-Thtlene pointed out.az2 = P(x). In his thesis [Chat 441. and many special cases have been studied. This method can then be applied to reduce the study of rational points on a variety over F which becomes rational over F” to the study of rational points on auxiliary varieties for which the Manin obstruction vanishes. Theorem 1. For such varieties. Let X be a Severi-Brauer variety over a number field. and become isomorphic to projective space. defined over a field k by the affine equation y2 . The theory of homogeneous spaces for linear groups and the Galois cohomology with coefficients in the (linear) group of automorphisms of a variety are also used in an essential way to study the problems which arise concerning projective non-singular varieties which do not necessarily have a rational point. A general theory was developed by Colliot-Thelene and Sansuc [CTS 873. or are k-birationally equivalent to principal homogeneous spaces under linear algebraic group varieties. 94 extends to other varieties.. Let G be a linear group variety over a number field Let V be a principal homogeneous F-space of G. Chltelet generalized to such varieties what was known before on tonics. one raises the question whether they satisfy Hasse’s principle.

For degree P = 3 or 4.. In [CTSSD 871 the reader will find a simple description of a projective non-singular completion of V. projecting on the x-line. say. 3. Supposek is a number field. is an affine non-singular surface. today when the degree of P is at least 8. and when the Chatelet surface over Q has solutions in Q. 3. a # 0..10 and 3.. is pseudo very ample. when P is an irreducible polynomial of any degree. and the image of the birational imbedding into P4 can be described explicitly... then V.lx 921 THE BRAUER GROUP OF A VARIETY 257 with a E k. Colliot-Thelene and Sansuc [CTS 823 have shown that the non-singular Hasse principle also holds for X. whose generic fiber is a curve of genus 0. and P E k[x] is a polynomial without multiple roots. a result of Iskovskih can be formulated as saying that -K. positive leading coeficient. we now have the following possibility: . This conjecture is a generalization of the conjecture that there are infinitely many primes of the form n2 + 1.. When k = Q. and X. and the Hasse principle is true for V. containing in particular the next two theorems. cf.9.. and such that P(Z) has no common prime factor.11. In [CTSSD 871 one will find a theory of the Chltelet surface when P has degree 5 4.. represents infinitely many primes. If there exists a k-rational point on V... Theorem 2. Zf P is irreducible over k then Br(X)/Br(k) = 0.. provided a conjecture of Schinzel is true: an irreducible polynomial with relatively prime integer coeficients. is k-t&rational. The above fact is complemented result. However. For simplicity we assume that k has characteristic 0. See also [C-T 861 and [San 851 for a more extensive survey of results in this direction and a more extensive bibliography. Thus V.6.. [Izk 721. Then: (a) (b) The Manin obstruction to the Hasse principle for X0. In line with Chapter I.. one does not know if there exist such solutions for infinitely many rational values of x.5. over number fields by the following Theorem 2. which we denote There is a morphism by Xo. is the only one..

but the weaker hypothesis would. the hypothesis that -K. Let X be a projective non-singular variety over a number field k. X. Then these close solution with Xi E RO. pseudo ample. including the two cases mentioned above [La 511. be the ring of convergent power series inside the ring of formal power series.) The question also arises to what extent one would need -K. . Let R.1 concerning elliptic curves. which are local rings such that R is the formal completion of R. 3. $3. I proved this result in the case when the rings arise from absolute values in the context of complete fields. be the maximal unramified extension of k.. c R be a subring of an integral ring.. .. . Let F.(Tl.[TJ in one variable and coefficients in R. Let f. namely: . .) = 0 be a jinite family pose this family equations have a to xi in the local (j = 1.) with xi E R. . I am lacking examples or counterexamples to make a coherent general conjecture. Then F. We say that RO is relatively algebraically closed in R if given a non-zero polynomial P(T) E R.T. which is sufficient or necessary so that the Hasse principle holds? How far back must one go to get such a condition.. .is a root of P in R then in fact CtE R. is pseudo ample.r) of polynomial equations with coefticients in R. We can characterize Fo also as the field obtained by adjoining to k all n-th roots of unity with n prime to p.11. R)... Examples of such a pair first arose as follows. Let k be a p-adic field or a power series field in one variable over a finite constant field of characteristic p. Local specialization principle. following Colliot-Thdlhe’s conjecture. Suphas a solution (x1. is relatively algebraically closed in F.. if ~1.7. Let F be the completion of FO. is pseudo ample would not cover the case of Manin’s Theorem 2.x. in other words is there a natural condition on K. . Is the Manin obstruction the only obstruction to the Hasse principle? (Perhaps only for cycles of degree 1. Then R. is the hypothesis that -K. contains some effective divisor sufficient? Because of Chapter I. .g. is relatively algebraically closed in R. e..258 Exwr~~cE OF(MANY) RATIONAL POINTS cx 031 2. I formulated the following principle for such a pair of rings (R. such that Xi lies arbitrarily ring topology of R.. LOCAL SPECIALIZATION PRINCIPLE Let R. Assume that -K. .

X.. 93. [Grbg 631. §41 ANTI-CANONICAL VARIETIES AND RATIONAL POINTS 259 Let k be a jield complete under an absolute value.. whereby a polynomial equation over the power series amounts to infinitely many equations among the infinitely many coefficients of those power series. b] and van den Dries [vanD 811. to the extent it exists. thus giving rise to the Greenberg functor [Grbg 613. Robba [Rob Xla. Other cases of the specialization principle have been proved by Bosch [Bos 811.) = 0. to specialize the solution from R to R. . . or only a finite . which is relatively algebraically closed in k. be a dense subfield.d) where d = cdj. Thus -K. obtained by truncating the equations modulo a power of the maximal . . r. I believe for the first time in the context of such diophantine problems. Such a refined version for finding a zero from an approximate zero was needed because a solution of a system of equations might be singular modulo high powers of the maximal ideal. I used the Newton approximation method. or equivalently. j=l ? . Contrary to the expectation of few. Greenberg. developed long before people became conscious of the interpretation of the inequality n > d in terms of the canonical sheaf. Let k. Then the above formulated specialization principle applies to the pair k. A finite system of polynomial equations over a complete local ring amounts to an infinite system of equations in the residue class field (under mild conditions on the local ring). In particular. where instead of the local ring topology. The theory of quasi-algebraic closure.. and the usual version of Hensel’s lemma could not be applied. we use the topology dejined by the absolute value. ANTI-CANONICAL RATIONAL POINTS VARIETIES AND As noted already in Chapter I. then the anti-canonical sheaf is 6(n . $4. In working with fields rather than the rings. and such that k is separable over k. if W is an algebraic set in Pnel defined as a complete intersection by a system of homogeneous equations &CT. in k. or in a one-variable situation.T. The typical case is that of power series. of degree dj. This procedure which arose first in [La 511 was schematized and functorized by M. . My conjecture [La 541 that the result would also hold for power series in several variables was proved by M. is ample if and only if n > d. amounts to a projective system of equations in these coefficients. Artin [Art 681. I proved the specialization principle for convergent power series in one variable inside the formal power series. .

Lower bounds in certain cases were obtained by Schmidt [Schm 851 by the circle method. extending and similar to Schanuel’s theorem for projective space. Let X be a projective non-singular variety over a number field F. For such a variety one expects an abundance of rational points if there is at least one.260 EXISTENCE OF (MANY) RATIONAL POINTS cx 041 number of rational points when the canonical class is ample. I know only very few results in this direction. when minus the canonical class is ample. Batyrev.(P) 5 log B. having to do with generalized flag manifolds. be one of the height functions (defined mod O(1)) associated with c. B) = number of points P E X(F) such that h.loglls(P)lI. Batyrev-Manin make some conjectures for which we need various defini- . Let H. Manin and Tschinkel [BaM 901 and [FrMT 893 have also considered a zeta function in the context of counting points. Let -$? be a line sheaf corresponding to the class c. for a~ F.. and let h. which would thereby bring in the whole machinery of zeta functions of automorphic forms into diophantine analysis via this route. aside from Theorem 12. c. Franke. Such a variety is also called a Fano variety. be the exponential height. Following Arakelov [Ara 74b] and Faltings [Fa 84~1. and even possibly the unirationality of the algebraic set. = exp h. (These are the usual conditions. In general.(P)= 1 . Let X be a projective non-singular variety. H. We say that X is anticanonical if -K. One expects certain asymptotic estimates for the number of rational points of bounded height. and we shall discuss these varieties. Define: N(X. The zeta function is defined by Of course one must normalize the height. one expects either none or many rational points. is ample. We mostly rely on some conjectures to give an idea of what may go on. where 11 [Iv is a norm on 3” su%h that llasllv = IlallUllsI/.) In one case. Let c E Pic(X.. Let s be a rational section defined and non-zero at a given rational point P. and II 1 JIV= 1 for almost all o. Then h. Let U be a Zariski open subset of a projective variety X over the number field F. Theorem 8.11 of Manin’s book [Man 741. as follows. Franke observed that this zeta function can be identified with a Langlands-Eisenstein series. F) be a divisor class.

c.Kxlc 0 on RNS+(X) ifK.) is linearly effective.[Ix> 941 tions. so &(c) is negative. so U(F) is finite for sufficiently . for all for all E > 0. If X is anti-canonical. Let ANTI-CANONICAL VARIETIES AND RATIONAL POINTS 261 /I”(c) = inf(a E R such that &(c. Let a(c) = inf{t E Q such that for some n > 0. B) = O(P+y N(U.#O otherwise. Conjecture 4. E > 0. Thirdly. The main part of Batyrev-Manin lies in the following: open dense set Conjecture 4. In addition.) In particular. and take c to be an ample class. For every E > 0 there exists a Zariski U = U(c. and & extends to a continuous function on the positive cone RNS+(X) of ample elements in the N&on-Severi group. we have: a(c) = . the linear equivalence class contains an effective divisor}. is ample. n(tc + K. c. sujiciently small means contained a fixed non-empty Zariski open subset. s) converges for Re(s) > o}. (Sufficiently large means containing some fixed finite extension. a(c) depends only on the class of c in NS(X). E) of X such that Ah-3 < 44 + E. Then again.1. for every suficiently large number field F and for all s@ciently small non-empty Zariski open subsets U of X we have IL(c) = a(c). the function a extends to a continuous function on RNS+(X). then X is pseudo canonical.2. Batyrev-Manin define another function a as follows. In case the rank of NS(X) = 1. that is. Then N(U. Then a(c) is negative. suppose K. if a(c) < 0 for some ample class c. B) # O(B-) Furthermore: The function CH j?“(c) depends only on the class of c in the NtronSeveri group NS(X). Furthermore. as for the function p.

262 EXISTENCE OF (MANY) RATIONAL POINTS CX? 041 small U. c. 702. Srinivasan Ramanujan (the inventor of the circle method). Phys. For further discussion. The Hardy-Littlewood method.2. $2 in a much wider setting. VAUGHN. 21 (1987). . Math. B) N yB”“‘(lOg B)‘. 6. 3. Sci. Then as in Conjecture 4. In addition.3. Cambridge University Press (1981). 706. Reviews. for sufficiently large F and for sufJiciently small U (which may depend on F). This is a special case of one of my conjectures expressed in Chapter I. namely: R. 701. which is also implied by Vojta’ conjectured quantitative height inequalities. K. 705. see especially pp. 3. J. Batyrev conjectured [Bat 903: 4. pp. Atle SELBERG. Springer-Verlag. and Ram Murty’s review of this paper. there exists a number y = y(X) 2 0 and an integer r 2 0 such that Conjecture N(X. 1989. see Franke-Manin-Tschinkel [FrMT 891. last paper.c + K. 698. where r is defined by the condition r + 1 = codim of the face of efictive cone where a(c). the circle method gives estimates of the conjectured type. RAMACHANDRA. in his Collected Works. 89e-11001. No. Let X be anti-canonical. For nonsingular complete intersections. p. The relevant bibliography was pointed out to me by Ram Murty. lies. Math. Reflections around the Ramanujan Centenary. 545-565. The above asymptotic expression would set Schanuel’s counting of points on projective space from Chapter II. Added in extremis in proofs: The circle method attributed to HardyLittlewood in the text has its origins in a letter from Ramanujan to Hardy. Colliot-ThCEne tells me that when the number of variables is sufficiently large with respect to the degree. C.1.

Theory of intersectionson an arithmetic surface. J. Mat. J. of Math. Izv. 60-630 J. 38 No. J. Nauk SSSR Ser. KOCHEN. 1205-1213 J. J. ARAKELOV. 213-230 M. USSR Izv. 6 (1974) pp. 35 (1971) pp. 5 (1968)pp. ADAMS. KOCHEN. Acad. ARTIN and J. 1083-1108 S. Amer. Some topics in differential algebraic geometry II: on the zeros of theta functions. ARTIN. Math. USSR Izo.pp. J. Vol. Amer. Mat. Silverman.Bibliography [Ad 663 [Ad 671 [Ara 713 [Ara 74a] W. 8 No. translation. Springer-Verlag. Nauk SSSR Ser. 1986. WINTERS. Asymptotic diophantine approximations to e. 1277-1302 S. 1. Akad. pp. translation: Math. Math. 40-408 M. Nat. 1179-l 192. ARAKELOV. Sci. International Congress of Mathematicians. Akad. ARAKELOV. Diophantine problemsover local fields I. Class field theory. Proc. 373-383 E. Benjamin 1968 J. 87 (1965)pp. USA 55 (1966)pp. Ann. Topology 10 (1971)pp. Proc. ARTIN and G. 6 (1974) pp. 5 (1971)pp. Ax. Arithmetic geometry. Cornell. Vancouver 1974. ARTIN. Diophantine problems over local fields III. ADAMS. II.Amer. Math. eds. Amer. On the solutions of analytic equations. 1269-1293. 83 (1966)pp. 277-291 M. A field of cohomologicaldimension1 which is not C.Asymptotic diophantine approximations and Hurwitz numbers. Families of algebraic curves with fixed degeneraties. Intersection theory of divisors on an arithmetic surface. 28-31 [Ara 74b] [Art 681 [Art 863 [ArW 711 [AT 68) [Ax 663 [Ax 721 [AxK 651 [AxK 661 W. Math. Invent. 717 J. 89 (1967)pp. 1167-1180 S.. J. N&on models. 71 (1966)p. J. Ax and S. Sot. 437-456 . Math. Math. Degeneratefibers and reduction of curves. G. 94 (1972)pp. Bull. TATE. 120. Ax and S.

