0% found this document useful (0 votes)
175 views

Shimura G. - Automorphic Functions and Number Theory (1968) PDF

This document contains lecture notes from a talk given by Goro Shimura at the Forschungsinstitut für Mathematik in Zurich in 1967. The notes provide a short, non-technical overview of recent results in the field of automorphic functions and number theory, with indications of basic ideas but no detailed proofs. References are provided for readers seeking a more complete presentation of the material.

Uploaded by

Paky Pakic
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
175 views

Shimura G. - Automorphic Functions and Number Theory (1968) PDF

This document contains lecture notes from a talk given by Goro Shimura at the Forschungsinstitut für Mathematik in Zurich in 1967. The notes provide a short, non-technical overview of recent results in the field of automorphic functions and number theory, with indications of basic ideas but no detailed proofs. References are provided for readers seeking a more complete presentation of the material.

Uploaded by

Paky Pakic
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 38

Lecture Notes in Mathematics

A collection of informal reports and seminars


Edited by A. Dold, Heidelberg and 6. Eckmann, Zurich

Series: Forschungsinstitut f i r Mathematik, ETH, Zurich . Adviser: K. Chandrasekharar

Goro Shimura
Princeton University, Princeton, New Jersey

Automorphic Functions
and Number Theory

Springer-Verlag Berlin Heidelberg New York


Preface

These notes a r e based on l e c t u r e s which I gave a t the


Forschungsinstitut f a r Mathematik, Eidgenossische Technische
Hochschule, Ziirich in July 1967. I have attempted to make a
s h o r t comprehensible account of the l a t e s t r e s u l t s in the field,
together with an exposition of the m a t e r i a l of an e l e m e n t a r y nature.
No detailed proofs a r e given, but t h e r e i s a n indication of b a s i c ideas
involved. Occasionally even t h e definition of fundamental concepts
m a y b e somewhat vague. I hope that this procedure will not bother
the r e a d e r . Some r e f e r e n c e s a r e collected in the final section in
o r d e r to overcome these shortcomings. The r e a d e r will be able to
find in them a m o r e complete presentation of the r e s u l t s given h e r e ,
with the exception of s o m e r e s u l t s of §lo, which I intend to d i s c u s s
in detail in a future publication.
It is m y pleasure to e x p r e s s m y thanks t o P r o f e s s o r s K.
Chandrasekharan and B. Eckmann for their i n t e r e s t in this w o r k ,
and f o r inviting m e to publish it in the Springer L e c t u r e Notes in
Mathematics. I wish a l s o acknowledge the support of the
Eidgenossische Technische Hochschule, Institute for Advanced Study,
and the National Science Foundation (NSF-GP 7444, 5803) during the
s u m m e r and f a l l of 1967.

Princeton, January 1968 G. Shimura

All rights re\crued. N o part uf this book m q be translated or reproduced in any form without wrincn permission from
Springer Veriag. 0 by Springer-Vdag Berlin. Hddelberg 1%8
.I,ibmty of Congress Catalog G r d Number 68-2>132. Printed in Germany. Title No. 7374
Contents Notation

We denote by 2 , Q, R and C respectively the ring of rational


i n t e g e r s , the rational number field, the r e a l number field and the
1 complex number field. F o r a n associative ring Y with identity ele-
Introduction
X
ment, Y denotes the group of invertible elements in Y, M (Y) the
Automorphic functions on the upper half plane, especially modular n
r i n g of a l l m a t r i c e s of s i z e n with e n t r i e s in Y, and GLn(Y) the
functions
group of invertible elements in M (Y), i. e. , Mn(Y)X . The identity
Elliptic c u r v e s and the fundamental t h e o r e m s of the c l a s s i c a l n
element of M (Y) i s denoted by 1 and the t r a n s p o s e of a n element
t h e o r y of complex multiplication n n '
t
A of Mn(Y) by A a s usual. When Y i s commutative, SLn(Y)
Relation between the points of finite o r d e r on an elliptic curve
denotes the group of a l l elements of M (Y) of determinant 1. F o r a
and the modular functions of higher level n
Abelian v a r i e t i e s and Siege1 modular functions typographical reason, the quotient of a space S by a group G will be
denoted by S/ G, even if G a c t s on the left of S. If F i s a field
The endomorphism ring of an abelian variety; the field of moduli
and x i s a point in a n affine (resp. a projective) space, then F ( x )
of a n a b e l i a n v a r i e t y with many complex multiplications
m e a n s the field generated over F by the coordinates (resp. t h e
The c l a s s -field-theoretical c h a r a c t e r i z a t i o n of K' ( (z)).
quotients of the homogeneous coordinates) of x. If K i s a Galois
A f u r t h e r method of constructing c l a s s fields
extension of F, G(K/ F) stands f o r the Galois group of K o v e r F.
The H a s s e z e t a function of a n algebraic c u r v e
Infinite Galois extensions with l -adic representations
F u r t h e r generalization and concluding r e m a r k s
Bibliography
1. Introduction

Our starting point i s the following t h e o r e m which was stated by


Kronecker and proved by Weber:

Theorem 1. E v e r y finite abelian extension of Q contained &


2ni/ m
a cyclotomic field Q(5)
- with a n m - t h root of unity 3 =e for-
s o m e positive integer m.

As i s immediately observed, 5 i s the special value of Qe expo-


2 niz
nential function e a t z = l/ m . One can naturally a s k the following
question:
2 niz
Find analytic functions which play a r o l e analogous to
- e

-f o-r a given algebraic number field.


Such a question was r a i s e d b y Kronecker and l a t e r taken up by
Hilbert a s the lzth of h i s famous mathematical problems. F o r an
imaginary quadratic field K, this was settled by the works of Kronecker
himself, Weber, Takagi, and Hasse. It t u r n s out that the m a x i m a l abelian
extension of K i s generated over K by the special values of c e r t a i n
elliptic functions and elliptic modular functions. A p r i m a r y purpose of
t h e s e l e c t u r e s is to indicate briefly how this r e s u l t can be generalized
f o r the number fields of higher d e g r e e , making thereby an introduction
to the theory of automorphic functions and abelian v a r i e t i e s . I will a h o
include s o m e r e s u l t s concerning the zeta function of an a l g e b r a i c curve
in the s e n s e of Hasse and Weil, since this subject i s closely connected
with the above question. F u r t h e r , it should be pointed out that the auto-
morphic functions a r e meaningful a s a m e a n s of generating not only
abelian but a l s o non-abelian algebraic extensions of a number field. and vice v e r s a . We s h a l l l a t e r d i s c u s s special values of automorphic

Some ideas in this direction will b e explained in the l a s t p a r t of the functions with r e s p e c t to r f o r an arithmetically defined l? with

lectures. compact #/r . But we f i r s t consider the m o s t c l a s s i c a l group l? =


S L (2). Since i s not compact in this c a s e , one has to impose
2 $/I'
a c e r t a i n condition on automorphic functions. It i s well known that
e v e r y point of f c a n be t r a n s f o r m e d by an element of T = SL ( 2 )
2
2. Automorphic functions on the upper half plane, into the region

e s p e c i a l l y modular functions

Let 5 denote the complex upper half plane:


No two distinct inner points of F can be t r a n s f o r m e d to each other
by a n element of T . Now $/I? c a n be compactified by adjoining a
2riz
point a t infinity. By taking e a s a local p a r a m e t e r around this
b
We l e t e v e r y e l e m e n t a = d) of GL2(R), with det ( a ) > 0 , act point, we s e e that $1 r becomes a compact Riemann s u r f a c e of
genus 0. Thus we define an automorphic function with r e s p e c t to l?
t o b e a meromorphic function on this Riemann s u r f a c e , considered a s

(2.1) a ( z ) = (az + b ) / (cz t d) a function on


/B .f.
In other w o r d s , l e t f be a r - i n v a r i a n t m e r o m o r -
6 1
phic function on For y = , we have y (z) = z t 1. Since
It i s well known t h a t the group of analytic automorphisms of ff is f (y ( 2 ) ) = f (z), we can e x p r e s s f (z) in the f o r m f (z) = Z 0 c e 27rinz
n=-w n
isomorphic to S L (R)/ { t l ). L e t r be a d i s c r e t e subgroup of with c c C. Now a n automorphic function with r e s p e c t t o r i s an
2 2 n
SL2(R). Then the quotient $ 1 ~ has a s t r u c t u r e of Riemann surface f such that c = O for a l l n < n f o r s o m e n , i. e . , meromorphic
n
such that the n a t u r a l projection $+ $/I? i s holomorphic. If
in the local p a r a m e t e r q = e
2riz
a t q = 0 . Such a function i s usually
$ /r i s compact, one can simply define an automorphic function on
called a modular function of level one. $ 1
f r e s p e c t t_o l? to be a m e r o m o r p h i c function on -$ invariant
Since is of genus 0,
a l l modular functions of level one f o r m a rational function field over C.
under the e l e m e n t s of I? . Such a function m a y be r e g a r d e d a s a
As a generator of this field, one can choose a function j such that
merornorphic function on the Riemann surface $11. in an obvious way,
h a s one of the following two normalized f o r m s :

(i) P(z) = z t A,
Obviously the function j c a n be c h a r a c t e r i z e d by (2.2) and the
(ii) P(z) = K z ,
property of being a g e n e r a t o r of the field of a l l modular functions
of level one.
with constants h and K . This can be shown, for example, by taking

Now l e t K b e a n imaginary quadratic field, and a a frac- the Jordan f o r m of a . In the f i r s t c a s e , we call a parabolic; in
the second c a s e , a i s called elliptic, hyperbolic, o r loxodromic,
tional ideal in K. Take a b a s i s {wl, w2) of 8t over Z. Since K
is imaginary, wl/ w2 i s not r e a l . Therefore one may a s s u m e that
according a s 1K I = 1, K r e a l , o r otherwise. In this classification,

wI/ Y2 '$ . by exchanging w and u2 if n e c e s s a r y .