New York. Artin’s theorem on analyA tic equations.Math. 21. 44 (1980). W. Springer-Verlag 1990 der [BoMM 891 J. BIRCH. BAKER. BISMUT and E. Preprint IHES 1988 F. 247-252 A. de preprint. 67.Proc. MORET-BAILLY. Acad. 204-216 A. Sur le calcul explicite des “classes Chern” des surfacesarithmitiques de genre 2. II.N&-on models. BISMUT and E. II. 220-228. Ann. rigid analytic version of M. Kenneth Ribet. Progress report for the Soviet-American meeting.London 1977. l-27 V. M. vol. Mordell conjecture revisited. The theory of linear forms in logarithms. 395-404 [BLR 901 S.Effective methodsin Diophantine problems I. BAKER. Symposia Pure Math. Phil. B. L. Counting points on algebraic varieties. Mathematika 4 (1957)pp. pp. BATYREV.Forms in many variables. 204-216. MANIN. J. BAKER. USSR Izv. RAYNAUD. Academic Press. 14 (1967) pp. II The Diophantine equation y2 = x3 + k. Paris 1989 . London 263 (1968)pp. London Ser. BOSCH. (1990). Contributions to the theory of Diophantine equations:I On the representationof integersby binary forms. VASSEROT. 779-781 [BiV 88b] J. 117-129. 21 (1972)pp. 55-72 [Born 901 E.translation: Math. 102-105 [Bir 611 B. 27 (1974)pp. BATYREV and J. BAKER. Sur le nombre des points rationnels de hauteur born&e des vari&s algkbriques.Points of finite order on an abelian variety.A sharpeningof the bounds for linear forms in logarithms I. BOST. Transcendence theory: advances and applications. 102-107. BATYREV and J. VASSEROT. 3 (1981)pp. The cone of effective divisors of threefolds. BOSCH. J. MESTRE. preprint The [Bos 813 S. Contemporary matics Vol. 33-36. Sot.264 [Ba 66+] [Ba 681 BIBLIOGRAPHY A. III. Izv. IV. III. Height pairing between algebraic cycles. ed. Mathematika 13 (1966) pp. BAKER. 286 (1990) pp. 1988 V. 15 (1968)pp. Trans. Sot. Ann. 17 No. F. preprint. Mat. BIRCH. 1-7 A. 173-208 A. C. Linear forms in the logarithms of algebraic numbersI. J. M. 20 AMS (1971)pp. Sci. Proc. 27-43 A. Current trends in arithmetical algebraic geometry. Acta Arith. Roy. Ecole Normale appear V. AMS 1985 Mathe- [Ba 711 [Ba 72+] [Ba 773 [Bat 901 [BaM 883 [BaM 901 [Be 851 [Bir 571 B. Roy.Math.Homogeneous forms of odd degreein a large number of variables. A (1961)pp. Comportement asymptotique de la torsion analytique associee aux puissances d’un fibrC en droites positif. MANIN. 245-263 [BiV 88a] J. pp. BOMBIERI. asymptoticsof the Ray-Singer The torsion associatedwith high powers of a positive line bundle. L~~TKEBOHMERT M. Paris 307 (1988)pp. II. 14 (1967)pp. Nauk SSSR Ser. BOWMOLOV. [Bog 801 Akad. and Ergebnisse Math. 24 (1973)pp. 24. 255 No. 195-205 and vol. BEILINSON. 1 (1981)pp.

Sci. Paris 217 (1943) pp. i Mat. 150-158 J. BRUMER and 0. DCmonstration de quelques lemmes de rehaussement. Fiz. Proc. Paris 212 (1941) pp. C. Arithmetic on curves of genus 1 (III). MCGUINNESS. Math. 127-168 C. Math. Math. Sci. The TateShafarevich group. CASSELS. W. Vanishing sums in function fields. The dual exact sequence. Sur le thtortme fondamental de la thkorie des points entiers et pseudo-entiers des courbes algkbriques. BRYUNO. MASER. CASSELS. 95 (1972) pp. CHABAUTY. 557-584 H. J. pp. S. giving proofs of the above. CASSELS. l-15 J. Arithmetic on curves of genus 1 (IV). Sot. Math. CASSELS. same title. Sur les points rationnels des variCtCs algkbriques dont l’irrtgularitk est supkrieure ?I la dimension. Arithmetic on curves of genus 1 (VII). J. 2 (1990) pp. AMS translation USSR Comput. A defect relation for equidimensional holomorphic mappings between algebraic varieties. reine angew. W. J. Zh. 217 (1965) pp. 427434 A. Springer Lecture Notes 980 (1983) D. BROWNAWELL and D. Continued fraction expansion of algebraic numbers. reine angew. W. Phys. BREEN. 12 (1962) pp. 38 (1963) pp. Sot. 4 (1964) pp.BIBLIOGRAPHY 265 [Br 831 [BrM 861 [BruM] CW 641 [CaG 721 [Car 291 [Car 331 [Cas 62a] [Cas 62b] [Cas 631 [Cas 64a] [Cas 64b] [Cas 651 [Chab 381 [Chab 411 [Chab 43a] [Chab 43b] L. Amer. Proof of the Hauptvermutung. Math. On the conjectures of Birch and Swinnerton-Dyer. CR. 214-215 (1964) pp. J. Sci. 5-31 J. CASSELS. Paris 189 (1929) pp. Sci. W. C. Acad. Bull. 413-415 [These two 1943 papers are partly incorrect. 244-248 J. The TateShafarevich group can be arbitrarily large.1 . The behavior of the MordellWeil group of elliptic curves. Annali di Math. 1022-1024 C. p. 144. reine angew. Arithmetic on curves of genus 1 (V). S. London Math. S. W. 23 No. Sot. Acad. GRIFFITHS. Sot. CR. S. Ann. CASSELS. 521523 H. 211 (1962). Math. 100 (1986) pp. 259296 J. Paris 217 (1943) pp. 211-221. J. CHABAUTY. 95-112 J. Arithmetic on curves of genus 1 (VI). See the comments in [La 781. 17 (1938) pp. reine angew. Math. Mathematika 7 (1933) pp. 375-382 A. Cambridge Philos. CARLFKJN and P. S. Two counter-examples. 336-338 C. of Math. D. Vichisl Mat. Fonctions theta et thkorkme du cube. Sur les zCros des combinations linkaires de p fonctions holomorphes don&es.R. W. 65-70 J. CHABAUTY. Acad. CHABAUTY. 216 (1964) pp. 180-189 C. 4 nr. CARTAN. Acad. Sur les Cquations diophantiennes likes aux unit&s d’un corps de nombres algkbriques fini. CARTAN. Arithmetic on curves of genus 1 (VIII). 2 (1964) pp. London Math. Proc. S.R.

Sci. 82 (1985) pp. Proc. COLLIOT-TH~Z. Wilstholz. Berkeley 1986. and G. Sot. Amer. ed. L. Math. COLLIOT-THRLBNE. London Math. Sup. CLEMENS. the generically surjective case. Intern. MORI. CHEVALLEY and A. CHAI. The Nevanlinna error term for coverings. ed. London Math. WEIL. Pub. Ann. J. Griffiths. Sot. Math. J. &ole Norm. 249-300 W. 106. 1985 C. 73-75 C. Math. Trans. Sci. Homological equivalence modulo algebraic equivalence is not finite generated. 223-251 R. J. Math. in: Topics in Nevanlinna theory. L. COLLIOT-TH~L&NE. J. CHAI and G. Acad. Invent. Math. SANSUC. Cherry. Nevanlinna theory for coverings. Sem. Les grands themes de Francois Chltelet. in: Topics in transcendental algebraic geometry. L. Ann. Pade approximations and diophantine geometry. Ergebnisse der Math. 570572 W.. 519 . Manin’s proof of the Mordell conjecture over function fields. reine angew. P. to appear J. IHES 58 (1983) pp. KOLLAR and S. Artin. Amer. SANSUC. 2212-2216 H. Stud. 22.R. 34 (1988) pp. FALTINGS. Math. Abh. 387-405 J. Enseign. Arithmetique des surfaces cubiques diagonales. Compactification of Siegel modular schemes. to appear C. S. CHATELET. 150-191 J. Springer Lecture Notes 1433 (1990) pp. 61 (1944) pp. Asttrisque 166 (1988) J. CHOODNOVSKY. Lang. Acad. Proc. CHAI. COLLIOT-T&L~NE. L. Some results about Abel-Jacobi mappings. Nat. Proc. 320 (1980) pp. CHERRY. pp. A note on Manin’s theorem of the kernel. COLEMAN. Lecture Notes Series 107. L. Math. CHEVALLEY. SALBERGER. CHOW. Sot. Univ. J. Springer-Verlag 1990 F. D. Degeneration of abelian varieties. Sci. KANEVSKY. Demonstration dune hypothbe de M. Enseign. 39 (1977) pp. 289-304 H. W. 64-653 J.266 BIBLIOGRAPHY [Chai 851 [Chai 901 [ChF [Chat 901 441 [Cher 901 [Cher 911 [Chev 353 [ChW 321 [Cho 551 [Chu 851 [Cl 831 [Cl 841 [ClKM [CoaW 881 771 [Co1 901 [C-T 863 [C-T 881 [CTCS 803 [CTKS 851 [CTSal 891 C. 253-275 D. 113-174 W. Abelian varieties over function fields. Hamburg 11 (1935) pp. to appear C. Springer Lecture Notes 1290 (1985) pp. Arithmetic on some singular cubic hypersurfaces. of Math. (3) 58 (1989) pp. COATES and A. L. Un theortme d’arithmetique sur les courbes algtbriques. l-108 J. L.L~NE and P. Congress Math. L. C. (1955) pp. eds. Princeton University Press (1984) pp. J. G. WILES. 19-38 H. in: Diophantine approximation and transcendence theory. CHERRY. COLLIOT-THRLBNE. CLEMENS. Descente et principe de Hasse pour certaines variites rationnelles. Paris 195 (1932) pp. Arithmetique des varietes rationnelles et problemes birationnels. On the conjecture of Birch-swinnertonDyer. CLEMENS. Higher dimensional complex geometry. CORAY and J. Variations sur un theme de Poincare. D. Math. L.

Math. DESCHAMPS and R. E. Ann. II. 66 (1968)pp. J. Invarianten und Normalformen elliptischer Funktionenkiirper.Proc. SANSUC. ibid. DELIGNE.No. MBNBGAUD. La descentesur les varittCs rationnellesII. History of the theory of numbers II. COLLIOT-TH~L~XNE. Formes modulairesde poids 1. 47-56 L. 30 (1966) pp.g”(t). 391-401 N. et sur une hypothtse de Schinzel. Roy. Izv. A. CORNALBA and J. SERRE. DEMJANENKO. in: SCminairesur Ies pinceaux arithmktiques:La conjecturede Mordell. Mat. DANSET. L. Ecole Norm. Math. SANSUC and P. DELIGNE and D. Transl. Washington DC 1920. Les constantesdes Cquationsfonctionelles.BIBLIOGRAPHY [CTS 821 267 J. Rational points of a classof algebraic curves. 374 (1987) pp.. 173-177 V. COLLIOT-T&L~NE and J. 37-107. J. 285-303 H. Math. SANSUC. 127-135 M. France 106 (1978) pp. 31 (1985)pp. 19 (1970)pp. London Ser A 272 (1963)pp. Sup. DEL!GNE and J. Sup. Sci. Akad. DAVENPORT. J. Norske Vid. K. Acta Arith. 47 (1941)pp. 33-53 [CTS 871 J. 16-28 P. Comp.Pub. Asterisque. DAVENPORT. On f3(t) . Acta Arith.SPminaire Delange-Pisot-Poitou. DEMJANENKO. 34 (1979) pp. 825-835 M. 51. MCthode du cercle adtlique et principe de Hassefin. Bull. SWINNERTON-DYER.On A4 + B4 + C4 = D4. Acta Arith. On a question of Lehmer and the number of irreducible factors of a polynomial. DAVENPORT. Math. J. IHES 35 (1969)pp. A.-P. Carnegie Institution. Sot. L. Ser. 1373-1396. COLLIOT-T&L&NE and J.127(1985)pp..reprinted by Chelsea1966 E. Ann Arbor Publishers1962 H. Mat. 279-287 DEURING. Z. 41 (1982)pp. Ann Arbor. J. 7 (1974)pp. 21 (1988)pp. 246-272 V. 507-530 V. 271-278. Amer. 373 (1987) pp. DELIGNE. with applications to the moduli spaceof curves. Zametki 3 (1968)pp. J. Lucien Szpiro. HARRIS. DEMJANENKO. An estimate of the remainder term in Tate’s formula. Math. Forrh. reine angew. Nauk SSSR Ser. 54 (1987) pp. 86-87 P. A. Enseign. MUMFORD. 375-492 [CTSSD 871 J. 184 (1988)pp. Analytic methods for diophantine equations and diophantine inequalities. No. (Trondheim) 38 (1965)pp. Applications rationnelles &parables dominantes sur une variCti du type ginCra1. Divisor classes associated with families of stable varieties. Sot. L. ELKIES. Cubic forms in 16 variables. DICKSON. Ann. 75-109 P. . Euler’sconjecture. 249-255 P. Notes 3 (1968)pp. Sot. translation: Math. Sur le principe de Hasse et l’approximation faible. Ecole Norm. Selsk. Duke Math. DOBROWOLSKI. 25 (1973-74) pp. 11 (1969/70). l-66 H. The irreducibility of the space of curves of given genus. Intersections of two quadrics and Chltelet surfacesI. Reprbsentations I-adiques. L. ed. Sci. II. Math. 72-168 [CorH 883 [Dan 853 [Dav 621 [Dav 631 [Dav 651 [Del 701 [Del 831 [DelM 691 [DelS 741 [Dem 661 [Dem 681 [Dem 741 [DesM 781 [Deu 413 [Di 201 [Do 791 [El 881 M. Math. 455-475 R.

349-366. FREY. FRANKE. to appear G. 1989:CR.pp. Paris 309 (1989)pp. FALTINGS. SOULS. SOULS. Arithmetic intersection theory. Math.Links betweenstable elliptic curves and certain diophantine equations. III. Acad. 108-118 . 33-114 G. 4 (1976)pp.268 [EsV 841 [EsV 901 [Fa 83a] [Fa 83b] H.B = C. 623-663 D. Rational Points. J. Preprint IHES.E. to appear G. 119 (1984)pp. 78 (1984)pp. SOULS.Links betweensolutionsof A . Ulm 1987. Carbondale 1979. GILLET and C. 51 (1987)pp. Rational points of bounded height on Fano varieties. Ann. GILLET and C.eds.Annales Universitatis Saraviensis. Sup. 421-435 G.Inoent. GILLET and C. ESNAULT and E. pp. Pisa. Indian Math.preprint. l-33 in G. Preprint IHES. Sci. Endlichkeitssltze fur abelsche Varietaten iiber Zahlkiirpern. FREY. Springer Lecture Notes 751 (1979)pp. 445-490 H. MANIN and Y. Conjectureson elliptic curves over quadratic fields.B = C and elliptic curves. Schlickewei. Calculus on arithmetic surfaces. 1988 H.Un thiortme de Riemann-Roth Grothendieck arithmetique. Invent. CR. Sot. eds. Effective bounds for semipositive sheavesand for the height of points on curves over complex function fields. FALTINGS. P. J. J. Acad. Moduli spaces. pp. GOLDFELD. Arakelov’s theorem for abelian varieties. 73 (1983)pp. Analytic torsion and the arithmetic Todd genus. G. Silverman. SeriesMath. Preprint IHES. ed. 21-54 G. SOUL@ Characteristicclasses algebraicvector for bundleswith hermitian metric. Links between elliptic curves and solutions of A . Scuola Norm. 117-145 G. 31-62 H. appearedin Topology 30 (1991)pp. Math. SOULS. Diophantine approximations on abelian varieties. l-40 G. Comput. Amplitude arithmetique. in: Number theory. Paris 307 (1988)pp. 9-27 G. English translation: in Arithmetic geometry. SerieIV. Springer-Verlag 1986. B. FALTINGS. 887-890 H. 1988 H. Nathanson. FALTINGS and G. VIEHWEG. 929-932 D. VIEHWEG. Math. of Math. 95 (1989) pp. 387-424 G. FREY.Ann. FALTINGS. W~~STHOLZ. Math. Sci.H. GILLET and C. Invent. Dyson’s lemma for polynomials in several variables. GILLET and C. 73 (1983)pp. Wirsing. 337-347 G. FALTINGS. Invent. Heights. Seminar Bonn/ Wuppertal 1983-1984. Vol. 1 (1986)pp. TSCHINKEL. M. Cornell. [FaW]. GOLDFELD. The classnumber of quadratic fields and the conjecture of Birch-Swinnerton-Dyer. BIBLIOGRAPHY [Fa 84a] [Fa 84b] [Fa 84~1 [Fa 901 CFaW [FrMT [Fr 861 [Fr 87a] [Fr 87b] [GiS 88a] [GiS 88b] [GiS 88~1 [GiS 88d] [GiS 891 [Go 761 [Go 793 891 ESNAULT and E. Springer Lecture Notes 1380(1989)pp. FALTINGS. in: Number theory.Math. in [FaW].Vieweg 1984 J.

pp. GROTHENDIECK. Benjamin 1969 N. 73 (1961) pp. 87-106 B. 41-74 M. 145-220 B. J. Springer-Verlag 1977 D. in: Chern Symposium 1979. A-304. Hsiang. Autour d’une conjecture de Serge Lang. Conference on Fermat’s Last Theorem.. Gkom. Amsterdam 1968 A. Invent. Ann. of Math. Math. Proc. Points de torsion sur les sous-varittts de variCtts abkliennes.(N). Schemata over local rings. HINDRY. 617-632 M. Ellis Horwood 1984. eds. Acta Math. No. Dix exposCs sur la cohomologie des Schemas. Pub. Ser. R. Algtbrique Bois-Marie 1967-1969. On the conjecture of Birch and Swinnerton-Dyer for elliptic curves with complex multiplication. 21-51 P. GRIFFITHS. Springer-Verlag 1980. 225-320 A. 21-35 P. 615-620 W. GREENBERG. Kobayashi et al.R. HINDRY. J. 219-236 B. GROSS. Rankin. GREENBERG. North-Holland. of Math. 256-266 M. Birkhiuser 1982. Algebraic geometry. Paris. 94 (1971) pp. Durham conference. Ann. ed. Sci. Ser. Lectures on forms in many variables. pp. ZAGIER. W. GROSS. Two applications of algebraic geometry to entire holomorphic mappings. Acad. Math. GREENBERG. II. Acad. HEATH-BROWN. Heegner points and derivates of L-series. C. III. S. 624-648 M. GREEN. 94 (1988) pp. GROSS. HINDRY. GREEN. GRIFFITHS and J. R. Math. 84 (1986) pp. (3) 47 (1983) pp. Paris. Heights in families of abelian varieties. London Math. 78 (1963) pp. A. IHES 25 (1965) pp. Springer Lecture Notes 288 (1972) pp. Amer. Modular forms. 225-257 M. 575-603 . 313-523 R. A-305 (1987) pp. of Math.BIBLIOGRAPHY 269 [Gra 651 CGrM781 CGrw891 [GrG 803 [Grbg [Grbg [Grbg 611 631 691 [Grlf 651 [Gri 711 [GrK 731 [Gros 821 [Gros 841 [Gros 901 [GrZ [Grot [Grot 861 681 721 [Ha 773 [H-B 831 [Hi 87a] [Hi 87b] [Hi 881 H. 1990 B. C. 9. Duke Math. 219-221 M. 130 (1973) pp. Ann. GRAUERT. HARTSHORNE. Complex analytic properties of certain Zariski open sets on algebraic varieties. Kolyvagin’s work on modular elliptic curves.R. Math. Cubic forms in ten variables. Nevanlinna theory and holomorphic mappings between algebraic varieties. W. Modeles de N&on et monodromie. GROTHENDIECK. Irreducible subvarieties and rational points. Points quadratiques sur les courbes. pp. 31 l314 M. 12 (1987) pp. GREEN and P. Sci. Groupe de Brauer I. Schemata over local rings II. GREENLEAF. GRIFFITHS. Skmin.-Y. 131-149 M. Mordell’s Vermutung iiber rationale Punkte auf algebra&hen Kurven und FunktionenkGrpern. pp. Chichester. Invent. 3 (1989) pp. Sot. SGA7 I. 100 (1978). J. 58 No. KING. (1965) pp. Holomorphic maps to complex tori. Heegner points on X. Math. Amer. GROSS and D.

Mat. Finiteness theorems in geometric class field theory. IITAKA. JournCes ArithmCtiques de BesanGon 1985. Amer. 232 (1978) pp. Sbornik 88 (1972) pp. Hyperbolic manifolds and holomorphic mappings. 386 (1988) pp. reine angew. Number Theory 24 (1986) pp. SpringerVerlag 1982 V. A. Heights and Arakelov’s intersection theory. 525-544 S. 32-98 P. HIRATA-KOHNO. 57 (1980) pp. translation: Math. A counterexample to the Hasse principle for a system of two quadratic forms in 5 variables. Minorations effective de formes 1inCaires de logarithmes sur les groupes algkbriques. to appear N. Tata Institute. to appear C. (1982) pp. Math. translation: Math. Ann. Sci. 285-314 Y. 93 (1988) pp. IITAKA. KATO and T. Sur les fonctions abCliennes singulibes I et II. 76. 175-189 S. 31-37. 30-36 J. 1 (1985) pp. Application of the conjecture on the Manin obstruction to various diophantine problems. LANG. HOOLEY. HUMBERT. pp. 153-166 . 233. Iwanami Shoten. Much. 419-450 N. Fat. ISKOVSKIH. 23-38 P. J. Complex analysis and algebraic geometry. Algebraic geometry. Nagoya Math. KAWAMATA. On nonary cubic forms. 107 No. J. P. IGUSA. 6 (1900) p. KAMIENNY. Lectures on Galois cohomology of classical groups. KUYUMAKI. 97-100 M. 1969 S. 575577 V. ISKOVSKIH. Texts Math. The dimension of fields and algebraic K-theory. Invent. HINDRY and J. Marcel Dekker 1970 S. IA Math 23 (1976) pp. The canonical height and integral points on elliptic curves. J. 279 J. Math. A. Tokyo Sect. Negative vector bundles and complex Finsler structures. KNESER. Univ. 57 (1975) pp. KOBAYASHI. Hypersurfaces solutions g une Cquation de Pfaff analytique. Notes 10 (1971) pp. Grad. HIRATA-KOHNO. AstCrisque 147-148 (1987) pp. Formes linkaires de logarithmes de points algCbriques sur les groupes algtbriques. Torsion points on elliptic curves and q-coefficients of modular forms. Tokyo 1977. HRILJAC. USSR Sbornik 17 (1972) pp. KOBAYASHI. IITAKA. Much. Lectures on forms of higher degree. Springer-Verlag 1978 S. 307-314 K. J. Mat. 5 (1899) p. Bombay. preprint D. Math. 253-257. SILVERMAN. Logarithmic forms of algebraic varieties. J. Math. Birational properties of a surface of degree 4 in I$‘. KANEVSKY. Invent. KATZ and S. Math. Zametki 10 (1971) pp. JOUANOLOU. On Bloch’s conjecture.270 [His [Hir [Hir 881 90a] 90b] BIBLIOGRAPHY [Ho 881 [Hr 851 [Hu 18991 Ck 781 [Ii 761 [Ii 773 [Ii 821 [Isk 711 [Isk 721 [Jo 781 [Kam [Kan 901 871 [KatK [KaL [Kaw 861 821 SO] [Kn 691 [Kob [Kob 701 753 M. 239-245 S. Enseign. Tata Institute for Fundamental Research. On logarithmic Kodaira dimension of algebraic varieties. J. 229-244 N.

IHES. Math. Universal bounds on the torsion of elliptic curves. Sot. KOLYVAGIN. 303-326 V. Amer. 522-540. Ann. Abeliun varieties. Division points on curves Ann. France 107 (1979) pp. KUBERT. 78 No. translation: Math. LANG. KOLYVAGIN. LACHAUD. KOLLAR. Modular units inside cyclotomic units. Amer. 1079 S. USSR Izu. Bull. KUBER~. 32 (1989) pp. Math. J. J. Sot. LANG. 66 (1960) pp. Math. LANG. 521-533 S. Ann.BIBLIOGRAPHY 271 [KoO 713 [KoO 75-j [Koll 891 [Koly 881 [Kol 903 [Ku 761 [Ku 793 [KuL [KuL 751 791 [Lac 821 [LandP [La 511 [La 541 [La 56a] [La 56b] [La 56~1 [La 591 [La 60a] [La 60b] [La 621 [La 641 [La 65a] 661 and T. France 84 (1956) pp. Ann. OCHIAI. 218 (1975) pp. 161-178 G. Math. Minimal models of algebraic threefolds: Mori’s program. LANG. LANG. reprinted by Springer-Verlag. Math. 555-563 S. Adrisque 177-178 (1989) pp. 3 (1956) pp. 86 (1964) pp. 193-237 D. Izu. KOBAYASHI and T. LANG. Math. Math. Proc. LANG. of Math. Amer. Amer. Diophuntine geometry. LANG. 38 (1979) pp. LANG. 373-390 S. 52 (1988) pp. Enseign. Counterexamples to Euler’s conjecture on sums of the powers. Math. LANG. KOBAYASHI . LANG. 28 (1982) pp. Mat. Math. Units in the modular function field I. of Math. 33 (1976) pp. Mappings into compact complex manifolds with negative first Chern class. LANG. Meromorphic mappings into compact complex spaces of general type. 55 (1951) pp. 136-148 S. 64 No. Sur les series L dune variete algebrique. Algebraic groups over finite fields. Integral points on curves. Sot. LANDER and T. Universal bounds on the torsion of elliptic curves. Une presentation adelique de la serie singuliere et du probleme de Waring. Bull. 67-96 D. Sot. Math. Bull. Sot. 229-234 S. SCminuire Bourbaki (1988-1989). KUBERT and S. Euler systems (1988). Sot. On quasi algebraic closure. Diophantine approximations on toruses. LANG. 1983 S. 523-541 V. (IV) LXX (1965) pp. Math. London Math. A. to appear in Birkhauser volume in honor of Grothendieck D. Some applications of the local uniformization theorem. 2 (1956) pp. 121-128 D. PARKIN. Diophantine applications. 240-249 S. Nauk SSSR Ser. Mat. Pub. 1960 S. Unramified class field theory over function fields in several variables. Amer. 285-325 S. 362-374 S. Puru Appl. Math. KUBERT and S. Invent. OCHIAI. A. Finiteness of E(Q) and III(E/Q) for a class of Weil curves. 385-407 S. Wiley-Interscience. New York 1959. Akad. 1962 S. J. Compositio Math. J. 31 (1975) pp. Interscience. Some theorems and conjectures in diophantine equations. Bull. 72 (1966) p. Japan 23 (1971) pp. 76 (1954) pp. 7-16 J. 139-169 L.