1
In this setting.
we exclude the identity transformation, which i s represented b y the
scalar matrices.
we have:
If a s GL2 (R) and d e t ( a ) > 0, a maps -f onto itself, and a
is
T h e o r e m 2. T h e maximal unramified abelian extension of K
elliptic if a h a s exactly one fixed point in
--
c a n be generated 9 j (w / w2)
1
over K. $ 8

hyperbolic if a h a s two fixed points in R U {m),


This is t h e f i r s t m a i n t h e o r e m of the c l a s s i c a l theory of c o m -
parabolic if a h a s only one fixed point in R U {m).
plex multiplication. To c o n s t r u c t ramified abelian extensions of K , No transformation in G L (R) with positive determinant i s loxodromic.
2
one needs m o d u l a r functions of higher level (see below) o r elliptic If we put
functions with p e r i o d s w w2 . Even Th. 2 c a n fully be understood
with the knowledge of elliptic functions o r elliptic c u r v e s , though such
a r e not explicitly involved in the statement. T h e r e f o r e , our next t a s k
i s to r e c a l l s o m e e l e m e n t a r y f a c t s on this subject. But before that, then i t c a n easily be verified that SO (R) i s the s e t of a l l elements 0.l
2
i t will be worth d i s c u s s i n g a few elementary facts about the fractional SL2(R) which leave the point i fixed. Therefore the map
l i n e a r t r a n s f o r m a t i o n s and discontinuous groups. \
Every a = % r GL (C) a c t s on the Riemann s p h e r e C U{m)
):
2 -1
by the rule (2.1). With a suitable element fj of GL2(C), P = S a c
gives a diffeomorphism of the quotient SL2 (R)/ Slb2 (R)onto # . (Here note that if f s a t i s f i e s (i), then f(p(z)) i s invariant under

It i s a fruitful idea to r e g a r d $. a s such a quotient. But I shall not z Hz + 1, hence f(p(z)) i s always meromorphic a t l e a s t in the

pursue this view point, f r o m which one can actually s t a r t investi-


domain O < ( q[ < r for some r > 0 ) .
gation in various directions. Let 9 b e the join of # and a l l the cusps of I? . Then J?

We s e e easily that a differential f o r m


a c t s on 9. One can define a s t r u c t u r e of Riemann surface on
fj*/rby taking 2 z ] a s a local p a r a m e t e r around the
-2 point s . (Actually the proof of the fact that
Y dxAdy (z = x t iy)
-%
*/I? i s a Hausdorff
space i s not difficult, but non-trivial. ) Then an automorphic function

on is invariant under SL2 (R). Therefore we can introduce an with r e s p e c t to r , defined above, i s nothing e l s e than a meromorphic

invariant m e a s u r e on # by means of this form. function on the Riemann s u r f a c e PIT,


regarded a s a function on
To speak of a n automorphic function for a r with non-compact -f . The above discussion about SL ( Z ) i s a special c a s e of these
2
facts. Now the following facts a r e known:
1 , we have to introduce the notion of cusp. Let J? be a d i s c r e t e
subgroup of SL2(R). We c a l l a point s of R U { m ) a=of J? if Proposition 1. j'/r -
i s compact if and only if d/r -has

there e x i s t s a parabolic element y of I7 leaving s fixed. Let --


a finite m e a s u r e with r e s p e c t t o the above invariant m e a s u r e .
Proposition 2. Suppose that
$/r has a finite m e a s u r e .

-
Then
I i s compact if and only if I' has no parabolic element.

As for elliptic elements, the following proposition holds :


Then one c a n find an element p of S L2 (R) s o that p(w) = s , and Proposition 3. k t r &a point of -$ fixed 9 an e c element
PTsp
-1
i s generated by (1O i) and possibly by -I2 . Then we define -
of r. -
Let = a az =z . -
Then Tz 2 a cyclic group of
an automorphic function on #. w x respect 2 r to be a meromorphic --
finite order. 6
function on -# satisfying the following two conditions:
Such a point z i s called a n elliptic point of I?, and the o r d e r
(i) f ( y ( z ) )= f ( z ) f o r a l l y e r . of r z .{+12}/ {+12} is called the o r d e r of the point z (with r e s p e c t
(ii) Lf s g
a cusp of I? a n d p k a s above, then f(p(z)) ka t o I?). Two elliptic points or cusps a r e called equivalent if they a r e
2riz
i z _n neighborhood o_f q = 0.
meromorphic function in q = e transformed to each other by elements of I?. If - f / r is of finite
periods in L i s a m e r o m o r p h i c function on C invariant under the
m e a s u r e , t h e r e a r e only a finite number of inequivalent elliptic
points and c u s p s , and the following f o r m u l a holds: translation u u +w for every w t L. Define complex numbers
g2, g3 and meromorphic functions P(u) and 8. (u) on C by

H e r e g i s the genus of p / r ; h i s the number of inequivalent

c u s p s ; C Z i s the s u m extended over a l l inequivalent elliptic points;


e is the o r d e r of z. F o r r = S L 2 ( Z ) , one h a s g = 0, h = 1, e Z =
2 o r 3 according a s z = c
l o r z = (-1 + c 3 ) / 2.
F o r e v e r y positive integer N, set
where C denotes the s u m extended over a l l n o n - z e r o w in L.

I'm) = { a c S ~ ~ (1 za) r l2 mod N. M ~ ( z ) ) . Then it is well-known that

An automorphic function with r e s p e c t to I'm) i s called a modular

---
function of l e v e l N.
(3. 3) The field of a l l elliptic functions with periods in L coincides
-
-
with C ( 9 , ), the field g e n e r a t e d b~ P -
and $g E r C.

Now l e t E be the algebraic curve defined b y


3. Elliptic c u r v e s and the fundamental t h e o r e m s
of the c l a s s i c a l theory of complex multiplication

L e t L b e a l a t t i c e in the complex plane, i. e . , a f r e e


H e r e we consider E a s the s e t of a l l points (x. y) satisfying (3.4)
Z - submodule of C of r a n k 2 which i s d i s c r e t e . Then C / L i s a '

An elliptic function with


with x, y in C , together with a point (w,a!. Then the m a p
compact Riemann s u r f a c e of genus one.
A. L C L. Let End(E) denote the ring of all such endomorphisms. It
can easily be proved that End(E) i s isomorphic to Z unless
gives a holomorphic isomorphism of C / L onto E in the sense of Q(w / w ) i s a n imaginary quadratic field. Assume that Q(w / w ) is
21 2 1 2
complex manifold. It is a l s o known that any elliptic curve (i. e . an imaginary quadratic, and put K = Q(w / ). Then End(E) i s iso-
1 ' 2
algebraic curve of genus one) defined over C i s isomorphic to a morphic to a subring of the ring 0' of a l l algebraic integers in K ,
curve of this type, and hence t o a complex torus. which generates K. In this c a s e we say that E has complex multi-
Take a b a s i s {ul, w ) of L over 2. We may a s s u m e that
2
plications. In particular, if L = Zwl + Zw2 i s an ideal in K , End(E)
w 1/ u 2 f
, Then one c a n easily show that i s isomorphic to Q . P u t jo = j(wl/ %). F o r a given L (or wl, w2),
one can find the equation (3. 4) s o that g2 and g3 a r e contained in
Q(j ). Moreover j i s a n algebraic number if E has complex multi-
0
plications.
defines a one-to-one correspondence between $11. & a l l the iso- Now write E a s E(&) if L = DL for an ideal 6t in K.
m o r p h i s m - c l a s s e s of elliptic c u r v e s . F u r t h e r m o r e we have an We choose the equation f o r E(8L ) s o that g 2 , g3 c Q(jo). Suppose
important relation we could somehow prove that K(j ) i s an abelian extension of K.
(Anyway this i s not the m o s t difficult point of the theory. ) Take a
p r i m e ideal J7/ in K u n r a m i f i e d i n K ( j ), andlet a = [T , KGo)/K]
(= the Frobenius automorphism of KG ) over K f o r 2 ). Then g2
One should note that the right hand side can be obtained purely a l - and g3 a r e meaningful. Therefore we can define an elliptic c u r v e
gebraically f r o m the defining equation (3. 4) for E, while the l e f t E(@L)O by
i s defined analytically. This coincidence of an algebraic object with
an analytic object h a s a deep meaning, though we know, f r o m (3. l ) ,
that g2 and g depend analytically on wl and w2 . We call the
3
number e x p r e s s e d by ( 3 . 5 ) the invariant o_f E. Two elliptic curves Then one has a fundamental relation:
have the s a m e invariant if and only if they a r e isomorphic.
Let us now observe that any holomorphic endomorphism of E =
C /L i s obtained by u H Xu with a complex number X satisfying
If we denote by j ( a) the invariant of E(& ), then (3.6) i s equi- The answer is affirmative but not unique. It may be said that the
valent to world of mathematics is built with a g r e a t harmony but not always
in the f o r m which \r*e expect before unveiling it. This certainly
applies to our present question. I s h a l l , however, f i r s t present a
comparatively simple answer which consists of the following t h r e e
F r o m the relation (3.6) o r (3.7), one c a n easily derive Th. 2 objects:
and a l s o the r e c i p r o c i t y law in the extension KGo) of K. H e r e I do (A1) abelian variety,
not go into d e t a i l of the proof of (3. 6), but would like to call the r e a d e r 1s (B' ) Siege1 modular function,
attention t o the following point: Although no elliptic curves appear in (C1) totally imaginary quadratic extension of a totally r e a l
Th. 2 , they conceal themselves in it through the above (3.6) and the algebraic number field.
following f a c t s .
At l e a s t this will include the above r e s u l t concerning elliptic c u r v e s

2 a s a special case. A different type of theory, which I feel r a t h e r un-


(3.8) -
The quotient $/I? in o n e - 5 - 0 2 correspondence with all
t h e i s o m o r p h i s m c l a s s e s of elliptic c u r v e s . expected, and which a l s o generalizes Th. 2, will be discussed l a t e r .