Sot. Elliptic curoes: Diophantine analysis. Invent. Bull. Complex multiplication. Equations diophantiennes exponentielles. LANG. LANG. Math. LANG. Sur une conjecture de Serge Lang. LANG and J. Springer-Verlag 1978 S. AddisonWesley. Lectures on Nevanlinna theory. 770-787 S. J. Amer. 1155125 S. Amer. J. Sot. 14 (1986) pp. Math. Math. 124-132 S. Acad. Springer-Verlag 1988 S.R. Une nouvelle demonstration du theoreme d’isogenie d’apres D. Addison-Wesley 1970. Grundlehren 255. Springer-Verlag 1983 S. LAURENT. Old and new conjectures in diophantine inequalities. Duke Math. LANG. 6177628 M. Math. 1986 S. Fundamentals of diophantine geometry. l-112 S. Ann. Report on diophantine approximations.272 BIBLIOGRAPHY [La 65b] [La 661 [La 703 [La 711 [La 741 [La 751 [La 781 [La 83a] [La 83b] [La 83c] [La 861 [La 871 [La 883 [La 90a] [La 90b] [La 90~1 [LaN [LaT [Lange [Lau 591 581 871 843 [Lau 863 [Lew 521 [Li 741 S. 81 No. France 93 (1965) pp. LANG. 1966 S. LANG. Lang. Amer. 177-192 S. Introduction to Arakelov theory. papers dedicated to Shafarevich. Birkhatiser 1983 S. 80 No. Introduction to diophantine approximations. 1 (1959) pp. C. 4355437 . Math. 23 (1990) pp. LANG. 659-684 H. LANG. 97 No. 80 (1958) pp. Sot. Bull. The error term in Nevanlinna theory II. 299-327 M. Seminaire de thdorie des nombres. Math. Amer. Hyperbolic and diophantine analysis. Introduction to complex hyperbolic spaces. Math. LANGE. 22 (1990) pp. Bull. 55 (1987) pp. Bull. Amer. edited by Artin and Tate. 5 (1974) pp. Arithmetic and geometry. Sot. LAURENT. TATE. LANG. Choodnovsky. W. 78 (1984) pp. 37-75 S. Higher dimensional diophantine problems. Abelian varieties with several principal polarizations. Sot. 635-677 S. LANG. Bull. et G. Springer-Verlag 1983 S. Principal homogeneous spaces over abelian varieties. Amer. Conjectured diophantine estimates on elliptic curves. 955118 S. 159-205 S. Amer. J. Algebraic Number Theory. reprinted by Springer-Verlag. 56 (1952) pp. J. V. LANG. Paris 1985-1986 D. Springer-Verlag 1987 S. Rational points of abelian varieties over function fields. Amer. LEWIS. LANG and A. Paris 279 (1974) pp. 77 (1971) pp. Cubic homogeneous polynomials over p-adic number fields. LANG. LANG. S. Bull. Math. LANG. LIARDET. 1 (1975) pp. Sci. Math. of Math. V. Transcendental numbers and diophantine approximations. Division points of elliptic curves and abelian functions over number fields. NBRON. LANG. J. Math. Sot. Cherry Springer Lecture Notes 1433 (1990) pp. in: Topics on Nevanlinna theory. 4733478 P. LANG.

107 (1933) pp. 37 (1964) pp. Mat. MANIN and Y. translation: Transl. Amer. MANIN.. 219-230 R. MANIN. Akad. 101-123 K. Acad. ed. 87 (1990) pp. algebra. 6 (1972) pp. 59 (1966) pp. Mat 22 (1958) pp. 3 (1969) pp. Progr. II. The p-torsion of elliptic curves is uniformly bounded. MASON. Math. 33 (1969) pp. 19-65. MANIN.. pp. Surveys 41: 2 (1986) pp. USSR lzv. Sot. Math. Cubic forms. Sot.BIBLIOGRAPHY 273 [Li 753 CJ-UYI [Mae [Mah 831 331 [Man 581 [Man 631 [Man 641 [Man 691 [Man 701 [Man 70b] [Man 721 [Man [MaT 741 861 [MaZ [Mas [Mas 721 831 84a] [Mas 84b] [Mass 811 Sur une conjecture de Serge Lang. Lu and S. Proc. C. Math. Mat. 1363-1390. 51-116 J. Amer. Mat. 149-157 R. (1975) S. Springer Lecture Notes 1068 (1984) pp. T.. MANIN. 213-222 P. Sb. Math. Akad. Cambridge Philos. TSFASMAN. 12 (1970) pp. H. MASON. Heights on families of abelian varieties. translation: Russian Math. YAU. Math. Mat. The hyperelliptic equation over function fields. USSR. Uspekhi Mat. SO-82 K. 189-234 J. Zzv. 1970 Tome 1 pp. 12. Zur Approximation algebraischer ZIhler I (fiber den griissten Primteiler binBrer Formen). Akad. Nauk SSSR Ser. Le groupe de Brauer-Grothendieck en gComttrie diophantienne. Nauk 41 (1986). Jager. Holomorphic curves in surfaces of general type.. Diophantine equations over function fields. Birkhsuser (1981) pp. MANIN. 59-78 J. MANIN. Mat. USSR. Nauk SSSR Ser. Math. MAEHARA. Nut. Lecture Notes Series 96. Noordwijkerhout 1983. Izv. 36 (1972) pp. 325-342 J. MASER. translation: Transl. . 691-730 J. Akad. Sbornik 89 (1972) pp. Proc. Ann. SSSR Ser. MANIN. Mat. 459-465.A. II. Akad. The refined structure of the N&on-Tate height. geometry and arithmetic. 401411 J. its variants and applications. 50 (1966) pp. Rational varieties. 1974. A. Parabolic points and zeta functions of modular curves. North-Holland. London Math. Actes CongrLs Intern. Ser. LIARDET. II. 1395-1440. translation: Math. Izv. Amer. Sbornik 83 (1970) pp. ZARHIN. MASON. 82-110 J. C. in: Number theory. C. Math. MAHLER. 27 (1963) pp. U. 737-756. Cambridge University Press 1984 D. Sot. MANIN. A finiteness property of varieties of general type. Algebraic curves. MANIN and J. 433-438 J. Sci. Nauk SSSR Ser. 28 (1964). Izv. 262 (1983) pp. 93 (1983) pp. Ser. Rational points of algebraic curves over function fields. 331-348 translation: Math. Nauk. A&risque 24-25. over fields with differentiation. translation: Math.. 171-181 R.S. Izv. Small values of the quadratic part of the N&on-Tate height. translation: Transl. 19-64 J. second edition (with appendix) 1986 J. Math. Izv. Equations over function fields. Sot. Ann. Nauk SSSR Ser. The Tate height of points on an abelian variety. Ser. Sot.

Proc. 44 (1978)pp. Invent. D. MAZUR. Someeffective estimates elliptic for curves. IHES 47 (1977) pp. MASER. 209-232 [Mil 86a] J. 281 (1988) pp. Math. 1299162 [Maz 831 B. 597-615 MAZUR. eds. l-24 D. eds. 179-192 [MorB 851 MORET-BAILLY. Math. Serre. Comps. On the Kodaira dimension of minimal threefolds. Invent. Modular curves and the Eisensteinideal. Cornell. 249-267 [Miy 881 Y. 183-266 B. 58 (1986) pp. MAZUR. International Congress Math. H. Rational points of abelian varieties with values in towers of number fields. 33-186 [Maz 781 B. Abelian varieties.Askrisque 129 (1985) . Ann. 103-150 [Mil 86b] J. Rational isogeniesof prime degree. 103-109 [Maw 911 Way 641 [Maz 721 D. 1088148 [Maz 771 B. MASSER. PERSSON. W~JSTHOLZ. of Math. in ditto. 100 (1990)pp. Ann. 53 (1984) pp. ed. MORDELL. Arithmetic of Weil curves. D. J. MIYAOKA. Sot. Torsion group of elliptic surfaces. 18 (1972) pp. MASTER and G. L. Erlangen 1988. Springer-Verlag 1986. Math. Linear relations on algebraicgroups. G. Modular curves and arithmetic. WDSTHOLZ. J.274 BIBLIOGRAPHY [Mass 843 [Mass 851 [Mass 881 [Maw 90a] [Maw 90b] Small values of the quadratic part of the N&on-Tate height on an abelian variety. B. period relations on abelian 79 W. in: Arithmetic geometry. Lange Springer Lecture Notes 1399 (1990) pp. Proc. 248-262. S. Pinceaux de varietes abeliennes. Math. 185-211 [Mes 863 J. B. Cambridge Philos. Barth. P. Compos. Jacobian varieties. 1-61 [Maz 761 B. MAZUR. A. 21 (1922)pp. Math.pp. Estimating isogenieson elliptic curves. MIYAOKA and S. On the rational solutions of the indeterminate equation of the third and fourth degrees. Math. A numerical criterion for uniruledness. W. Small values of heightson familiesof abelian varieties. MASER. CambridgeUniversity Press 1988. Silverman.-F. 153170 D.eds. of Math. P. Formules explicites et minorations des conducteurs de varietts algebriques. in: New advances in transcendence theory. MILNE. 167-212 CMirP 891 R. in: Modular functions of one variable. W~~STHOLZ. D. Invent. MAZUR. MAZUR. MASSER and G. MESTRE. 65-69 [Mord 221 L. Warsaw (1983)pp. 3 (1964) pp. in: Diophantine approximation and transcendencetheory. G. 124 (1986) pp. MORI. MIRANDA and U. Rational points on modular curves. No. Math. Baker. S.pp. MILNE. Pub. 109-148 D.Math. J. Wiistholz Springer Lecture Notes 1290 (1985) pp. Minimal varieties and isogenies. 25 (1974) [Maz 741 pp. Math. Binary forms over number fields. pp. 325-332 [MiyM 863 Y. in: Arithmetic of complex manifolds. ed. MASSER and G. Zagier Springer Lecture Notes 601 (1976) pp. Ann. 72 (1989)pp. to appear MAY. Math. Comput. Invent.

Raynaiid. Ann. Inst. 1-21 . J. Tokyo-Nikko. Res.BIBLIOGRAPHY [Mori [Mori [MOM 821 871 831 275 [Mum [Mum [Mum [Mum [NaN 651 661 741 831 811 [Ne 521 [Ne 551 [Ne 641 [Ne 651 [Nev 251 [Nev 533 [Nev 701 [No 81a] [No Sib] [No 851 [No 871 [Och 771 COgg 671 S. IHES 21 (1964) pp. NOGUCHI. Analytic functions. M. Birkhauser 1983 M. Proceedings of Symposia Pure MathematicsAMS 46 (1987) pp. MUMFORD. Modeles minimaux des varietes abtliennes sur les corps locaux et globaux. NEVANLINNA. 287-354 D. Ann.2nd ed. 4 (1965) pp.Oxford University Press 1974 D. Math. Springer Lecture Notes 1016 (1983) pp. Lemma on logarithmic derivative and holomorphic curves in algebraic varieties. 27-46 J. On holomorphic curves in algebraic varieties with ample irregularity. pp. SpringerVerlag 1953 R. N~GUCHI. N~RON. (2) 116 (1982) pp. 133-176 S. Zur Theorie der meromorphen Funktionen. MUMFORD. Problemes arithmetiques et geometriques rattaches a la notion de rang dune courbe algebrique dans un corps. of Math. Math. On the equations defining abelian varieties I. 87 No. Elliptic curves and wild ramification. Shioda. T. Sci 90 (1981) pp. Math. of Math. Math. MUKAI. Moduli spaces of holomorphic mappings into hyperbolically imbedded complex spaces and locally symmetric spaces. NBRON. 1 (1966) pp. J. Proc. 213-233 J. Amer. Amer. 269-288 in S. 43 (1977) pp. A higher dimensional analogue of Mordell’s conjecture over function fields. l-99 R. Ann. eds. MUMFORD. Invent. MORI. Math. Math. MORI and S. 258 (1981) pp. Math. Springer-Verlag 1970 J. Invent. Tokyo 1955. 82 No. Publ. Math. 207-212 J. Invent. 249-331 R. 83 (1981) pp. 93 (1988) pp. Proc. N~CXJCHI. 21 No 1 (1985) pp. 83-96 A OGG. MORI. 101-166 A. Quasi-functions et hauteurs sur les varittis abeliennes. Math. N~GUCHI. France 83 (1952) pp. MUMFORD. Math. Abelian varieties. 15-34 T. Pub. Nagoya Math. International Symposium Algebraic Number on Theory. Polarizations on an abelian variety. Hyperbolic fiber spaces and Mordell’s conjecture over function fields. Arithmetique et classes de diviseurs sur les varietts algebriques. NEVANLINNA. The uniruledness of the moduli space of curves of genus 11. Appendix. 2 (1965) pp. Indian Acad. 125-128 A. Sci. in: Algebraic geometry. 139-154 A. Acta Math. Sci. A remark on Mordell’s conjecture. OCHIAI. Sot. 89 (1967) pp. Eindeutige analytische Funktionen. 334-352 D. NORI. NARASIMHAN and M. translation and revision of the above. 46 (1925) pp. J. Threefolds whose canonical bundles are not numerically effective. Bull. 1007-1016 D. Classification of higher dimensional varieties. NEVANLINNA. NBRON. NBRON. Tutu lectureson Theta I. 361-482 A.

l-45 L. PECK. Paris 1980/81. . Modular correspondences. M. N. Math. 244-259 S. Une propriete de specialisation continue. USSSR Izv. 431-476 K. 11 (1985)pp. Toulouse (1990). Hauteurs et isogenies. 71 (1983) pp. 71 (1949) pp.Schtmas en groupes de type (p. RIBET. and N. 32 (1968). Math. 199-234 R. Math. Sometimeseffective Thue-Siegel Roth-Schmidt Nevanlinna bounds. Annales de la Fat. in: Algebraic geometry. F. ROBBA. PARSHIN. Math. 243-270 A. 1-19 M. 387-402 M. 241-280 [OS 851 [Par 681 [Par 731 [Par 861 [Par 891 [Pat 851 [Pet 491 [Ra 64-651 [Ra 741 [Ra 83a] [Ra 83b] [Ra 83~1 [Ra 851 [RDM 621 [Ri 90a] [RI 90b] [Rob 81a] M. edited by Artin and Tate.translation: Proc. 2 (1989)pp. Akad. dedicated to I. 207-233 M.p). ed. Shioda. 1986 A. heights and isogenies of abelian varieties.Around the Mordell conjecture for function fields and a conjecture of Serge Lang. Math. N. appearingwith delays P. Izv. Anal. 4 (1962) pp. 347-389 A. RAYNAUD. N. PARSHIN.Caracteristique d’Euler-Poincare dun faisceau et cohomologie des varietes abeliennes. OSGOOD. METROPOLIS. Math. Numer. J. Bourbaki 196465 #286 M. Algebraic curves over function fields. N. Math. number theoretic-differential equations approach A to generalizing Nevanlinna theory.On modular representationsof Gal(Q/Q) arising from modular forms. Invent. OSGOOD. J. RAYNAUD. France 102 (1974) pp. Bull. Math. Finitenesstheoremsand hyperbolic manifolds. Arithmetic and geometry. 26 . Asterisque 127. the application of ramified coverings to the On theory of Diophantine Geometry. Trudy Mat. Inst. Gottingensis. RAYNAUD. Schriften. From the Taniyama-Shimura conjecture to Fermat’s last theorem. 1145-1170 A.No. Sbornik 180 No. Continued fraction expansionsof algebraic numbers. eds. Steklov Inst. Raynaud. Lucien Szpiro. RAYNAUD. 100 (1990)pp.Sous-varietesdune variete abtlienne et points de torsion. RAYNAUD. RIBET. Shafarevich. Sonderforschungsbereichs Geom. Hardy-Littlewood method and diophantine The analysis in light of Igusa’s work. J. T. 2 (1968)pp.S&m.Preprint IHES. PARSHIN. Birkhluser 1983 M. Springer Lecture Notes 1016 (1983) pp. or better. PATTERSON. Sci. Amer. 23 (1981) pp. p. l-15 C. . Groupe d’dnalyse Ultrametrique. F. Mat. 1985pp. RAYNAUD. Steklova 132 (1973).Courbes sur une variete abelienneet points de torsion. RICHTMYBR. Seminairesur les pinceaux in arithmetiques: La conjecture de Mordell. J.276 [OS 811 BIBLIOGRAPHY C. Diophantine equations in algebraic number fields. Invent. Nauk SSSR Ser. 68-84 K. Math. DEVANEY M. Indian J. translation: Math. Number Theory 21 (1985) pp. des Sci. pp. PARSHIN.