j(w 1w ) i s the invariant of an elliptic curve E isomorphic


(3. 9)
1 2 --
-
to C / (Zwl Zw2).+
(3.10) 2 Q(ul/ uZ) 2 imaginary quadratic, End@) i s non-trivial. 4. Relation between the points of finite o r d e r on
an elliptic curve and the modular functions of
L e t u s now consider the question of generalizing Theorems 1 and
higher level.
2 t o the fields of higher degree. We observe that t h e r e a r e three
objects:
Before talking about abelian v a r i e t i e s , l e t us discuss the topic
(A) elliptic c u r v e ,
given a s the title of this section. Any hasty r e a d e r may skip this
(B) modular function,
section, and come back afterward.
(C) imaginary quadratic field.
F i x a positive integer N. Observe that any point t on E
Among many possible i d e a s , one c a n take the m o s t naive one, namely
such that Nt = 0 can be expressed a s
ask whether t h e r e exist generalizations of (A), (B), (C)whose i n t e r -
relation i s s i m i l a r to that of the original ones, a s described in (3.8-10).
Theorem 3. L e t K be a s above, and @
, an ideal in K. Take
@L L, and l e t 6 = Zw1 t Zw2 wifh wl/ w2 6 6. Suppose that
with integers a , b. Now, f o r each o r d e r e d pair (a, b ) of integers g Z ( 6t ) g 3 ( 8 L ) # 0. Then the maximal abelian extension f K i s
N
such that (a, b) ? (0, 0) mod (N), we can define a meromorphic g e n e r a t e d over K & j (wl/ w2) and the fab (wl/ w2) for a l l N, a , b ,
N
function f (z) on
ab
by # ---
with a fixed . 8L .
N
It should be observed h e r e that fab(wl/ w2) i s a special value
of an elliptic function and a special value of a modular function of
level N a s well. This coincidence will not n e c e s s a r i l y be retained
in one of our l a t e r generalizations.
where z = w / w and L = Zwl t Zw2
1 2
.
This i s possible because the We note that g
2
(a
) = 0 o r g3 (a
) = 0 according a s K =

right hand s i d e depends only on z = w / w


1 2'
Then a((-1t a)/ 2) o r K = Q ( m).
In these special c a s e s , we c a n still
hT
obtain the s a m e type of r e s u l t by modifying the definition of fLY
ab
N N suitably.
fab(z)=f z (a,b)=_(c,d)mod(N)
cd
The function field C ( i , f
N
), with a fixed N, is a Galois ex-
o r (a, b ) (-C, -d) mod (N).
ab
tension of C(j) whose Galois group i s isomorphic to r ( l ) / r ( N ) . {&12};
By a simple calculation, we can show that, for every a r SL2(Z), the l a t t e r group i s isomorphic to S L (Z/ NZ)/ (21 ). Since our purpose
2 2
i s to construct number fields by special values of functions, i t i s
meaningful to take Q, instead of C, a s the basic field. Then we
obtain:
N N
Therefore, f (2) = f (a ( 2 ) ) for a l l (a, b) if and only if a belongs
ab ab
Theorem 4.
N
Q(j, fab) i s a Galois extension o_f Q(j) whose
to ?I (N). { i J L } . It follows that j and the b:f , for a l l (a, b ) ,
Galois group i s isomorphic t o GL2(Z/ NZ)/ {f12}, and the following
generate the field of a l l modular functions of level N. Roughly speak- statements hold.
ing, the modular functions of level N c a n be obtained f r o m the in-
(i) F o r e v e r y a E G L 2 ( Z / NZ), the action of a a n element
v a r i a n t of elliptic c u r v e s and points of o r d e r N on the curves.
we have the following r e s u l t which i s an analogue of Th. 1 for an
Now
---
of the Galois group is given by fib -
f N with (c d) = (ab ) a .
cd
I-+
N
(ii) If y E S L ( Z ) , the action of y mod (N)
2
Q(j, f ) i s
ab -
imaginary quadratic field. N
obtainedh 'f'(z)rj y ( y ( z ) ) for YE
Q ( j , fab).
N
(iii) Q(j, f ) contains 5= e 2nil N, and
a r GL2 ( Z / N Z ) 6
ab - We call such a Riemann form on Cn/ L. Take a basis
n
sends 5 2 6 det(a) {vl, . . . , v 2n ) of L over Z , and r e g a r d the elements of C a s
column vectors. Then we obtain a m a t r i x
We shall l a t e r extend this theorem to the field of automorphic
functions with r e s p e c t t o a m o r e general type of group.

of nX2n type, which may be called a p e r i o d m a t r i x for Cn/ L. De-


fine a m a t r i x P = (p..) of s i z e 2n by p.. = f * ( v i , v.). T h e n t h e
U 1J J
above a r e equivalent to the following (Ri-3):
5. Abelian v a r i e t i e s and Siege1 modular functions
i ) 'P = - P ;

A non-singular projective v a r i e t y of dimension n , defined over (R;) p.. Z;


1J
i s called a n abelian v a r i e t y if i t i s , a s a complex manifold, iso-
C,
morphic to a complex torus of dimension n.
5 ) 52p-l . = 0, & -~ ---
1 a ~ -t~I . i s a positive definite -
An elliptic curve i s
hermitian matrix.
nothing but a n abelian v a r i e t y of dimension one. We know that any
one dimensional complex t o r u s defines an elliptic c u r v e , but such The m a t r i x P (or its i n v e r s e ) i s called a principal m a t r i x of 52 .
i s not t r u e in the higher dimensional case. To explain the n e c e s s a r y Assuming these conditions, l e t A be a projective variety i s o -
n
condition, l e t L be a lattice in the n-dimensional complex vector morphic t o C / L. Shifting the law of addition of cn/L to A , we
space cn , a d i s c r e t e f r e e Z-submodule of rank 2n in C n
i. e . , . c a n define a s t r u c t u r e of commutative group on A. Then the map
n
Then the complex t o r u s C / L h a s a s t r u c t u r e of projective variety,
and hence b e c o m e s an abelian variety, if and only if t h e r e exists an A X A 3(x, y ) C , x + y E A
R-valued R-bilinear f o r m G ( x , y) on Cn with the following
properties: can be e x p r e s s e d rationally by the coordinates of x and y. This i s
'
classically known a s the addition theorem of abelian functions.
In general, a projective variety A, defined over any field of
any c h a r a c t e r i s t i c , i s called an abelian v a r i e t y , if t h e r e exist rational
mappings f : A X A -+ A and g: A --t A which define a group s t r u c -
t u r e on A by f(x, y) = x + y, g(x) = -x. Additive notation i s used since any such
group s t r u c t u r e on a projective variety can be shown to be commuta- corresponds t o such a n abelian variety. Obviously
tive. It should be observed that such a variety defined over C , As a n analogue of S L (R), we introduce a group
2
being a connected compact commutative complex Lie group, m u s t be
isomorphic to a complex torus.
If n = 1, t h e r e i s a single universal family of elliptic curves
p a r a m e t r i z e d by the point of #. If n > I, however, t h e r e a r e F o r every U = [: 1 1 6 Sp(n, R ) with a , b , c , d in Mn (R), we
infinitely many f a m i l i e s of abelian v a r i e t i e s depending on the elementary define the action of U on
divisors of P , a s shown in the Supplement below. But we s h a l l f i r s t
fix our attention to one particular family by considering only abelian
v a r i e t i e s f o r which P = J
n
. where
Put

When n > I, we can define an automorphic function with r e s p e c t to


Under t h i s assumption, l e t q and w2 be the s q u a r e m a t r i c e s of s i z e Sp(n, Z ) to be a meromorphic function on invariant under
n composed of the f i r s t and the l a s t n columns of respectively. Sp(n, Z). Fortunately, if n > 1, i t i s not n e c e s s a r y to impose any
-1
One c a n show that w i s invertible. P u t z = w u1 If we change the . f u r t h e r condition like that we needed in the c a s e n = 1. Such a
n
coordinate s y s t e m of C by w2 , we may a s s u m e that 52 i s of the function i s us'ually called a Siege1 modular function (of degree n
form and level one).
Put r = Sp(n, Z). Now one can a s k whether the quotient
fillr
i s in one-to-one correspondence with all the isomorphism c l a s s e s of
abelian v a r i e t i e s of type (5.1). This i s a l m o s t s o but not quite. To
Now it c a n be. shown (see Supplement below) that z i s s y m m e t r i c and g e t a n exact answer, we define 2n r e a l coordinate functions x (u),
1
Im(z) i s positive definite. We denote by $-- the s e t of a l l such z ... 2n
n
, x (u) (u E C ) by u =
n
xi(u)vi , and consider a cohomology
of degree n. Thus every abelian v a r i e t y , under the assumption that c l a s s c on A represented by a differential f o r m
P has the f o r m (5. l ) , c o r r e s p o n d s to a point of , though z i s
#n
not unique for a given abelian variety. Moreover, e v e r y point of
-
and (A, c ) respectively.
~ Then the coordinates of the point (z)
belong to k, a n d ~ ( z= )Y)(zl
~ ).

of degree 2. Such a c is called a polarization of A, and the s t r u c - In (iii), we of c o u r s e consider A a s a projective variety de-
t u r e (A, c ) f o r m e d by A and i t s polarization c i s called a p o l a r i z e d fined by s o m e homogeneous equations. Now one can prove that the
abelian variety. T&n f n r e p r e s e n t s a11 the isomorphism c l a s s e s cohomology c l a s s c i s r e p r e s e n t e d by a divisor on A (i. e. an (n - 1)-

-of polarized abelian v a r i e t i e s of type (5. l ) , the isomorphism being de- dimensional algebraic s u b s e t of A). If the defining equations f o r A
fined in a n a t u r a l way. and such a divisor have coefficients in a field k , we s a y that (A, c )
O u r next question i s about the existence of s o m e functions s i m i l a r i s defined over k. If o i s a s in (iii), the t r a n s f o r m s of the equations
to j and the analogue of (3.5). F i r s t one should note that t h e r e e x i s t s by o define an abelian v a r i e t y together with a d i v i s o r , which t u r n s out

--
a Z a r i s k i open s u b s e t V o_f 5 projective v a r i e t y V* and a holomorphic to be a polarized abelian v a r i e t y of type (5. l ) , which we write a s (A, cIa .
mapping of fn into V which induces a biregular isomorphism of We can actually prove a s t r o n g e r statement than (iii), which i s roughly
$n/r Onto V. T h i s was proved by W. L. Baily using the Satake a s follows:
compatification of I . We c a l l such a couple (V, y) a model (iv) If (A' , c 1) 2 specialization of (A, c ) o v e r Q, z'
for Sn/r. This is-not sufficient f o r our purpose. In f a c t , in the corresponds 2 (A' , c ' ), then ((A1, c 1), (zl )) specialization
c a s e n = 1, the function (aj + p)/(yj + 6) plays a r o l e of ? for -
of ((A, c ) , cf(z)) over Q.
F o r details we r e f e r the r e a d e r to the paper [ZO] in $12.
any [: GL2(C). Of c o u r s e one can not r e p l a c e j b y such a
Thus plays a r o l e s i m i l a r to j. It i s analytic on f n , and
function in Th. 2. F u r t h e r m o r e , we would like t o have an analogue of
a t the s a m e t i m e , it i s a rational expression of the coefficients of defin-
(3.5). T h e r e f o r e a f u r t h e r refinement i s n e c e s s a r y , and can be given
ing equations f o r (A, c ) , a s explained in (iii). F r o m (i) it follows that
a s follows:
the coordinates of y ( z ) generate the whole field of Siege1 modular
T h e o r e m 5. T h e r e e x i s t s a couple (V, ) with the following
functions of degree n. The couple (V, y) c a n b e c h a r a c t e r i z e d by
properties:
these p r o p e r t i e s (i, i i , iii). Namely, if (V' , cj?') i s another couple
(i) (V, y) i s a model for
fnfr . with the s a m e p r o p e r t i e s , t h e r e e x i s t s a biregular isomorphism f of
(ii) V is.defined over Q.
V onto V' defined over Q such that y' =f 0 (o . Moreover, from
(iji) k t (A, c ) a polarized abelian v a r i e t y with a P of type
(iii), we s e e that the field Q ( 'f'(z)) has an invariant meaning f o r the
(5. l ) , defined over a subfield k f C, and a & isomorphism of k

-
into C. L e t z and z' be points on fn corresponding t_o (A, c )
isomorphism c l a s s of (A, c ) . We c a l l i t the field of moduli of (A, c ) . T h e r e f o r e , to discuss 52 satisfying (Ri-3), we may assume
Actually we c a n prove a l l these things without assuming P = J
n
. r0 -el
with e a s in the above lemma. Let Y p be
F o r each choice of P (or r a t h e r f o r a choice of elementary divisors thatP=Le ol
of P ) , one obtains a group r (see Supplement below) acting on
%n
the space of a l l such 51, and l e t
P
and a couple (Vp , yp) with the properties (i, ii, iii) modified
suitably. F u r t h e r , by considering the points of finite o r d e r on the
abelian v a r i e t i e s , one c a n generate the field of automorphic functions
with r e s p e c t t o congruence subgroups of Sp(n, Z ) ; one then obtains
a theorem analogous to Th. 4.
The next thing t o do i s the investigation of special m e m b e r s of In particular G
P
= Sp(n, R) if e = 1
n
. If 52 c Yp and U r G
P '
our family of abelian v a r i e t i e s , analogous t o elliptic c u r v e s with com- t
then a U r Yp . ObviouslyB J B = P, hence BG B-I = Sp(n, R).
n P
plex multiplications. This will be done in s6.
Now write = (v v l ) with two elements v and v' of Mn(C). Then,
Supplement t_o s5. To d i s c u s s the families of abelian v a r i e t i e s
from (R;), we s e e easily that
of a m o r e g e n e r a l type, for which P i s not n e c e s s a r i l y of the f o r m
(5. l ) , f i r s t we r e c a l l a well known

Lemma. L A P b_e invertible alternating m a t r i x of s i z e 2n

---
with e n t r i e s in 2. Then t h e r e e x i s t s an element U of GL
2n
(Z) e l . - ve -1 . t-v ' ) i s positive definite.

-
that
The l a s t fact implies that v and v1 a r e invertible. F r o m these
relations i t follows that e v l -'v i s s y m m e t r i c and has a positive
definite imaginary p a r t , i. e. , ev' - l v r
n
.
If z $n and U =[: t] r Sp(n, R ) , then (z ln)U r Y
Jn'
hence by the above r e s u l t , (z ln)U = A(w
ln) with A r M (C) and
n
-
= 0 mod
w t $n . Then one obtains w = (az t b ) ( c z t d)
- 1 This shows .
that
--
where the e
i -
a r e positive integers satisfying eitl (ei).
the action of U on can actually be defined. Since the action
-fn
In g e n e r a l , two d i v i s o r s X and Y a r e called algebraically
of U-I c a n b e defined, U gives a holomorphic automorphism of
equivalent, if t h e r e exist a divisor W and i t s specializations W
Now s e t 1
and W2 over an algebraically closed field such that X - Y = W - W
1 -2
If t h e universal domain i s C , then the algebraic equivalence of di-
v i s o r s coincides with the homological equivalence. Moreover, if a

It c a n e a s i l y be s e e n t h a t rp i s a d i s c r e t e subgroup of Sp(n,R). Then divisor X r e p r e s e n t s the cohomology c l a s s c obtained f r o m a

fn/Fp r e p r e s e n t s a l l the isomorphism c l a s s e s of p o l a r i z e d abelian Riemann f o r m , then 3X i s ample. E v e r y abelian v a r i e t y , defined


over a field of any c h a r a c t e r i s t i c , h a s a polarization, since i t always
v a r i e t i e s with principal m a t r i x P.
The notion of polarization can a l s o be defined in the c a s e of h a s a n ample divisor.

positive c h a r a c t e r i s t i c . Given a n abelian v a r i e t y A defined over a Now a polarized abelian variety i s a couple (A, X ) formed by
field of any c h a r a c t e r i s t i c , and given a divisor X on A, l e t L be a n abelian v a r i e t y A and a polarization of A. This definition
the l i n e a r s p a c e of a l l r a t i o n a l functions on A whose poles a r e con- i s equivalent to the previous one, if the universal domain i s C. An

tained in X (even with multiplicities). Take a b a s i s {fo, fl, . . . , fN) i s o m o r p h i s m of A of A' i s called an isomorphism of (A, x ) to

of L over k, and consider the m a p (A', xt) if it sends to X I . For a given (A, x),we can
prove that t h e r e e x i s t s a field k with the following p r o p e r t i e s :
0

A 3 x (fo(x), . . . , fN ( x ) )projective
~ N-space. (i) If (A, x) i s defined over k,
-
then k i s contained & k.
-
0
(ii) F o r a n isomorphism o of k into the universal domain,

We c a l l X ample if this i s a b i r e g u l a r embedding of A into the (nu. EO) 5 isomorphic f_o (A, X) if and only if
-- o i s the identity

projective space. Now a polarization of A i s a s e t of d i v i s o r s mappins on k - .


on A satisfying the following conditions: If the universal domain i s C , k i s uniquely determined by the

(1) contains a n ample divisor. i s o m o r p h i s m c l a s s of (A, x), and is called the field of moduli of
(2) _II X X' belong to , then t h e r e a r e two positive (A, XI. This of c o u r s e coincides with Q ( Y, ( 2 ) ) in the special case

integers m mt such that mX & m ' X t a r e algebraically P = J n .

equivalent. .
(3) 5 m a x i m a l s e t satisfying the above two conditions.
6. The endomorphism-ring of a n abelian variety; L e t k be a field of definition f o r A and the elements of End(A),
the field of moduli of a n abelian v a r i e t y and l e t D b e the vector space of a l l l i n e a r invariant differential forms

with many complex multiplications on A, defined over k. If zl, . . . , zn a r e the complex coordinate
n
functions in C , then d zl, . . . , dzn a r e considered a s invariant dif-
F o r an abelian v a r i e t y A, we denote by End(A) the ring of f e r e n t i a l f o r m s on A, and one has
a l l holomorphic endomorphisms of A. If A i s isomorphic to a
complex t o r u s Cn/ L , e v e r y endomorphism of A c o r r e s p o n d s to
a n e l e m e n t T of M (C), r e g a r d e d a s a C-linear transformation
n
on Cn , satisfying T (L) C
L. T h e r e f o r e End(A) i s a f r e e Z-module Now e v e r y A E
*
End(A) a c t s on D a s usual; denote the action by A*.

of finite rank. L e t End


Q
(A) = End(A) 8 =Q,
and W = Q. L. Then W i s Then X H < can be extended to an anti-isomorphism of End
Q
(A)
n
a vector s p a c e o v e r Q of dimension 2n, which spans C over R, and into the ring of linear transformations in D. F r o m the relation (6. l ) ,
End (A) i s isomorphic to the ring we obtain
Q
--
(6. 2) This anti-isomorphism equivalent to the t r a n s p o s e of t h e
complex representation o_f End (A).
Q
Let b e a Riemann f o r m on cn/L. For every T E End
Q
(A),
F o r each e l e m e n t of End (A), consider the corresponding element T
Q one can define a n element T~ of EndQ(A) by
of Mn(C). Then we g e t a faithful r e p r e s e n t a t i o n of End (A) by complex
Q
m a t r i c e s of s i z e n , which we c a l l the complex r e p r e s e n t a t i o n of End (A).
Q
On t h e other hand, with r e s p e c t t o a b a s i s of W over Q (for example,
{vl,. . . , v 2n ) c o n s i d e r e d i n s5), we obtain a r e p r e s e n t a t i o n of End
Q
(A)
H e r e we identify an element of End (A) with the corresponding element
by rational m a t r i c e s of d e g r e e 2n, which we c a l l the rational r e p r e - Q
of M n ( C ) Then p $ a positive involutinn o_f EndQ(A). In g e n e r a l ,
sentation of End
Q
(A). As an e a s y e x e r c i s e of l i n e a r a l g e b r a , one c a n
a n involution of an associative a l g e b r a S over Q (or R ) i s , by
prove :
definition, a one-to-one m a p p of S onto S such that
L e m m a . , The rational r e p r e s e n t a t i o n f End (A) equivalent
Q
to the d i r e c t s u m of the complex r e p r e s e n t a t i o n of
------ End (A) and i t s
Q
complex conjugate.
to s e e that the direct factors of F 8 R a r e a l l r e a l , i. e. , F i s
Q
totally r e a l . F u r t h e r , if [K : F ] = 2, the d i r e c t f a c t o r s of K @ R
Q
a r e a l l C, i. e . , K i s totally imaginary.
Conversely, l e t F be a totally r e a l algebraic number field, K
a totally imaginary quadratic extension of F , and p the non-trivial
P
Such a p i s called positive if T r ( x x ) > 0 for 0 f x r S, where
automorphism of K over F. Then p is a positive involution of K.
T r denotes the t r a c e of a r e g u l a r representation of S over Q.
We fix such F , K, p, and consider a triple (A, c , 8 ) formed
If an a l g e b r a S over Q o r R h a s a positive involution p,
by a polarized abelian variety (A, c) and an isomorphism 8 of K
then S has no nilpotent ideal other than (0). In fact, if x , f 0,
into End (A) such that the map 8 (a) H 8 (aP) i s exactly the r e s t r i c -
belongs to a nilpotent ideal, then T r (xy) = 0 for e v e r y y c S, but Q
tion of the involution of End (A) obtained a s above. (Note that End (A)
this i s a contradiction, since ~r (xxP)> 0. It follows that S i s s e m i - Q a
m a y be l a r g e r than 8 (K).) We a s s u m e a l s o that 8 (1) i s the identity
simple. If e is the identity element of a simple component of s ,
of End (A). Take cn/L and W a s above. Then W may be regarded
then e e P f 0, hence e P = e. It follows that p i s stable on each Q
a s a vector space over K , by means of the action of 8 (K). Let m
simple component of S. Thus the classification of S and p can be
be the dimension of W over K, and g = [F : Q]. Then we have ob-
reduced t o the c a s e of simple algebras.
viously
If S i s an a l g e b r a over Q with a positive involution p, we
c a n extend p t o a positive involution of S 8 *R . In particular ,
(6.4) n = gm.
consider the c a s e where S i s a n algebraic number field, and use the
l e t t e r K instead of S. Put
Now r e s t r i c t the complex representation of End (A) to 8 (K). Then
Q
we obtain a representation of K by complex m a t r i c e s of s i z e n.
In this situation, we s a y that (A, c , 8 ) of type (K, 9).Since K is
a field, 9 i s equivalent to the d i r e c t sum of n i s o m o r p h i s m s of K
Then [K : F] = 1 o r 2. By the g e n e r a l principle we just mentioned,
into C. By our choice of K, there a r e exactly 2g isomorphisms of
p i s extended t o a positive involution of the tensor product K BQR
K into C , which can be written a s
which i s a d i r e c t s u m of copies of R o r C. F r o m this fact i t i s e a s y
7
71
with a suitable choice of g isomorphisms rl, .. ., 7 among them. s ,. . . , s . Take any f r e e 2-submodule 6t of K of rank 2g.
g
Let r and s V be the multiplicity of 5 and p r V in 9 , re- Put
spectively, o r symbolically, put