Cunad. 505-524 P. A&risque 147-148 (1987) pp. and the Noether formula for arithmetic surfaces. Groupe de Brauer et arithmktique des groupes algCbriques linbaires sur un corps de nombres. 55-62 J. Math. Invent. 189-201 W. reine ungew. 187-253 W. 443-449 W. F. 63 (1987) pp. SALBERGER. Actu Math. A metrical theorem in diophantine approximation. Tate-Shafarevich groups and L-functions of elliptic curves with complex multiplication. Heights in number fields. Pub. 371-404 T. J. Invent. to appear K. Conductor. 244-296 T. SCHMIDT. surfaces cubiques et intersection de deux quadriques. Lectures on old and new results on algebraic curves. discriminant. Zero cycles on rational surfaces over number fields. Math. Simultaneous approximation to algebraic numbers by rationals. 327 (1981) pp. 175-197 P. Principe de Hasse. Ru and P. J. reine ungew. PropriCtC d’approximation pour les ClCments algkbriques. SCHANUEL. 183-207 S. 89 (1987) pp. Math. Duke Math. Springer Lecture Notes 785 (1980) W. Mathematika 2 (1955) pp. Invent. 70 (1964) pp. ROTH. SCHMIDT. Math. Bull. l-20 M. SANSUC. SALBERGER. Band 5. Actu Math. Math. Birkhauser 1984 W. Math. J.{2n + 1 hyperplanes in general position}. 91 (1988) pp. (1985) pp. RUBIN. 527-560 K. 619-631 W. Compositio Math. pp. Enseign. The density of integer points on homogeneous varieties. 881 [Sal 863 [Sal 881 [Sam 66a] [Sam 66b] [San 811 [San 851 [Sch 641 [Sch 793 [Schm 601 [Schm 701 [Schm 711 [Schm 801 [Schm 841 [Schm 851 [Schn 361 P. 89 (1987) pp. Some observations on motivic cohomology of arithmetic schemes. Approximation to algebraic numbers. Amer. SCHMIDT. DMV Seminar. 17 (1971) pp. SAMUEL.BIBLIOGRAPHY [Rob 81b] [Rot [RuW 551 901 277 [Ru 87a] [Ru 87b] [Sai 891 [Sai. SAITO. 125 (1970) pp. SCHMIDT. Global units and ideal class groups. RUBIN. SCHMIDT. Invent. Birkhauser 1986. J. J. Integral points of P” . SAMUEL. Math. Diophantine approximation. Math. 22 (1960) pp. SCHANUEL. 3-14 K. ROBBA. SLminaire de thdorie des nombres 1985-1986.-M. Rational approximation to algebraic numbers. 57 (1988) pp. 175 (1936) pp. SANSUC. Math. 51 l-526 S. Math. 12-80 J. Math. Sot. SCHMIDT. 262-263 S. On the arithmetic of conic bundle surfaces. Bombay 1966 P. 182-192 . WONG. ComplCments g un article de Hans Grauert sur la conjecture de Mordell. 151-173 P. Tata Institute of Fundamental Research. 98 (1989) pp. Analytische Methoden fir Diophuntische Gleichungen. SAITO. Paris. IHES 29 (1966) pp. J. Bull. France 107 (1979) pp. iiber die Approximation algebraischer Zahlen. Sot. SCHNEIDER. JournCes arithmktiques de Besanqon 1985. On heights in number fields.

Mordell-Weil lattices and sphere packings. SCminaire Delange-Pisot-Poitou #19. Benjamin 1968 J-P. 492-517 I. Facteurs locaux des fonctions zeta des vari&s algCbriques. Profinite groups. Inst. pp. 440-481 G. 195-197 . 15 (1972) pp. Math. SERRE. III Proc. Invent. 1969-1970 J-P. Japan 24 (1972) pp. Tate’s conjecture on the endomorphisms of abelian varieties. in Faltings-Wiistholz et al. BIBLIOGRAPHY [See 671 [Se1 511 [Ser 681 [Ser 69-701 [Ser 721 [Ser 871 [Ser 891 [SeT 681 [Sha 631 [ShT 671 [Shat 721 [Shi 71a] [Shi 71b] [Shi 733 [Shi 891 [Shio 721 [Shio 731 [Shio 89a] [Shio 89b] [Shio 89~1 SCHAPPACHER. Ann. Preprint 1989 T. J. 268-271. Japan Acad. and geometry. Am. On elliptic modular surfaces. 1984 R. Mittag-Leffler (1963) pp. (1972) G. Proc. in Singular integrals. Tokyo 1973. Congr. SELMER. SHAFAREVICH. PropriCtCs galoisiennes des points d’ordre fini des courbes elliptiques. 199-208 G. Math. University of Tokyo Press. J. 31 (1963) pp. of Math. pp. Conf. Int. The diophantine equation ax3 + by3 + cz3 = 0. 6 (1989) pp. SHIODA. 186-196 T. Sot. SHIMURA. 179-230 J-P. Japan Acad. Vieweg. Duke Math. Algebraic cycles on certain K3 surfaces in characteristic p. SHIMURA. 8 (1967) pp. SHAFAREVICH and J. 20-59 T. No. 10 (1967) E. 259-331 J-P. SHIMURA. SHIODA. SHATZ. Math. Stud. Mordell-Weil lattices and Galois representations I. 300-303 T. 917-920 S. Ann. 85 (1951) 203-362 and 92 (1954) 191-197 J-P. The rank of elliptic curves. SERRE. The Galois representation of type Es arising from certain Mordell-Weil groups. 357364. II Ser. Transl. 296-299. Sot. SHIMURA. Sympos. Introduction to the arithmetic theory of automorphic functions. arithmetic. Transl. of Math. Sur les reprtsentations modulaires de degrt: 2 de Gal@/Q). Acta Math. Vieweg 1989 J-P. 1962. On elliptic curves with complex multiplication as factors of the Jacobian of modular function fields. 25-39. Nagoya Math. 88 (1968) pp. SHIODA. Ann. Springer-Verlag. Abelian l-a&c representations and elliptic curves. SERRE and J. 97 (1973) pp. Lectures on the Mordell-Weil theorem. SEELEY. Rational points. Tokyo (1975) T. 163-176. Algebraic number fields. 43 (1971) pp. SHIODA. Princeton University Press. 54 (1987) pp. 1971 G. Internat. Manifolds-Tokyo. SERRE. 65. Amer Math. 65 (1989) pp. Proc. Pure Math. 283-294 I. Collected Works. 67. Good reduction of abelian varieties. 21 (1989) pp. of Math. Yutaka Taniyama and his time.278 [Schapp 841 N. Ser A. Sot. SERRE. TATE. On modular forms of half-integral weight. II. Complex powers of an elliptic operator. SERRE. London Math. Proc. TATE. Math Sot. SHIODA. J. Bull. Proc.

SILVERMAN. Integer points and the rank of Thue elliptic curves. Math. Wiss. Defect relations for holomorphic maps between spaces of different dimensions. Math. 81 (1985) pp. Math. Representations of integers by binary forms and the rank of the Mordell-Weil group. ICM talk. J. Integral points on abelian varieties are widely spaced. J.I. 1990 T. Japan Acad. Stminaire Bourbaki. 1990) T. 555-565 J. 55 No 1 (1987) pp. Duke Math. 378 (1987) pp. 2 (1990) pp. Kl. 71-106 J. The arithmetic of elliptic curves. 197-211 J. Math. Math. 57-60 (accompanied by a preprint: Construction of elliptic curves with high rank via the invariants of the Weyl groups. Abh. Geometric d’Arakelov des surfaces arithmetiques. J. Nachr. SILVERMAN. Amer. 2 (1984) pp. Ser A. Compositio Math. Kyoto. Math. London Math. South.BIBLIOGRAPHY [Shio 90a] 279 [Shio 90b] [Shio 9Oc] [Sie 291 [Sie 691 [Slbg 851 [Sil 811 [Sil 821 [Sil 83a] [Sil 83b] [Sil 83~1 [Sil 84a] [Sil 84b] [Sil 851 [Sit 861 [Sil 871 [Sil 871 [Sil 871 [Siu 871 [So 891 [St 891 T. J. 1989-1989 G. SILVERBERG. SILVERMAN. A quantitative version of Siegel’s theorem: integral points on elliptic curves and Catalan curves. Stickelberger elements and modular parametrizations of elliptic curves. Construction of elliptic curves over Q(t) with high rank: a preview. 279 (1987) pp. STEVENS. Math. Arithmetic distance functions and height functions in diophantine geometry. L. 28 (1983) pp. Einige Anwendungen diophantischer Approximationen. reine angew. Math. SIEGEL. preprint. SILVERMAN. SILVERMAN. 342 (1983) pp. Invent. SILVERMAN. 107 (1985) pp. SILVERMAN. Invent. No. 6333648 J. L. SHIODA. Phys. Proc. Inuent. 395-403 J. SILVERMAN. l-7 J. The S-unit equation over function fields. Math. lnoent. 213251 C. 51 No. SHIODA. SIU. Abschatzung von Einheiten. SILVERMAN. Duke Math. 98 (1989) pp. . On the Mordell-Weil lattices. 60-100 J. Gottingen (1969) pp. 1990 C. SILVERMAN. SILVERMAN. J. Proc. Preuss. Integer points on curves of genus 1. 75-106 . 66. 253-266 J. reine angew. 74 (1983) pp. (1929) pp. SILVERMAN. Heights and the specialization map for families of abelian varieties. 3-4 J. Mordell-Weil groups of generic abelian varieties. Sot. Math. Akad. Sot. 71-86 A. 395-404 J. 61 (1987) pp. T. 193-216 Y. 66 (1982) pp. Lower bound for the canonical height on elliptic curves. Duke Math. Divisibility of the specialization map for families of elliptic curves. Wiss. SIEGEL. Theory of Mordell-Weil lattices. 281-292 J. 48 (1981) pp. 95 (1984) pp. Lower bounds for height functions. 41-69 C. Springer-Verlag 1986 J. J. Ann. Cambridge Philos. Akad. SHIODA.

TATE. TATE. Sur la dimension diophantienne des corps p-adiques. AstPrisque 86 (1981) L. Cong. pp. TATE. 516-518 L. TATE. TATE. pp. 306 J. C. Acta Arith. in Proc. Variation of the canonical height of a point depending on a parameter. SCminaire Bourbaki 1965-1966. 613-623 R. Sci. letters to Serre. On the conjectures of Birch and Swinnerton-Dyer and a geometric analogue. The higher dimensional cohomology groups of class field theory. Math. Math. Cambridge Tract in Mathematics 80.280 [SWD 521 BIBLIOGRAPHY CSZP 811 CSZP 841 [Szp 85a] [Szp 85b] [Ta 521 [Ta 581 [Ta 621 [Ta 651 [Ta 66a] [Ta 66b] [Ta 671 [Ta 68-79) [Ta 741 [Ta 831 [Ter 661 [Ter 773 [Tu 831 [Ue 731 [Ul 901 [vanD 811 [Vau 813 [Vio 851 A solution of A5 + B5 + C5 = D5 + E5 + F5. TATE. C. Acad. 294-297 J. Summer Conference at Arcata. 127-130 J. in Arithmetical algebraic geometry. Invent. 2 (1977) pp. No. Math. Springer-Verlag 1967 J. On Dyson’s lemma. Ann. Internat. 2 (1966) pp. 179-206 J. Classification of algebraic varieties I. Math 105 (1983) pp. ULMER. 93-111 J. Lefschetz Centennial Conference. The arithmetic of elliptic curves. pp. Sup. Cambridge University Press 1981 C. Duality theorems in Galois cohomology over number fields. SWINNERTON-DYER. Math. WC groups over p-adic fields. Math. Math. Pisa 12 (1985) pp. 48 (1952) pp. TATE. 58. Proc. VIOLA. 3 (1973) pp. 279-293 L. Paris 262 (1966) p. J. Invent. Algebra 73 (1981) pp. Contemp. 21 June 1968 and 12 October 1979 (appendix to the first letter) J. Math. Math. Contemp. 1958 J. Sot. TATE. 27 No. 105-135 P. TUNNELL. SCminaire Bourbaki. TATE. Proc. Scuola Norm. Ann. On universal elliptic curves over Igusa curves. Invent. A&risque 127 (1985) J. J. 1986. Invent. TERJANIAN. SZPIRO. 34 No. Algebraic cycles and poles of zeta functions. of Math. Amer. 72 (1983) pp. SZPIRO. TERJANIAN. Philos. UENO. Harper and Row 1963. 377-381 L. p-divisible groups. Cambridge . Endomorphisms of abelian varieties over finite fields. Un contrexemple g une conjecture d’Artin. Skminaire sur les pinceaux arithmittiques: La conjecture de Mordell. 99 (1990) pp. 287-294 G. Math. 612 G. The Hardy-Littlewood method. 251-260 L.R. 134-144 J. SZPIRO. 1985. A specialization theorem for p-adic power series converging on the closed unit disc. Stockholm (1962) pp. Comp. SCminaire sur les pinceaux de courbes de genre au moins deux. (1974) pp. A classical diophantine problem and modular forms. Small points and torsion points. van den DRIES. 56 (1952) pp. of Conference on Local fields (Driebergen). 67. Prksentation de la thCorie d’Arakelov. 277-342 D. VAUGHAN. SZPIRO. 288-295 J. TATE. 323-334 K.