It can easily be shown that L is a lattice in cg , s o that c g / L is


u
Note that a P u i s the complex conjugate of a for e v e r y a r K and a complex torus. Take a n element 5 of K s o that
e v e r y i s o m o r p h i s m a of K into C. F r o m the above l e m m a it follows

that ~ t = ~ s(Vr)(rV
~ + +
p7 ) i s equivalent to a rational representation
of K. T h e r e f o r e we have
Define a n R-valued alternating f o r m E ( x , y) on cg by

Since i s of degree n and n = m g , we have


where x v ( r e s p . y i s the f h component of x (resp. y). and t

(6.7) rv + s V= m (v = I , ..., g ) . i s a positive integer. F o r a suitable choice of t , we s e e e a s i l y that

6 becomes a Riemann f o r m on c g / L, hence cg/ L i s isomorphic


In p a r t i c u l a r , if m = 1 (and hence n = g ) , either r or s t o an abelian variety A. From 6 we obtain a polarization c of
i s 0. Exchanging r V and pry if n e c e s s a r y , we may a s s u m e that A. F o r every a r K , the diagonal m a t r i x with diagonal elements
7
,. . . , a
7~
a defines an element of Endo(A), which we write B(a).
.-
In p a r t i c u l a r , if aOt C a , the m a t r i x sends L into L, hence
3).
i.e., ip - z;=~T~. 8 ( a ) r End(A). Thus we obtain
one can prove that any (A, c , 8 ) of type
(A, c , 8 ) of type
(K, 9)
(K, Actually
i s constructed in
F o r a given K and 7 , the existence of (A, c , 8 ) of this way. If 8t i s a fractional ideal in K , and 0 denotes the
type (K, $1, with 3 - 1'"" 7g
c ~ , ,~ can
T be
~ shown a s follows. For ring of algebraic integers in K , C
then 8 ( 0 ) End(A). If n = 1, our
every s E K, l e t u ( s ) denote the element of cg with components
(A, c , 8 ) i s nothing but an elliptic c u r v e isomorphic to C/Bt 7. The class-field-theoretical characterization
(provided that r1 i s the identity map of K). of K' ( y ( z ) )
Now taking a period m a t r i x f o r A, we obtain a point z of
$n
a s in $5. H e r e we a s s u m e that (A, c ) i s such that P = J , and L e t us f i r s t r e c a l l the fundamental t h e o r e m s of c l a s s field
n
8( U ) C End(A). L e t (V, y) b e a couple a s in T h e o r e m 5. Let theory. On this topic, I s h a l l give a n exposition which i s somewhat
K(y (2)) b e the field g e n e r a t e d over K by the coordinates of the out of mode, since such will be m o s t convenient to d e s c r i b e the
point y ( z ) . One m a y naturally a s k a question: field K1 ( y ( z ) ) .
-
Is K ( 40 (z)) the m a x i m a l unramified abelian e x t e n s i o n f K? L e t F be a n algebraic number field of finite d e g r e e , Z an
This i s s o if n = 1, a s a s s e r t e d by T h e o r e m 2. But if n > 1, integral ideal in F , and .jL a (formal) product of r e a l archimedean
this i s not n e c e s s a r i l y t r u e . To c o n s t r u c t the maximal unramified p r i m e s of F. F o r an element a of F , we write
abelian extension of K, we s h a l l l a t e r d i s c u s s a function which i s
rather different from . However, even though (P is not a
function with the expected p r o p e r t y , y(z) h a s s t i l l an interesting
number t h e o r e t i c a l p r o p e r t y , which i s roughly described a s follows: if t h e r e e x i s t two algebraic integers b and c in F such that

over 5 a = b / c , b =_ c =_ 1 mod $ , and b , c a r e positive for e v e r y


-T h e o r e m 6. L e t K' be the field g e n e r a t e d Q
archimedean p r i m e involved in &. We denote b y I ( F , Z ) the
a " for a a c K. T h e n K1 ( (2)) & unramified abelian
group of a l l fractional ideals in F p r i m e to T , and by P ( F , tg)
extension of K1 . the subgroup of I ( F , 2') consisting of a l l principal ideals (a) such
V
I t c a n be shown that K' is a l s o a totally imaginary quadratic that a 1 mod* Z U.
extension of a totally r e a l a l g e b r a i c number field. Obviously L e t M b e a finite abelian extension of F. For every prime
K' = K if n = 1. However, both c a s e s K = K1 and K f K' can ideal in F unramified in M, the Frobenius automorphism
happen if n > 1. E v e n the d e g r e e s of K and Kt over Q m a y be [g , M/ F ] i s meaningful. Let .? be the relative discriminant of
different. The field K1 ( y ( z ) ) i s not n e c e s s a r i l y the maximal un- M over F. Then we can define [BL , M/ F] for every eL e I(F, 3)
ramified abelian extension of K1 . Then how big i s Kt ('-f'(z))? We s o that
s h a l l a n s w e r this question in the following section.
again by r v . Put T = Hrv , and
H' = { y c G I Ty = T I .
i s a homomorphism of I ( F , 2) into the Galois group G(M/ F) of
M over F. We have now Then f r o m our definition of K' (see Th. 6 ) , we observe that K t
i s the subfield of S corresponding to H I . Since H t T -1 = T -1 , we
Theorem 7. The m a p (7.1) $ s u r j e c t i v e , and i t s k e r n e l con-
tains P ( F , 36) f o r G.
.
c a n find elements ol , . . , oh of G s o that
- some

T h e r e f o r e , if Y i s the k e r n e l , G(M/ F) i s isomorphic to


I(F, J ) /Y. M o r e o v e r , a p r i m e ideal J in F i s fully decomposed

in M if and only if f c Y. The converse of Theorem 7 is given by


Counting the number of e l e m e n t s , we s e e that [K' : Q] = 2h. More-
T h e o r e m 8. For e v e r y g r o u p Y ' of ideals in F such that o v e r , f o r e v e r y element a (resp. ideal ? ) in K ' ,

fok s o m e 7 a s &, t h e r e e x i s t s a unique abelian extension M of


i s a n element (resp. ideal) in K .
F such that Y t n I ( F , $3)i s the k e r n e l of the m a p This follows e a s i l y f r o m (7. 2).
Now l e t I' be the group of a l l ideals /e in K 1 such that

where /Q i s the r e l a t i v e d i s c r i m i n a n t of M over F.


with an element f3 of K. I t can easily b e s e e n that I t contains
One c a n actually show that Y t C I(F, &). We c a l l this M
P ( K t , (1)). Now Th. 6 i s refined in the following way: -
----
the c l a s s field over F corresponding to Y t .
Coming back to K , rl,. . . , 7g and K' of $ 6 , l e t us take the
Theorem 6 ' . Kt ( 9 (2)) ----
i s exactly the c l a s s field over K 1
correspondingt_o I t .
s m a l l e s t Galois extension S of Q containing K, and denote by G
the Galois group of S over Q. L e t H be the subgroup of G c o r r e s - F u r t h e r m o r e , we have an analogue of the relation (3.6). T o
ponding t o K. Extend each rV to an element of G, and denote it d e s c r i b e i t , l e t us denote by A ( & ) the abelian v a r i e t y isomorphic
to cg/L with L defined by (6.8) for a n ideal in K. Take a T o prove (7.3), we have to introduce the notion of reduction of

field k of definition f o r A ( a ) containing K' ( y ( z ) ) . L e t u b e a n algebraic v a r i e t y modulo a p r i m e ideal. L e t V be a variety in

an isomorphism of k into C such that u = [T , K1 ( y ( z ) ) /K t ] on a projective s p a c e , defined over a n algebraic number field k. Let 7
K' ( (z)) f o r a p r i m e ideal 8 in K' . Then we have b e a p r i m e ideal in k , and k( ) the r e s i d u e field modulo
;P -
uA
We consider the s e t f of a l l homogeneous polynomials vanishing on

(7.3) A ( a )a 5 isomorphic fo A(O(,~-'), where % = . V, whose coefficients a r e 7 -integers. F o r each f r ,f , we


consider f mod , which is a polynomial with coefficients in k ( P ).
F u r t h e r we c a n obtain r a m i f i e d abelian extensions of K' by m e a n s Then we define the reduction of V modulo , denoted by V[ 1, to
of the points of finite o r d e r on A. be the s e t of a l l common z e r o s of the polynomials f mod f? f o r all
L e t us b r i e f l y indicate how Th. 6 and (7. 3) c a n b e proved. f c 8. If V i s an abelian variety defined over k, then one c a n show

F i r s t l e t us d e r i v e Th. 6' f r o m (7.3). Let A = A ( 6 t ) and k be a s that V[ 7] i s a n abelian v a r i e t y defined over k( -


P ) for all except a

above, and T an i s o m o r p h i s m of k into C. To simplify the m a t t e r ,


l e t us a s s u m e that (iii) of Th. 5 i s t r u e f o r the p r e s e n t A even if we
finite number of
P. F o r such a , reduction mod
e v e r y element of End(V) i s well defined, and gives an element of
of

d i s r e g a r d the polarization; namely a s s u m e End (V[FI).


We apply these f a c t s to the above A ( m ). It i s not difficult to
7 . obtain A ( 8t ) defined over a n algebraic number field k.
(7.4) A isomorphic to A ~f and only if y(z) = ~(2)' . We a s s u m e
that k contains K' ('fJ (2))and i s Galois over K' .
By the principle
(This is t r u e if g = 1, but not n e c e s s a r i l y s o if g > 1. ) Now we observe (6. 2), we can find n l i n e a r l y independent l i n e a r invariant differential

that A((%) and A ( c ) a r e isomorphic if and only if 8t and t be- forms w, ... , wg on A, rational over k , s o that

long to the s a m e ideal c l a s s . T h e r e f o r e , the notation being a s in (7. 3 ) ,


we s e e that A(UL )* i s isomorphic t o A ( ) if and only if b is a ( a € 0 ; v = I , . . . , g).
principal ideal in K. Combining this fact with (7.4), we conclude
that a p r i m e ideal in K ' decomposes completely in K' ( ( p ( z ) ) L e t u s a s s u m e , f o r the s a k e of simplicity, that K i s n o r m a l over

if and only if fl:=l ih i s a principal ideal in K.


the d e s i r e d r e s u l t , but not quite.
This i s a l m o s t
We could not obtain the condition
Q, K = K' , and the c l a s s number of K is one, though Th. 6 '
comes somewhat t r i v i a l under the l a s t condition. By (7. 2 ) , we can
be-

about N ( -$?), since we d i s r e g a r d e d the polarization. A c a r e f u l analysis


of polarization leads to Th. 6' .
put a = 7
-1
Let 2 be a p r i m e ideal in K of absolute degree one, Then it i s not difficult t o lift the isomorphism to that of A ( m ) O to
A A '
u
and l e t b = m:=l 7A . Take a p r i m e ideal f? in k which divides
~ ( b - l m ) ,hence (7.3).

2 , and consider reduction modulo . Indicate the reduced objects


b y putting tildes. F r o m (7.5) we obtain

8. A further method of constructing c l a s s fields

.- As I mentioned in §3, there a r e s o m e other ways of generalizing


if % = (b) with a n integer b in K. Let x be a generic point of A
Theorem 2. F o r example this can be done by considering special
over k. Then the relation (7.6) shows that every derivation of %(x)
-CV values of automorphic functions with r e s p e c t to a discrete subgroup of
vanishes on k(O (b)x), hence
S L (R) obtained f r o m a quaternion algebra.
2
A quaternion algebra over a field F i s , by definition, an algebra
B over F such that B 8
F
F i s isomorphic to M2 @) , where F
where p i s the rational p r i m e divisible by 2. Since
denotes the algebraic closure of F. F o r our purpose we take F to
be a totally r e a l algebraic number field of finite degree. Then one
can prove that

we obtain

where D i s the division ring of r e a l Hamilton quaternions, g =

over K ,
On the other hand, if
then mod p
a i s the-Frobenius substitution for
can be identified with xp .
P
Therefore
[ F : Q],
r > 0,
and r is an integer such that
and r e g a r d B a s a subset of BR
0 <_ r <_ g.

.
- -
We assume that
AO How many such B do
(7. 7) shows that, if A = A ( a), t h e r e e x i s t ? I shall answer this question afterward.

7 8 A((~L)O mod f'2 2 isomorphic f_o A(b


- 1 OL) mod 7.
For a e B, let a a be the projections of a to the (or i t s compactification when B = M2(Q)) i s a compact Riemann
1""' r
t s u r f a c e , and an automorphic function with r e s p e c t to
f i r s t r f a c t o r s M2(R). W e d e n o t e b y B t h e s e t o f a l l a in B ( 2. ) i s a
such that d e t ( a ) > 0 f o r v = 1 , .
V
..
, r . Then e v e r y element a of m e r o m o r p h i c function on 8 invariant under r ( t) (satisfying
t a n additional condition when B = M2 (Q)).
B a c t s on the product -$r of r copies of the upper half plane $,
the action of e a c h a
V
on -$
being defined a s in §2. If we denote
t em ark.In S2 we considered only d i s c r e t e subgroups of SL (R).
( a ) the reduced n o r m of a to F, then B i s the s e t of 2
by NB,
a l l a such that N ( a ) i s totally positive.
The action of a n element a of r ( ) on 5
i s that of al, i. e . ,
B/ F the projection of a to the f i r s t factor M2(R) of BR . The element
O b s e r v e that B i s of dimension 4g over Q. By a n o r d e r in
a l m a y not be contained in S L (R). But this does not produce any
B, we understand a subring O of B, containing Z, which i s a 2
difficulty, since we only have to consider
f r e e Z-module of r a n k 4g. An o r d e r i s called maximal, if i t i s not
contained p r o p e r l y in another o r d e r . T h e r e a r e infinitely many m a x i -
m a l o r d e r s in B. We f i x a m a x i m a l o r d e r 0 in B , and put

in place of r ( Z ).

The group of the above type was f i r s t introduced by P o i n c a r 6

F u r t h e r , f o r e v e r y integral ideal 1: in F , put about 80 y e a r s ago in the c a s e F = Q, and l a t e r F r i c k e considered


the g e n e r a l c a s e . They d i s c u s s e d t e r n a r y quadratic f o r m s instead of
quaternion algebras.
We have to define " s p e c i a l pointsn on % relative to ,
Then r and r ( T ),
t
a s subgroups of B , a c t on 5' . o n e can analogous t o w / W of Th. 2, where we s h a l l examine the values of
1 2
show t h a t /r( ) i s compact if and only if B has no z e r o - automorphic functions. F o r this purpose, we notice:

divisor other than 0. F o r example, if B = M2(Q), we can s e t


(?' = M 2 ( Z ) , hence r = S L 2 ( Z ) , and %i/
sInot
? compact.
em ma.
- - Let M &a totally imaginary quadratic extension of F -
Now l e t .us a s s u m e r = 1. Then $/r (Z) i s compact unless --
which i s isomorphic t o ? quadratic subfield of B over F. Then the

B = M2 (Q). T h e r e f o r e , a s was d i s c u s s e d in S 2 , the quotient $ . / I ?( ) following a s s e r t i o n s hold.


If
(l) f i s _s F - l i n e a r isomorphism o_f M into B , then Thus and z correspond to j and w / w
1 2
of Th. 2. We
t
f (M) - (0) contained 2 B , and e v e r y element of f(M) -F c a l l such a couple (V, y ) a canonical model f o r $ I T ( 7 ). If
has exactly one fixed point on # which i s common to a l l elements
- (V, 'p ) and (V' . (PI ) a r e two canonical models f o r the s a m e $ 1 (~7 ),
of f(M) - F.
- then we can show the existence of a biregular isomorphism 5 of V onto

(2) -
If 'ICM denotes the r i n g o f i n t e g e r s M, then t h e r e Vf , rational over k, such that = 6 0 . In this s e n s e , (V, (i7 )
exists a n F - l i n e a r i s o m o r p h i s m f o_f M into B such that f ( % M) C
-- 0 . i s uniquely determined f o r $/I?( Z ). It may be worth noting that the
m a x i m a l abelian extension of M can thus b e generated, over the maxi-
The f i r s t a s s e r t i o n i s quite e a s y to prove, but the second needs
m a l abelian extension of F , by special values of some specific automorphic
a somewhat deep f a c t of a r i t h m e t i c of quaternion algebras.
functions.
We a r e going to take the fixed point of f (M) -F a s our special
T o answer the questions about B and M, denote the
point" . One can of c o u r s e a s k a question: What kind of M can be let Fk
completion of F with r e s p e c t t o a n archimedean or a non-archimedean
embedded in B ? Leaving this question a s i d e for a while, we a r e now
ready t o s t a t e the main r e s u l t :
prime 2 of F. Put B
2
= B @ F
F ?'
.
Let PB b e t h e s e t o f a l l ?
such that B i s a division algebra. Then we have the following a s -
T h e o r e m 9. T h e r e e x i s t s a couple (V, y) f o r m e d by & p r o j e c t i v e sertions:
8
-
non-singular c u r v e V and a holomorphic mapping o_f $. into V (8. 2) PB & a finite s e t with an even number of p r i m e s .
with the following p r o p e r t i e s : (8.3) F o r any finite s e t P with a n even number of archimedean
(i) g i v e s a b i r e g u l a r isomorphism of $/r (T ) into V.
--
o r non-archimedean p r i m e s of F , there
--- e x i s t s a quaternion a l g e b r a
( 9 s u r j e c t i v e u n l e s s B = M (Q).) over
2 B F, --
unique up to F-linear i s o m o r p h i s m , such that P = P
B '
(ii) V i s defined over the c l a s s field k over F corresponding (8.4) _A quadratic extension M f F can be F - l i n e a r l y embedded
-
to P(F, Z C ~ )w, h e r e a0
" is the product of &l archimedean p r i m e s
-
in B if and only if M @ FF2 - ---
i s a field f o r e v e r y P P J 3 .
-
of F. (For notation, s e e § 7 . ) These r e s u l t s a r e special c a s e s of H a s s e s f s t h e o r e m s on s i m p l e
(iii) L e t M and f be a s in (2) of the above L e m m a , a& z the a l g e b r a s over algebraic number fields. Observe that g -r f a c t o r s of
fixed point of t h e e l e m e n t s of f (M) - F
- f . T h e n t& composite
(8.1) correspond to the archimedean p r i m e s of P
B '
-
of k ( and M i s exactly the c l a s s field over
(2)) M corresponding The r e c i p r o c i t y law for the extension M. k( Cp (2)) over M c a n
-
to P ( K Z ). b e described explicitly in t e r m s of the special points (z). F o r
simplicity, l e t u s consider only the c a s e where the c l a s s number of
F in the n a r r o w s e n s e i s one, i. e . , P(F, Go) = I @ , (1)). F o r every (Such a n a ~ not unique. ) With such an element a , one has
p r i m e ideal J2 in F, let 16 be the ring of ? - i n t e g e r s in F
9 I '
and l e t
J B
= % 0 . Then
2
i s a m a x i m a l o r d e r in B L e t
t 8'
U( ) denote the g r o u p of a l l e l e m e n t s a of B such that a is a
unitof 8
8
forall 3 dividing .Z , and l e t U ( t) b e the subgroup H e r e we s a y that f i s normalized if
of U( ) consisting of a l l a s u c h that a Z% I mod ru for all
dividing r. I t c a n e a s i l y be shown that U( t )/ U ( T )
3 2
i s isomorphic
(0 f a r M).
to ( 0 l f C 7 ) ~ (see Notation). F o r every a c U( t y, put