On the rational points of abelian varieties over Z. The finiteness theorems of Faltings. On the secondmain theorem of Nevanlinna theory. M. 206. A lower bound for linear forms in logarithms. Ann. 37 (1980)pp. 375-484 A. Paris 1935 A. Bol. WEIL. J.Ann. WALDSCHMIDT. 257-283 J. K. in Lang. J. (1987) P. ljber die BestimmungDirichletscher Reihen durch Funktionalgleichungen. Introduction to Arakelou theory. SC. VOJTA. 128 (1954) pp. 168 (1967) pp. et Ind. Act. Invent. 412-444 A. Arithmetic discriminantsand quadratic points on curves. Math. Ann. Math. L. WEIL.Amer. 155-178 P. WEIL. Algebraic points on curves. Wiistholz. 29-39 J. Nachr. 98 (1989)pp. Springer Lecture Notes 1239. Mordell’s conjecture over function fields. Sur les critkres d’kquivalence en gtometrie algkbrique. 111 (1989) pp. Mat. Amer. WEIL. On the conjectures of Mordell and Lang in positive characteristic. Sur les coefficients de Fourier des formes modulairesde poids demi entier. 493-512 A. Faltings and G.R. Arithmetic on algebraic varieties. in Rational points by G. Acad. The conjectural arithmetic Riemann-Roch-HirzebruchGrothendieck theorem. 95-127 A. A height inequality for algebraicpoints. Math. 154-202 P. WEIL. VOLOCH. J. 60 (1981) pp. 149-156 L. WALDSPURGER. Dyson’s lemma for products of two curves of arbitrary appear M. J. L’arithmttique sur les courbesalgtbriques. Math. Math. Amer. 107-113 P. P. preprint Max Planck Institut 1991. Brasil. 53 3 (1951)pp. 1988pp.BIBLIOGRAPHY 281 [Vo 871 [Vo 881 [Vo 89a] [Vo 89b] [Vo 89~1 [Vo [Vo [Vo [Vo [Vol 90a] 90b] 9Oc] 90d] 851 [Vol 903 [Wa 803 [Walp 811 [We 281 [We 351 [We 51-j [We 541 [We 553 [We 571 [We 671 [Weng 911 [Wi 871 [Wo 891 [Wu 841 VOJTA. On algebraic groups and homogeneous spaces. 2 (1957)pp. P. VOLOCH.pp. Diophantine approximations and value distribution theory.Inuent. WENG. of Math. Vieweg Verlag 1984. VOJTA. submitted for publication in C. VOJTA.-extensions of number fields. Math. 33-53 A. J. 98 (1989)pp. WEIL. 77 (1955)pp. WONG. VOJTA. of Math. A refinement of Schmidt’s subspace theorem. with the title: Le thCorbme conjectural arithmktique de R-R-H-G. Math. Ann. Siegel’s theorem in the compact case. WINGBERG. ArithmCtique et gkomttrie sur les variCtts algbbriques. 279 (1987)pp. Acta Arithm. VOJTA. Acta Math. Sot. Pures Appl. WEIL. Giitt. 115-138 P. W~~STHOLZ. . Math. VOJTA. Diophantine inequalitiesand Arakelov theory. 549-583 G. No. Ann. Sci. 111 (1989) pp. Math. VOJTA. 9-24 P. P. 16 (1985) pp. 281-3151 A. Zum Beweis des Torellisschen Satzes. Hermann. VOJTA. Diagonal equationsover function fields. 489-518 P. Aspectsof Mathematics. 52 (1928)pp.

240244 J. ZACIER and G. 38 (1974). Mat. Notes 19 (1976) pp. 52 (1987)pp.translation: Math. 2 (1975)pp. 24 (1974)pp. Zametki 19 3 (1976)pp. Endomorphismsand torsion of abelian varieties. Math. USSR Sb. Numerical investigations related to the L-series of certain elliptic curves. Sb. Anal. Akad. translation: Fun&. 147 (1976)pp. Fibered algebraic surfaceswith low slope. Mat. Zametki 21 (1977) pp. J. 451-461 J. 39 (1975). 8 no. 399-410 XIAO GANG. Indian Math. Nauk SSSR Ser. On the difference of the Weil height and the NCronTate height. Akad. 8 (1974) pp. translation: Math.282 [Wu 85] G. Mat. Math. Zzv. USSR Zzv. W~~STHOLZ. Math. J. Endomorphisms abelian varieties over fields of finite of characteristic. A finitenesstheorem for unpolarized abelian varieties over number fields with prescribedplacesof bad reduction. Math. ZARHIN. translation: Math. 255-260 J. Mat. 461-470. 51-69 J. A finitenesstheorem for isogeniesof abelian varieties over function fields of finite characteristic. 95 (1974) pp. 309-321 J. 9 no. 31-34. Duke Math. Zzv. Appl. A new approach to Baker’s theorem on linear forms in logarithms I and II. 301-303 J. Prilozh. Sot. edited by A. 79 (1985)pp. ZARHIN. Cambridge University Press 1988. Abelian varieties in characteristic p. 8 (1974) pp. 393-400. 35-51 . Isogeniesof abelian varieties over fields of finite characteristic. BIBLIOGRAPHY [Wu 881 [Xi 873 [ZaK 871 [Zar 74a] [Zar 74b] [Zar 74~1 [Zar 751 [Zar 761 [Zar 771 [Zar 851 [Zar 871 [Zi 761 WUSTHOLZ. Nauk SSSR Ser. ZIMMER. 477-480 J. 737-744. USSR Zzv. Endomorphisms abelian varieties and points of finite of order in characteristic p. Notes 21 (1977)pp.translation: Math. 415-419 J. in New Advances in transcendence theory. Mat. ZARHIN. A remark on endomorphisms abelian varieties over of function fields of finite characteristic. 131-145 H. Ann 276 (1987)pp. A new approach to Baker’s theorem on linear forms in logarithms III. Z. Anal. pp. Znvent. Baker. Fun&. in Diophantine Approximation and Transcendence Theory Springer Lecture Notes 1290 (1985) pp. 449-466 D. 3 (1974) pp. ZARHIN. ZARHIN. 189-211 G. KRAMARZ. ZARHIN. 54 (1987)pp. ZARHIN. ZARHIN.

236-239 equationsfor 26. 118-121. 20. 178. 77 seealso Algebraic families. 74-82. 65 Abelian logarithm 237 Abelian varieties 16. Mordell-Weil. 198 cotangent bundle 181 vector sheaf 20 Analytic torsion 173. 79. 49. 63-67. 158-162. 67. 175 . 33 Algebraic families 12.Index A abc conjecture 29. 78. 62. coordmatesand integral points 217 Atline variety 2 Ahlfors on Nevanlinna theory 199. 688100. 15. LangN&on. Parshin.BirchSwinnerton-Dyer. Polarization. 189 of heights 76-82 of pseudo-canonical varieties 24. 23-27. I-adic representations. 118-121. 182 Algebraically hyperbolic 16. 20. Finiteness.Function field case. 22. N&on-Severi. Silverman-Tate Algebraic integers 54 Algebraic point 3 in Vojta’s conjecture 50. Jacobian. Moduli. N&on model. 189. Raynaud. 28.Gauss-Manin. 221 of abelian varieties 27. 198 anti-canonical class 19. 25 split 12. Manin’s method. 181-183. 28. Descent. 222-225 Algebraic specialset 16. 18. 32 Albert-Brauer-Hasse-Noether theorem 252-253 Algebraic equivalence 30. Function field case. 158-162. Semisimplicity.Tate property. Theorem of the kernel. 47. 221. 179 Ample 7. 62. 181. 74-82. 67 canonical class 14. 220. Torsion points. Manin-Zarhin. 187. Rank. 17. 203 Ahlfors-Schwarz lemma 185 Ahlfors-Shimizu height 200 Albanesevariety 31. 192 seealso Fibration and Generic fibration. 22. 20. 24. 232. 158-162. 178. MasserWustholz. 25. 101-122. 192. Subvarieties. Trace (Chow) Absolute case 12 Absolute conjecture 67 Absolute norm 55 Absolute value 44 Adams type for e 214 Adeles 93 Adjunction formula 168 Admissiblemetric 165 Affine bounded 207 Affine coordinate ring 4 . Faltings. 18. Semistable. 25-42. 11-15.

232 Bloch (Andre) conjecture 182 Bloch (Spencer)conjectures 34 Bogomolov 20. 230. 67. 179. 248 CC inequalities 151. 223 denominator 217 height: seeHeight. 228. upper bound Boundsfor generatorsof finitely generatedgroup 240-243 Brauer group 250-258 birational invariant 252 exact sequence 255 unramified 252 Brauer-Grothendieck group 250. 226 Brody’s theorem 179 Brody-Green hypersurface 22. 14. 259-262 Arakelov degree 168 height 169 inequality 151 metric 165 Picard group 167 Shafarevich conjecture in function field case 104 theory 71. 146. 240 Baker-Feldman inequality 235 Basic Hilbert subset 41 Basic isogenyproblem 121 Batyrev 20. 146.142 Bismut-Vasserot 174. 64. 169. 171 discriminant and Vojta’s inequality 171 Euler characteristic 173 Picard group 167 surface 166 Todd class 173 variety 171 Artin conductor 71 conjectures on Ci fields 246-247 Artin-Tate 250. 139140 Coates-Wiles 136 Gross-Zagier 139. 119. 91. 204. 23 Canonical height 63. 98. 136. 20. 64. 151. 152 Ci property 245-249 Canonical bundle 185 Canonical class 11. 166 . 21. 197 relative 146 zero 19-21. 252 specialization 252 Breen 80 Brody hyperbolic 178. 231 volume form 166 Arithmetic Chern character 173 Chow group 174 discriminant 64. 152 on a curve 11 on moduli space 119 on projective space 14. 63. 67. 20. 152. 181. 92. 258-262 canonical varieties 15. 222. 225.INDEX Anti canonical class 15. 239. 183. 137. 169 in Nevanlinna theory 202 Canonical metric 165. 94-96. 237. 151 Bombieri simplification of Vojta’s proof 233 Borel’s theorem 183 Bosch 259 Bounded degree 56. 19. 163-171. 253 Artin theorem on local specialization 259 Artin-Winters 150 Ax-Kochen theorem 247 Ax theorems one-parameter subgroups 182 quasi-algebraic closure 246 B Birch-Swinnerton-Dyer conjecture 34. 262 Batyrev-Manin conjectures 260262 relation with Vojta conjecture 261 Beilensonconjectures 34 Biduality 33 Birational map 5 Birationally equivalent 5 Birch 148. 184. 186 Brownawell-Masser 66 Brumer-McGuinness on averagerank 28 Bryuno 214 C Baily-Bore1 compactification 118 Baker 235. 197 in higher dimension 14 inequalities 151. 245.

82 Chai’s theorem of the kernel 158 Chai-Faltings 118. 160 Nevanlinna theory 204 Parshin construction 104. 202. 24 de Rham cohomology 153-160 Decompositiongroup 83. 145. 136 Complex torus 125 isomorphism classes 126 Complexity of divisor 203 Conductor 51.106 ramified 104105. PO 128 Cubic forms 22. 256. 228. 202 Cherry on Nevanlinna theory 204. 103 Chow trace 26 see also Lang-Neron. 169 Canonical variety 15 Carlsson-Griffiths on Nevanlinna theory 201 Cartan’s theorem 197-198 Cartan-Nevanlinna height 194 Cartier divisor 6. 73. 14 divisor class group 6 Cassel-Guy example 254 Cassels-Tate pairing 89 Chabauty 36. 61. 211 Coverings applied to diophantine approximations 219. 257 Chbtelet-Weil group 87. 38 Chai on cubical sheaves 80. 97-99 of elliptic curve 97-99 seealso Modular elliptic curves d(F) or d(P) 55 Davenport 48. 71. 146. 81 Curvature 185. 169. Polarization Deligne on Faltings proof 120 . 33 Neron model 70. 98 Connection 154 Conormal sheaf 145. 118 Coates-Wiles theorem 136 Cocycle 153 Coleman’saccount of Manin’s method 153-161 Colliot-Thelene 249-258 conjecture 254 Compact caseof Vojta conjecture 67 Completeintersection 4 Complex multiplication 29. 254 Cyclotomic character 133 extensions 29. 259 of an imbedding 145. 260-262 units 57 Counting function 193. 170. 34 Co-Lie determinant 116. 146. 32. 122 Characteristic polynomial 85 Chltelet Brauer variety 256 surface 22. 120. 247 de Franchis theorem 13. 39. 247. 85. 223 Chevalley’s theorem on quasi-algebraic closure 246 Chevalley-Weil theorem 224 Chow group 7. 33. 79-81. 186 Cuspidalgroup 128 ideal 129 Cusps 127 Cycles 31. 247 units 141 D Canonical sheaf 119. 224. 88 Chern form 184. Trace of an abelian variety Circle method 248 Classfield theory 105 Clemens curves 21. 88. 250 Cubical sheaves 80. 174 in higher codimension 34 Chow-Lang 69.(N) by X. 112 Degree Arakelov 168 of canonical sheaf 152 of divisor on a curve 9 of hypersurface 4 of isogeny 35 of line sheaf 144 of metrized line sheaf 116 of polarization 35 on a singular curve 144 with respectto Riemannform 238 seealso Isogeny.INDEX 285 Congruent numbers 135 Connectedcomponent 31. 225 torsion points 39 X. 49. 222 descentand heights 37. 169 Constant field 12 Counting algebraicintegers 57 points 58.