If we define ? by ?(a) = f 6) f o r a E M, then we s e e that e i t h e r f


-
or f i s normalized.
It should be observed that (iv) of Th. 10 i s a generalization of
Now we have
(3. 7 ) . F u r t h e r , if B = M2(Q), = M 2 ( Z ) , and r = (N)with a
(9

T h e o r e m 10. ---
T h e r e e x i s t s a s y s t e m of b i r e g u l a r i s o m o r p h i s m s positive integer N, the function field of V i s exactly the field
N
R ( a ) of V t o
7 - v U b ), defined f o r each a r U ( f ) and rational over Q(i, f ) considered in 93. T h e r e f o r e the f i r s t t h r e e p r o p e r t i e s of
ab
k, with the following p r o p e r t i e s : R ( a ) in Th. 10 m a y be r e g a r d e d a s a generalization of Th. 4.
(i) R ( ~ ) o @ ) R(p) = R ( a p ) .
Example. Let us consider the c a s e F = Q(3 + 5 -1) with 5 =
(ii) R ( a ) = R ( p ) if a - l p r Uo(E). e2ni1
f o r d = 7, 9, o r 11. By ( 8 . 3 ) . t h e r e e x i s t s a unique quaternion
(iii) R ( y )[ (z)] = 'p(y (2)) f o r e v e r y z r $. if y r ro). a l g e b r a B over F , f o r which P
B c o n s i s t s of a l l but one archimedean
(iv) L e t M, f , z be a s in (iii) of Th. 9 (still under the
p r i m e s of F , the exception being the archimedean p r i m e of F c o r r e -
condition f ( WM) C 0 ) . Suppose t h a t f & n o r m a l i z e d & the sense
sponding to the identity m a p of F. The p r e s e n t F has the c l a s s num-
--
defined below. Let 3 b e a n ideal in M p r i m e to Z. , and l e t
b e r one in the n a r r o w sense. Moreover, a l l the maximal o r d e r s in
B a r e conjugate to each other under the inner automorphisms of B
defined by the elements of B+ . Take a maximal o r d e r 73 in B , and

----
Then t h e r e e x i s t s an element a fU ( Z ) such that f ( 8 ) O = a g . .
define I? = r(l) a s above. Then one can prove that $1 i s of extension K of F and consider (A, c , 8 ) of type (K, 9) in the
genus 0, and r modulo i t s center i s generated by t h r e e elements s e n s e of 56 with a representation 3 of K such that
y 2 , y 3 , y d of o r d e r 2, 3, d, y 2 y 3 y d = 1.
respectively, satisfying

(If d = 7 , the m e a s u r e given by (3.4) is 11 42, which is the s m a l l e s t

value of (3.4) for a l l possible l?. ) These t h r e e elements have unique


fixed points on $, one for e a c h , which we denote by w2. w3 , wd . where T,, and P a r e a s in 56. Then i t can be shown that the (A, c , 8 )
Then t h e r e e x i s t s a unique m e r o m o r p h i c function on $. such of this type a r e parameterized by the point on gr, and t h e r e is a dis-
that C ( y ) is the field of a l l automorphic functions on 5 with continuous group I? acting on $r such that $r/l" is inone-to-one
r e s p e c t to J?, and (i) (w2) = 1, f' (w3) = 0, (wd) = UJ . If we denote
correspondence with a l l the isomorphism c l a s s e s of such (A, c. 8 ) .
by V the complex projective line, then gives a biregular i s o m o r - Taking K suitably, we can identify J?' with the above I? (in the case
phism of $/r onto V. Now we can prove that this (V, p) $32 r = 1). F o r this family of (A, c , 8 ) , we can find a model (V' , )
canonical model for #/l7 . By (8.4), for e v e r y totally imaginary of #/rf with the properties analogous to those of Th. 5. If M and
quadratic extension M of F , t h e r e e x i s t s a F - l i n e a r isomorphism z a r e a s in (iii) of Th. 9, the corresponding (A, c , 8 ) i s such that
f of M into B such that f ( % M) C c. L e t {zl, . .. , z 9 } be a s e t End
Q
(A) contains a n isomorphic image of K 8 M.
F
The coordinates
of r e p r e s e n t a t i v e s for the r-equivalence c l a s s e s of the fixed points of of (z) generate a n abelian extension of the nature described in Th. 6 ' .
f(M) -F f o r a l l such f. Then q i s exactly the c l a s s number of M, This couple (V1, ) is the f i r s t approximation to the desired (V, yl).
and f r o m (iii) of Th. 9 and (iv) of Th. 10, we obtain: F r o m infinitely many such (V' , ), depending on the choice of K , we

..., c a n construct a canonical model (V, y).


(8.5) The values ?(al),
4 --
?(Z ) f o r m a complete s e t of
conjugatesf ?(z 1) over F, and for each i , M( y ( z i ) ) i s the maxi- If B = M (Q), t h e r e i s a family of elliptic c u r v e s , for which the
2
value of j is the modulus. F o r the basic field F of higher d e g r e e ,
m a 1 unramified abelian extension of M.
t h e r e is no such standard family of a b e l i a n v a r i e t i e s , though infinitely
Thus has a s t r o n g resemblance t o the c l a s s i c a l modular
many families of abelian v a r i e t i e s can be loosely associated with a
function j.
given $ 1 . It i s a n open question whether there exists any family
Unfortunately, the proofs of Theorems 9 and 10 a r e long and
of geometric s t r u c t u r e s , other than the above (A, c , 8 ), of which our
v e r y complicated. T h e r e f o r e I have to content myself with a rough
canonical model (V, y) i s a natural variety of moduli.
sketch of the m a i n ideas. We take a totally imaginary quadratic
9. T h e H a s s e zeta function of a n algebraic curve where T = 0) and the c l a s s number of F in the n a r r o w s e n s e i s
one, i. e. , every ideal in F i s a principal ideal generated by a
L e t V be a projective non-singular curve of genus h defined totally positive element. In this c a s e , f o r e v e r y right 0 -ideal 8t ,
over an algebraic number field k. F o r every p r i m e ideal J2' in k , there exists a n element a in B' such that
= a 8. .
l e t k ( 2 ) denote the residue field modulo 2 . Considering the Let us introduce cusp f o r m s and Hecke operators with r e s p e c t

equations for V modulo & , we o b t a i n a c u r v e V [ p ] over k ( & ' t o the group I? = r p ) . F o r e v e r y f = b 6


d) c GL2 (R) with det (f) > 0,

(see §7). It can be shown that V[ ] i s non-singular, and of genus h for Put
almost all J2 . (We c a l l such 2 good. ) Therefore one can speak
of the z e t a function of V[f ] over k ( p ) which i s of the f o r m

1
Z (u) = Z (u)/ [(I - u)O - N ( ~ ) U ) ] , L e t m be a positive integer. By a cusp f o r m of weight m w
&
h respect
B 8
1
-
to r, we understand a holomorphic function f (2) on $, satisfying
where u i s a n indeterminate, and Z (u) i s a polynomial of degree 2h. the following two conditions:
Y
The H a s s e zeta function of V over k, denoted by Z ( s ; V/ k ) with com- (i) f ( y ( z ) ) j ( y , z)m = f ( z ) f o r a l l y c r.
plex v a r i a b l e s , i s now defined by (ii) s a cusp of I-', a n d p , q a r e a s in $ 2 , then
f (P (z))j(P, 4 holomorphic & q = 2niz,
---
and vanishes a t q = 0.
The l a t t e r condition is n e c e s s a r y only when B = M (a), since
2
I? h a s no cusps otherwise. All such functions f o r m a vector s p a c e of
the product being taken over a l l "good" 8. Qt i s important to con- finite dimension over C, which we write S (r). If m = 2,
m
the map

sider also "badf1 &, which we s h a l l not discuss h e r e . ) Now one

c a n m a k e the following

Conjecture. Z (s; V / k ) can be continued to the whole s -plane


gives an isomorphism of S ( r ) onto the space of differential f o r m s of
-
and s a t i s f i e s 5 functional equation. 2
the f i r s t kind on $D, hence the dimension of S (r) i s equal t o
2
The purpose of this section i s t o verify this conjecture for the genus of f/r.
the curve which i s a canonical model for f /r ( 7: ) in the s e n s e of
§8. F o r the sake of simplicity, we shall consider h e r e only the c a s e
t
F o r every a E B , we note that the double coset F a r can (see (8.2-4)), and the second over the remaining p r i m e s in F. More-
be decomposed into a finite number of one sided cosets: over, D (s) can be holomorphically continued to the whole s -plane,
m
and s a t i s f i e s a functional equation:
r a r = urZlair (disjoint).

Then for f c S (I?), we define ( T a r ) f by


m m

-1 -1 m
(raTlmf = z d 1 f ( a i (z))j(ai , 2) .

In this way we obtain a l i n e a r transformation ( r a r ) m on Sm ( r ) ,


which i s of c o u r s e independent of the choice of the representatives
a F o r e v e r y integral ideal 8t in F , l e t T(BL), denote the
i '
of a l l distinct ( r a r ) , such that a E 0 and (N ( a ) ) = 61. . where P; means the s e t of non-archimedean 2 in P
B ' The last
B/ F
few stand for the usual gamma function.
Then we define a Dirichlet s e r i e s D ( s ) , whose coefficients a r e linear
m
Now we have
endomorphisms of S (I?), by
m
Theorem 11. Let (V, y ) &a canonical model for f/r . TZ,
---
f o r a l m o s t a l l p r i m e ideals
3 & F, the z e t a function o_f V z d 2 ,
over the residue field mod 2 , -
coincides with
where (9t r u n s over a l l integral ideals in F. It can be shown that
2
D (s) converges for sufficiently l a r g e Re (s) and has a n Euler product:
m det[l - T(7I2u + N(J?)u I/ [(I - u)(l - N(Z)u)]
1
Z (u) is the
B --
(i. e . , ,., Euler 2 -factor
of Dm (s) &
y m = 2 and N(?)-' = u). -
-
Thus Z (s; V/ F) coincides &
y det(D (s)) u p to a finite number
2
of
2 -factors.