286 Deligne (continued) on L-function 92 Deligne-Mumford 150 Deligne-Serre representation 132 Demjanenko estimate of Nbon height 72 fibering of Fermat surface 23 Demjanenko-Manin on split function field case 79 Descent 85, 90, 91, 160, 219 in coverings 160, 219 with Selmer groups 90 Determinant of vector sheaf 144 Diagonal hyperplane 183 Different 120 Differential form 10 Differentials 115 Differentials of first kind 11, 136 Dimension 4 Diophantine approximation on toruses 233-243 Diophantine approximation to numbers 213-216 Dirichlet box principle 234 Discriminant 50, 55, 69 in coverings 224 of elliptic curve 69, 96 Distance 177 Division group 37, 38, 161 see also Hindry, Raynaud and Voloch theorems Divisor 6 Divisor classgroup 6, 33 Divisor classes and heights 58-61, 194-196 and NCron functions 213 Dobrowolski 243 Dolbeault operator 172 Dual variety 33 Dyson’s lemma 229, 231, 232 E Effective divisor 6 Effective divisor class 7 Eigenform 131, 132 Eigenvalues of Frobenius 85, 91, 131 of Laplacian 173, 175 Eisensteinideal 129 quotient 129 series 260 Elkies exampleof integral points 50 Elkies on Fermat hypersurface 23


Elliptic curve 12, 21, 23, 25-27, 49, 50, 96, 132, 135, 139-142, 162 conductor 97, 98 diophantine approximation 235, 236 fibers of family 21, 23 Frey 132, 135 height of generators 99, 100 integral points 50 isomorphism 96 L-function 97-98, 137-142 minimal discriminant 97 modular 130- 142 periods 93, 95, 97, 125,236 rank 28,42, 92, 139-142 rank one over the rationals 137- 142 Tate curve 162 seealso Torsion points Error function in Nevanlinna theory 203, 224 Error terms in secondmain theorem 199-204, 224 Lang conjecture 200 Esnault-Viehweg inequality 152 on Roth theorem 229 Euler characteristic 172 arithmetic 173 Euler on Fermat 22 Euler product for a variety 91 Exact sequence de Rham cohomology 156 group cohomology 88, 250, 255 Lang-Tate 87 Selmerand Shafarevich-Tate group 88-90 Exceptional set in Vojta conjecture 67 in Schmidt-Vojta theorem 215, 222 Exponential on Lie groups 236 F Faltings canonical height 119 finitenessof I-adic representations 114 finitenessof rational points 12, 18, 36, 230, 232 formula for the degree 120 height 74, 117-123, 238 inequality on abelian varieties 220, 237 integral points on abelian varieties 220, 221


287 I-adic representations 112 polarizations 103 principally polarized abelian varieties 117 rational points 12, 13, 37-39, 130, 187 rational points by Parshin method in function field case 187 rational points in division groups 37-38 rational points in Eisenstein quotient 130 rational points on modular curves 130 rational points on toruses 37-38 Shafarevich-Tate group 89, 253 see also Finitely generatedgroups, Mordell-Weil, Shafarevich conjecture,Torus, Unit equation First kind 11, 136 First main theorem 194 Forms in 10 variables 247 Fourier coefficients 130 Franke 260-262 Frey polynomial 51 Frey’s idea for Fermat 132, 134, 135 Frobenius automorphism 84, 85, 95, 113, 114, 132, 133 isogeny 84 Fujimoto 183, 184 Function field 4, 12 Function field case 12, 18, 23, 24, 27, 28, 39, 45-47, 62, 74, 76-82, 92, 104, 145, 178, 187, 192,221 abelianvarieties 39, 187 Birch-Swinnerton-Dyer 92 Mordell conjecture 62, 143-162, 230 product formula 52 quadratic form 74 Shafarevichconjecture 104 torsion 27 Functional equation 98 G g2 and g3 12, 126 Galois groups 39, 42 of torsion points 39, 83, 132,238 Galois representations 39, 83, 132 Gauss-Manin connection 154- 161 on abelian varieties 158 Gelfond 235 General position 183,215

positivity of canonical sheaf 170 semisimplicity and Tate conjecture 111-115 stable height 117, 238 subvariety of abelian variety 36 Fano variety 260 Fermat 11, 22, 23, 48, 62, 64, 132, 135, 181, 225 Brody-Green perturbation 181 curve 11 Euler 22 fibrations 23 Frey elliptic curve 132, 135 hypersurface 22, 23 modular curve correspondence 23, 225 Ribet’s reduction of last theorem to Taniyama-Shimura 132 Taniyama-Shimura implies Fermat 132 theorem for polynomials 48 unirational for low degree? 22, 23 Vojta’s conjecture implies Fermat asymptotic 64 Fibration 18, 20-24, 35, 36, 183, 256, 257 by tonics 253, 255, 256, 257 of Chltelet surface 256, 257 of Fermat 23 of generic quintic threefold 21 of K3 surface 20 of Kummer surface 20 of subvarieties of abelian varieties 35, 36, 183 see also Algebraic families and split algebraic family Finite representation at a prime 134 Finitely generated extensions 12, 15, 16, 27, 33, 42, 111, 112 group 26-32, 36-40, 240-243 Finiteness I 106 Finiteness II 107 Finiteness Brauer group 254 Chow group generators 34 curves with good reduction 104 Faltings heights 120 integral points 50, 217, 221 isogenies 106-109, 115, 120, 121, 122, 128, 238, 239 isomorphism classes of abelian varieties 106, 107, 111, 113, 115, 117 isomorphism classes of curves 103



General (type) variety 15, 17 Generalized Jacobian 106 Generalized Szpiro conjecture 51 Generic complete intersection 18 1 hypersurface 21, 181 quintic threefold 21 Generic fibration 18, 20-23 Chatelet surface 257 Fermat 22, 23 K3 surface 20 Kummer surface 20 quintic threefold 21 Generically surjective 5, 24 Genus 10 Genus formula in terms of degree 11 Geometric canonical height 146 conditions for diophantine bounds 176 fiber 149 genus 14, 15 logarithmic discriminant 146 Gillet-Soule 1733175, 230, 232 theorem 174 theory 173-175, 230, 232 Global degree 53, 168 Goldfeld on rank 28 Good completion of N&on model 80,

Hall conjecture 49 Harder 255 Hardy-Littlewood circle method 248 Hartshorne conjecture 20 Hasse conjecture 98 eigenvalues of Frobenius 85 Hasse principle 248-258 hyperbolicity connection 254 non-singular 248 HasseeDeuring l-adic representations 83 Heath-Brown 247 Hecke algebra 129 correspondence 129 involution 129, 139 operators 130 Heegner point 138-141 Height 43, 51, 53-67, 70, 72-82, 85, 86, 99, 100, 117-123, 146, 152, 153, 168, 169, 171, 193-200, 202, 203, 222, 224, 227, 233-243, 252, 260 algebraic families 62, 74, 76-82 algebraically equivalent to zero 59 Arakelov degree 168, 169 as a norm 73, 85, 86, 236, 241, 242 associated with divisor class or line sheaf 58-61, 1944196 bounded 56-58, 62-67, 99, 233 canonical 63, 64, 67, 146, 169, 202 canonical coordinates on abelian varieties 77 Cartan-Nevanlinna 193-198 Faltings 74, 117- 123, 227, 238 finite extensions 51 inequalities: see below upper and lower bounds intersection numbers 169 Lie 119 lower bound 73, 74, 100 Nevanlinna theory 1933200, 202 normalized 260 pairing 12-76 transform 203 upper bound 62267, 99, 100, 152, 153, 160, 170, 171, 222, 224, 227 see also Birch-Swinnerton-Dyer, Modular elliptic curves, Properties of Heights, Regulator Hensel’s lemma 249 Hermite theorem 99, 114

reduction 68-70, 91, 103-106, 113 Grauert’s construction 147-149 Green and Fujimoto theorem 183 Green example of Brody but not Kobayashi hyperbolic 183 Green function 164, 165, 209 Green-Griffiths 20, 179, 180, 182, 186 Bloch conjecture 182 conjecture 179, 180 Greenberg functor 259 Greenleaf theorem 248 Griffiths complement of a large divisor 226 function 185 Griffiths-King on Nevanlinna theory 201, 204 Gross on Birch-Swinnerton-Dyer 9496 GrosssZagier theorems 139- 142 Grothendieck on semistable reduction 70 Group variety 16, 20, 67, 2555258 see also Abelian varieties, Toruses Good


289 complement of 2n + 1 hyperplanesin generalposition 221 Faltings theorem on abelian varieties 220 function field case 221 higher dimensionalfunction field case 189,221 and hyperbolicity 22, 189, 226 Integralizable 217 Intersection 144, 167,209 number 144, 167 pairing 144 theory and Weil functions 209 Involution on modular curve 129 Iskovskih surface 253 Isogeny 28, 34, 35, 108, 113, 121, 122, 128,238-239 boundson degreeby Kamienny 28 boundsby Masser-Wustholz 121, 122,238, 239 boundsby Mazur 128 problem 121 theorem 108 Isomorphismclasses toruses 107, of 126 Iwasawatheory 30

Hermitian manifold 177 vector sheaf 166- 168 Hilbert irreducibility 40-42 application 42, 162 Hilbert subset 41 Hilbertian 41 Hindry theorem 37, 38 Hindry-Silverman lower bound on height 74 Hirata-Kohno theorem 237 Hodge index 168 Holomorphic special set 179, 182 Homomorphisms of abelian varieties 26 Hooley 247 Horizontal differentiation 156 Humbert 103 Hurwitz genus formula 37, 105 Hyperbolic 16, 17, 25, 177-192, 225-228, 254 algebraicity 16, 17, 179 Brody 179 complement of a large divisor 226 Hasse principle connection 254 hypersurfaces 180 Kobayashi 178-181, 184 metric on disc 177 Mordellic connection 25, 179, 186 Parshin’s method 187 Hyperbolically imbedded 190-192, 225 Hyperbolicity ampleness connection 181 integral points connection 225-227 Hyperplanes in projective space, complementary set 183 Hypersurface 4 Hypersurface generic 21 I Ideal class group 55 Igusa zeta function 250 Imbedding 5 Index formulas 141 of Heegner point 140 in Schneider-Roth theorem 229 in Vojta’s theorem 231 Inertia group 84, 112 Infinite descent 85 Integral points 22, 189, 217-222, 226

Jacobian 32, 102, 103 Jensen’s formula 193 Jouanolou theorem 148

K3 surface 20, 23 Kamienny on torsion 28 Kanevsky 250 Kato-Kazumaki conjecture 247 Kawamata fibration 36 Bloch conjecture 182 structure theorem 35, 182 Khintchine function 198, 199, 223 theorem 213, 214, 234 Kneser 255 Kobayashi chain 178 hyperbolic 178, 184 semidistance 178, 186 Kobayashi conjecture on hyperbolic hypersurfaces 180 hyperbolicity 177-192, 225, 226 hyperbolicity and (1, l)-forms 185, 186



Kobayashi conjecture on hyperbolic hypersurfaces (continued) hyperbolicity and pseudoample cotangent bundle 181 theorem on ample cotangent bundle 181 Kobayashi-Ochiai 24, 25, 192 Kodaira criterion for pseudoample 9, 20 Kodaira-Spencer map 157-161 Koizumi-Shimura theorem 113 Kollar 20 Kolyvagin theorem 89, 139 Kubert on torsion points 28 Kubert-Lang 37, 141,225 Kummer surface 20

I-adic representation 82-85, 107-l 15, 238 L-function abelianvariety 9l-98 elliptic curve 97, 98, 139-142 local factor 95, 97 seealso Birch-Swinnerton-Dyer L2-degree 173 Lander and Parkin 23 Lang error term in Nevanlinna theory 199-201, 203 on integral points 218 theorem over finite fields 87 Lang conjectures Ax theorem 182 bound for regulator and ShafarevichTate 99 diophantine 15-20 diophantine approximation 233237 division points 37 exceptional set in Vojta 67 Fermat unirationality 22 finitely generatedgroups 36 function field case 13 Greenleaftheorem 248 Green’stheorem 182 hyperbolic imbedding 190, 191 hyperbolicity 17, 179, 181, 186 integral points 50, 220, 225 lower bound on height 73, 74, 100, 243 Mordellic property 15, 16, 25, 36, 179

pseudoMordellic 17, 180, 181 reduction to ordinary abelian varieties 162 upper bound on height 99 Lang-NCron theorems 27, 32, 74, 75 and theorem of the kernel 159 Lang-Stark conjecture 50 Lang-Tate exact sequence principal for homogeneous spaces 87 Lange on polarizations 103 Langlands-Eisenstein series 260 Laplace operator 173 Lattice points in expanding domain 57 Laurent theorem 38 Lehmer conjecture 243 Level N structure 118 Level of modular form 130, 131, 133135 of representation 133-135 Lewis theorem on forms in 10 variables 247 Liardet theorem 37 Lie determinant 116 Lie height 119 Linear group varieties 256 Linear torus 37 Linearly equivalent 6 Lipschitz parametrizable 57 Local completeintersection 145 degree 53 diophantine conditions 176 exact sequences 251 factor of L-function 95 parameter 10 ring 5 specializationprinciple 258-259 Locally bounded 207 Logarithm on Lie groups 234-239 Logarithmic discriminant 55, 146 height 43 Lower bound conjectures 74, 100,243 Lu-Yau 180

Maehara 24 Mahler 58, 217, 220 Manin Brauer group 253 constant 140, 141 counting 260-262 cubic surfaces 23, 48


Mordellic 15, 16, 25, 36, 179 Moret-Bailly 170 Mori on Hartshorne conjecture 20 proof of Ueno’s theorem 35 theoremson rational curves 19, 20 Mori-Mukai 20 Multiplicative height 54 Mumford 20, 23, 26, 61, 62 equationsfor abelian varieties 26 gapsbetweenheights of points 61, 62 N Narasimhan-Nori theorem 103 Ntron algebraicfamiliesof N&on functions 213 function 210, 212 model 19, 69-71, 79-82, 94, 95, 115-117, 120 pairing 212 rank 41, 42 specializationtheorem 41 symbol 212 theorem on Mordell-Weil 26 N&on-Severi group 30, 32-34, 77, 79, 149, 261 N&on-Tate height 72-75, 82, 85, 86, 241, 242 and Weil height 72 estimates Demjanenko 72 by estimates Zimmer 72 by N&on-Tate norm 73 quadratic form 72-75 Nevanlinna theory 192-204 for coverings 204 Newton approximation 249, 259 Noether and Galois groups 42 Noether’sformula 152, 168 Noguchi 36, 183, 184, 190-192 Non-degenerate 202 Non-singular 4 Hasseprinciple 248-258 rational point as birational invariant 249 Norm form 245 Norm as height 73, 85, 86, 236, 241, 242 Normal crossings 191 Normalized differential of first kind 136 Normalized theta function 209

elliptic curves 253 function field case of Mordell 13, 37, 153-161 letter 158 obstruction 250, 253-258 unirationality 23 Manin-Mumford conjecture 37, 38 Manin-Zarhin equations for abelian varieties 26, 77 height with canonical coordinates 77 Mason theorem 48, 65 in several variables 66 Masser lower bound on height 74, 240, 243 Masser-Oesterle abc conjecture 48 Masser-Wustholz theorem 121, 238239 replacementof Raynaud theory 122 May’s theorem 58 Mazur Eisenstein quotient 129 points in cyclotomic extensions 29 torsion group 28, 127-130, 134 Measure hyperbolic 186 Mestre 28, 170 Metrized vector sheaf 167 Minimal discriminant 73, 97, 134 height 73, 74 height conjecture 100 model 97 N&on differential 140 Miranda-Persson on torsion 27 Miyaoka 151 Miyaoka-Mori 20 Modular elliptic curve 132, 136, 1388142 representation 133, 134 units 37, 141 Moduli space 118, 119 Mordell conjecture 12, 106 Faltings proof 107-121 function field case 13, 143-162 Vojta’s proof 230 Mordell objection to Riemann-Roth 230 Mordell-Weil group and units 142 in abelian extensions 29 Shioda lattice 75 theorem 26, 27 seealso Rank, Specialization,Torsion points

42. MasserWustholz. 118. 119. 118.236. 256 Principal polarization 102. 121 Product formula 45. 33 variety 33 Picard-Fuchs group 157. 1799181 hyperbolic 180-181 Mordellic 17. 35. 179. 170 hyperbolic method 149. 121 seealso Humbert. 162 Ordinary absolutevalue 44 Osgood 216. 221 method with canonical sheaf 149 proof of Raynaud theorem in function field case 189 Shafarevichimplies Mordell 104. 239 lattice 125 relations 239 v-adic 93 Pfaffian divisor 148 Pit(X) 6. 67. 239 degree 35. 180-181 Pseudoample 9. 52 Projective bundle 147 variety 2 Proper set of absolutevalues 53. 257. 122. 159 Poincare class 33 Polarization and polarized abelian variety 34. 121. 34 R Ramanujan’s taxicab point 23 Ramification counting function 197. 97. 36. 95. 119. Lange. 259 Quasi-projective variety 3 Quintic threefold 21. 121. 19. 103.244. 1)-form 185 Positive cone in N&on-Severi group 261 Positivity of canonical sheaf 170 of Weil functions 208 Power series 247 Principal homogeneous spaces 85-91. Moduli space. 92. 35.292 Northcott theorem Number field 12 0 56 INDEX Ochiai on Bloch conjecture 182 on Ueno-Kawamata fibrations 36 see also Kobayashi-Ochiai Ogg on bad reduction 71. 103. 33. 238. 103. 103. 119. 180. 181. 187-189 inequality 170 integral points in function field case 189. 102. 139-142 . 105 Rank average 28 cyclotomic extensions 29 Demjanenko-Manin criterion 79 elliptic curve 28. 221 P p-adic absolutevalue 44 Parshin construction 104. 144 Picard group 6. 17. 239 principal 102. 179-181 Pythagorean triples 135 Q Quadratic form. 35. 258 anti-canonical class 244. 198 Kodaira condition 9 Pseudofication 15. 105. 105 Parshin-Arakelov proof of Shafarevich conjecture in function field case 104 Peck 248 Period 93. 98 One-parametersubgroup 183 Order at p 44 Order of a function at a divisor 6 Order of the conductor 71 Ordinary abelian variety 161.238.201204 divisor 196. 67. 199. 211 Pseudo canonical variety 15. Torelli Polynomial equations 3 Positive (1. 58 Propertiesof height in Nevanlinna theory 194-196 in number theory 58-61 Proximity function 193.202 order 196. 198. 125. seeN&on-Tate Quadratic forms in 9 variables 249 Quasi function 207 Quasi-algebraicclosure 245-248. 122. 258 canonical class 15.

92. 237 N&on-Severi group 261 rank 1 over the rationals 138 141 see also Birch-Swinnerton-Dyer. 121. 39 Noguchi theorem 183 Semisimplicity of I-adic representations 107-111. 216. Gillet-Soule Robba 259 Rosenlicht 106 Roth theorem 215 geometric version 2 18 Rubin on Shafarevich-Tate group 89 RuWong theorem 221 S S-integers 214 Saito. 260. I-adic representation Residue class field 3 Restriction of scalars 245 Ribet Galois representations for Fermat 132 theorem on Fermat 134 torsion group in cyclotomic fields 29 Ricci form 185. T. 170 Faltings height 119 formula for the degree of the Lie sheaf 120 function field case 39 Parshin’s proof 189 theorems 37 torsion and division points 37. 71 Saito. 18820. 230. 202 Richtmayer-Devaney-Metropolis 214 Riemann form 209 Riemann-Roth 173-175. 38. 98 of Mordell-Weil group 93. 255 Samuel proof in characteristic p 161 Sansuc 249-258 linear group varieties 256 Schanuel counting 262 theorem 58. 149-151 finitely generated group 240-243 generic case 27 high rank by N&on specialization 42 Mestre 28 Mordell-Weil group 28. 254 Salberger 249. 214. 71 curve 150 reduction 70. S. 20. 16.INDEX 293 see also Galois representation. 256 Rationally equivalent 6. 7 Raynaud bad reduction 71 conductor 11 division points 37. 260 Schinzel conjecture 257 Schmidt theorem 215. Brumer-McGuinness. 234 Schneider method 229 Second main theorem 196-204 Selmer example 89 group 88-91 Semiabelian variety 36. 38. N&on. 37. 244. 98 Relations 242 Relative canonical class 146 cohomology group 154 GausssManin connection 155 tangent sheaf 173 Relative case 12 Relatively algebraically closed 258 Representation finite at a prime 134 ’ . 71. 232. 170 Reduction homomorphism 84 and Voloch theorem 162 Reduction modulo a prime ideal 68 Regular differential form 14 Regulator 55.113-115 Semisimplification 133 Semistable abelian variety 70. 117. 93. 222. Zagierr Kramarz Rational curves 16. 23. 250. Shafarevich-Tate. 233 in Roth theorem 214 in Vojta’s proof 230 objection by Mordell 230 see also Bombieri. 247 differential form 10 function 4 group variety 16 map 5 point 2 point on Chatelet surface 257 points in completions 249 variety 6. 120. 139142. 22.

178. 106. 237 Serre-Tate theorem 113 Severi-Brauer 256 Shafarevich coniecture 104. 259 Split algebraic family 12. 135.294 INDEX Serre conjecture for Fermat 134 I-adic representations 134. 218 Siegel lemma 232 Sign of functional equation 98. 94 Torelli’s theorem 102. 220. 103 Torsion points 27-29. 253-255 Shatz 250. 136 on Taniyama 131 Shioda on generic torsion points 27 on lattices from Mordell-WeilLang-N&on groups 75 Siegel 37. 35. 139 Silverberg on generic torsion points 27 Silverman theorem on heights in algebraic families 78. 111. 232 Sp (Special Set) questions 18. 36. 111. 109. 253 Shimura correspondence 141 Shimura on modular elliptic curves 132. 138 Taniyama-Shimura implies Fermat 134 Tate duality of cohomology over p-adic fields 87 module 82 property 107. 99. 134.(A) 82. 24. 202. 96. 79 conjecture on algebraic families of heights 81 Silverman-Tate theorem 77 Simple normal crossings 196. 98. 232 on integral points 217. 173-175. ShafarevichTate. 82-85. 67. 217. 115 theorem on algebraic families of heights 81. 79. 17-23. 111 implied by Vojta conjecture 227 imolies Mordell 106 Shafarevich-Tate exact sequence 88 example of high rank in function field case 28. 245 Special set 16. 136. 228. 232 ?. 127-130. 159 Theta divisor 102 Theta functions 209 Thue-Siegel theorem 220. 42. 82 see also Lang-Tate. 94. 112. 138. 257 on Fermat 22 see also Birch-Swinnerton-Dyer Szpiro conjecture 51 positivity of canonical sheaf 170 Raynaud theorem connection 170 T Tangent sheaf 173 Taniyama conjectures 131 Taniyama-Shimura conjecture 131. 161 Stark 50. 79. 182. 188 Stably split 159. 228. 220 Green’s theorem 182 special set 36 Ax theorem 182 Sum formula 45 Support 7 Swan conductor 71 Swinnerton-Dyer 249. 78. 179. Silverman-Tate Terjanian example 247 Theorem of the kernel 158. 58. 21. 181-183. 140. 84 on Brauer group 252 on sections 40 Specialization principle (local) 258. 92 group 88-91. 218. 135 torsion points 39. 238 . 237 local L-factors 92 semisimplicity 111. 62. 192 Stability 150 Stable Faltings height 117 Stably non-split and finiteness of rational points 187. 240 Stevens conjecture 141 Stall on Nevanlinna theory 204 Strongly hyperbolic 185 Subvariety 3 Subvarieties of abelian varieties 20. 223 Siu 198 Soult: 168. 220. 245 and exceptional set 67 holomorphic 179 holomorphic and algebraic are equal 182 Special variety 17-20 Specialization map and homomorphism on abelian varieties 41. 25. 39. 78. 232 Faltings theorem 36. 139.

223 Voloch division points in characteristicp 161. 152. 231.Semiabelian varieties Totally geodesic 189 Trace of an abelian variety (Chow) 26. 20. 63. 171. 16.INDEX 295 diophantine approximation 238 function field case 27 Galois group 39 I-adic representations 82-85 uniformity conjecture 28 seealso Kamienny. 64. 226. 216 U Vector sheaf 144 Very ample 7 Very canonical 15 Viola 229. 227 compact case 67 exceptional set 67. 152 inequality with arithmetic discriminant 171 integral points 226 proof of Faltings’ theorem(Mordell conjecture) 230-232 Vojta’s conjectures 50. 114. 71. 113. 215. 222. 222 inequality and Nevanlinna theory 204 inequality in function field case 147. 257 Uniruled 20 Unit equation 37. 183 Ulmer on L-function 92 Unigrouped 20 Unipotent group 71 Unirational 6. 66. 215. 230-232 (1. 64. 36 Ueno-Kawamata fibration 36. Masser-Wustholz. Mazur. generalized Chevalley-Weil 224 estimates sections 174-175 for improvement of Cartan’stheorem 197-198 improvement of Schmidt theorem 215. 1)-form conjectureand Shafarevichconjecture 226. 197-198. Kubert. 193-195. 1)-form theorem 201 dictionary 193-195 estimatefor discriminants. 172. 251 good reduction 113 representation 112 seealso Coverings Upper half plane 124 V (1. 204. 216. 233-239 seealso Abelian varieties. 233 V&4) 83 Vojta theorems 147. 67. 216. 216. Miranda-Persson Torus 37. 222 Fermat curve 64 general 222-224 higher dimension 66 imply abc conjecture 64 relation to Batyrev-Manin 261 uniformity with respectto the degree 63. 185 W Valuation 44 van de Ven 151 van den Dries 259 Variety 2 Waldschmidt 241 Waldspurger’s theorem 137 Weierstrass functions 25 Weight 3/2 136. 132 Translation on N&on model 80 Tschinkel 260-262 Tsen’stheorem 246 Tsfasman 250 Tunnel1on congruent numbers 13% 137 Twisted elliptic curve 139 Type for a number 213. 226-227 Ueno fibration 35. 244. 22. 141 of a modular form 136 Weil algebraicequivalencecriteria 33 divisor 6 . 204. 74 Trace of Frobenius 97. 216 Type of meromorphic function 200. 201. 222-224. 174175. 162 unit equation 66 Volume 166. 38. 220 Units counting 57 Unramified Brauer group 252 Chevalley-Weil theorem 224 correspondence betweenFermat and modular curves 225 extension 247. 219.

.Weil (continued) eigenvalues Frobenius 85. 128 Y Yau 151 Y. 255 Zeta function 56. 122 semisimplicityand Tate conjecture 109. of 249 function 164. 173. 93-98 as Eisenstein series 260 of elliptic curve 97. 254. 209.(N) X.(N) Z 127 Zagier-Kramarz on rank 28 Zarhin points in abelian extensions 29 principal polarization 119. 210 function as intersectionnumber 209 function associated with a hyperplane 216 height 45 height as sum of local Weil functions 210 I-adic representations 83 Weng’scommentson Gillet-SoulC 174 Wild ramification 71 Wong on integral points 221 on Nevanlinna theory 199. 112 Zariski topology 3 Zero cycle 31.201. 93-98. 192. 139-142 of Laplace operator 173 of number field 56 with heightsof points 260 Zimmer 72 .(N) 127-131 127. 203 Wronskian 197 Wronskian method in Roth theorem 229 Wustholz on Baker inequality 235 x X. 114. 91. 98. 260 of abelian variety 91. 198. 121.(N) and Y. 32.207.

Sign up to vote on this title
UsefulNot useful