The proof of this theorem i s roughly a s follows. For every


where the f i r s t product is taken over a l l the p r i m e ideals 2 in P
B a E B
t
. define a s u b s e t X ( I ? a r ) of V X V by
Therefore the equality

We s e e t h a t X(l?al?) is a n algebraic correspondence, which depends


o n l y on F a r and not on the choice of a . We o b s e r v e that T (
7 I2
= (rar), with any e l e m e n t a of 0 n Bt such that f = (NBI (a)). should hold..
F o r such a n a , we w r i t e X ( I ' a r ) a s X L e t A denote the diagonal of V X V. Then
L e t M, f , and z b e a s in (iii)
2 -
of Th. 9 and (iv) of Th. 10.
T a k e M s o that 2 decomposes into two p r i m e ideals 7 and
7'
in M. We can find a n e l e m e n t P of 60Bt s o that f ( =PW .
Then 2 = NMl ( 7 = (NB1 (PI), hence (rPrI2= T (f) 2 by the The relations (9.1) and (9. 2) show that any s y m m e t r i c function of
above r e m a r k . Therefore ~ ( ~ - ' ( z )X ) y ( z ) c X? . Now l e t and tm can be obtained f r o m a correspondence of V by reduction
T = [q , M ( l (z))/ MI. By (iv) of Th. 10, we have ~ ( 2 ) '= ~ ( ~ - ' ( z ) ) , modulo 8. In p a r t i c u l a r , f o r e v e r y positive integer m , we have
h e n c e , putting y = y ( z ) , we obtain yT X y c X Consider now
2 .
r e d u c t i o n modulo a p r i m e divisor of a suitably l a r g e field, which
h

divides 2
, and denote the reduced object by tilde. Then y = y
7

hence f o r a polynomial P , which i s determined by


m

--dud log (1 - XU + yu 2) = *
x Pm(x, y)um-l
m =I

Let rr denote t h e Frobenius correspondence on V X V , i. e . ,


N
'V rV

with indeterminates x, y, and u.


the l o c u s of x X x N ( f ) f o r x c V. The above discussion shows that
hl L e t I[Y] denote the number of fixed points of a correspondence
h a s infinitely m a n y points in common with t m , the t r a n s p o s e
X,
of . Since trr
i s i r r e d u c i b l e , we have
t
mcT8 . F r o m our
Y, i. e . , the intersection number of Y with
the zeta function of V mod 2 , then
A . If Z
2
(u) denotes

definition of X i t can e a s i l y be s e e n that t~ = X Therefore


2 ' P
+ . Again f r o m the definition of X
8
. 8 '
we s e e that
.(xXV)
N N N
X c o n s i s t s of N ( Y ) t l p o i n t s f o r a g e n e r i c x on V.
Y
hl
10. Infinite Galois extensions with l -adic representations.
But if Y i s a correspondence on V X V, we have I[Y] = 1[Y], and

by the Lefschetz fixed point t h e o r e m ,


So f a r we have been interested only in the construction of abelian
extensions. Now we a r e going to show that the above canonical model
for $/TIf') can be employed to obtain m e a n i n g f d non-abelian ex-
i tensions of a number field, with some pleasant features.
where H (V) denotes the usual r e a l i-th cohomology group of V. For
L e t us call (V, f', R ( a ) ) of Th. 10 a canonical s y s t e m of level .t.
Y =Xrar), it i s e a s y to s e e that
F o r every integral ideal 8L in F, we fix a canonical s y s t e m of level

tr(Y 1 1
H (V)) = 2 . ~ e [ t r( l ? a r ) 2 ] ,
& , and denote i t by (Va , t,
RoL( a ) ) . (We a r e still assuming
that the c l a s s number of F i s one in the n a r r o w s e n s e , though our dis-
cussion c a n actually be done without this assumption. ) Let k a denote
since S2 (I?) is isomorphic to the space of differential f o r m s of the
the c l a s s f i e l d o v e r F corresponding to P ( F . a&), where 4x0 is
f i r s t kind on V = $ 1 ~ . Also i t is obvious that
2 the product of all archimedean p r i m e s of F (see $7). As i s stated in
t r ( Y ( HOW)) = tr(Y ( H (V)) = the number of right c o s e t s in r a r .
One can f u r t h e r show that t r (l?al?)2 is r e a l . Therefore we have
Th. 9, 10, Va and R a ( a ) a r e defined over k a . If eL C d , we
c a n obtain a rational m a p

defined over kw , such that

In view of (9. 3), we obtain

-d 1
du log 1 (u) = (det[l - T ( & )2u t N ( j )u2]-'} ,
2 du log

hence our theorem.


When 6L = b , this m e a n s the uniqueness of canonical s y s t e m of on f sothat f'*(z)=y, andput y = G ( z ) . ) Let
6t Y
level a. In the g e n e r a l c a s e , the m a p (10.1) defines a Galois c o v e r - denote the composite of the ka (ya ) for a l l
By Prop. 1, a.
ing. If we consider the c u r v e s over the u n i v e r s a l domain C , then 4 is a Galois extension of k r (y). Our purpose i s to investigate
Y
the Galois group i s isomorphic t o r ( $i )/r( cz )E% , where E % the Galois group of &
over k (y) and the behavior of the
Y
denotes the group of a l l units e of F such that e =- 1 mod 1 . If Frobenius automorphisms (when y i s algebraic) with r e s p e c t t o

we take kOL a s the field of definition, we obtain: c e r t a i n representations of the Galois group.

Proposition 1. Let y1 be an a r b i t r a r y point of Vm a& y =


F o r every p r i m e ideal f in F, l e t F1 . f. BI , and Oy
be the localizations a s defined in g8. F o r m the product lt of the
T b , &(yt ). (y' m a y b e generic o r algebraic. ) Then k o, (y' ) depends X
groups 0
1
(see Notation) f o r a l l 1, with the usual product top-
only on y, and i s a finite Galois extension o_f k a. (y). F o r e v e r y
7 c G(k., (yl ) / k (Y)), t h e r e e x i s t s a n element P of U(a,) nuo( b ) ology. ~e't atdenote the subgroup of a consisting of the elements

--
such that y1
7
= Rot,(P)(yl), & T = o on k, . ( u l ) with ut c 0; such that u
I-
= 1 mod f D for every
t .
F o r simplicity, we a s s u m e
Proposition 2. Take a point z of # s o that L(z)
= y' , and (10.2) {Y r r(Z)1 v(z) = z ) = E
-
let r Z= {y I- ( 2)1 y (z) = z ) . Then the following statements hold: t

(i) r ( e e ) . r z = { Y c r ( b ) I R,(Y)(Y') = y t ) . f o r a point z on $ such that


c ( z ) = y. This i s satisfied f o r all
(ii) F r o m the correspondence 7 H P described in P r o p . 1, one except a finite number of points on Vr .
Under the assumption, we
a r e going to define an injection
--
obtains a n i s o m o r p h i s m o_f G(ka(y1 )/ k (y)) into

- a.Let 1 k t (y)).
(iii) If -
y is g e n e r i c on Vg k% -
' this isomorphism i s
where E% i s the closure of EI in s r G(&
Y
surjective, a n d rZ= E b . F o r every C '2, we find, by P r o p . 1, an element em = 5 ( a )of
u ( an) u o ( t ) * '7
that YOL = Ra (ta)(y, 1. It i s not difficult t o choose
Let us now fix an integral ideal 2: in F and an a r b i t r a r y the elements $ s o that
OL
point y on Vy : Take a point ya on Va for each 6LC Z so
that T1; = y% i f 01 C 1, and yr =
a ( ~) a J . (For example, choose a point Z
F o r each p r i m e ideal f, the sequence { ( ( l n ) } n = l , 2 , ... converges Next l e t us consider the c a s e of an a r b i t r a r y algebraic point
to an e l e m e n t u of C; . We define J ( r ) to b e the element of y of Vr . Take any c e n t r a l simple a l g e b r a A over Q , and
r e p r e s e n t e d by (u ). We can verify that J i s actually a consider a representation
t t
continuous injection. 3 depends on the choice of the sequence of points
{y }. But it i s unique up to inner automorphisms of
OL
Er .
F u r t h e r we have:
satisfying the following two conditions:
Proposition 3. If y i s g e n e r i c on Vr o ~ kr , J issur- X X

jective.
tr @) p is rational over Q, B
-
and A being considered
algebraic groups over Q.
m
L e t u s now d i s c u s s the points y whose coordinates a r e algebraic (2) p (a) = NF/ (a) for every a e with an integer m,
n u m b e r s ; our main i n t e r e s t i s of c o u r s e in this c a s e . F i r s t we con- independent of a.
s ider the point fixed by an imaginary quadratic subfield of B. Let
F o r e v e r y rational p r i m e I , l e t QI denote the field of I -adic
M, f , and z b e a s in (iii) of Th. 9, and XM the ring of integers i n
i n t e g e r s , and l e t B
I = B QQQI, AI = A @,aI . F r o m p one can
M. Then f ( X M ) C 0. P u t naturally obtain a representation

and view a s a subgroup of %. Then, f r o m (iv) of Th. 10, we


obtain
Let aI denote the product of the groups 0 f o r a l l prime
factors t of I , and Eg
I
the c l o s u r e of Er in d. Put
Proposition 4. If y = Tr(z) M Ck (y), then the image
& (R )/ E r.
= b . Combining J with a n a t u r a l homomorphism of
-
of J contained
t m,/B, to a:/ I
ET , we obtain a homomorphism
If M is not contained in k (y), we have to consider a l a r g e r
group which contains rb;?f) fir a s a subgroup of index 2 , but we shall
not go into details.
Now we m a k e the following assumption: Suppose this i s t r u e , and l e t f. denote the principal polynomial
of p, J (cr . Then one
2
can define a zeta function c(s; k y ; p) of
(3) p ( e ) - 1 f o r a l l e c Ey . o
1 2
k Yassociated with the representation p &
In view of (2), this i s automatically satisfied if either m i s even,
o r the units of Er a r e a l l totally positive. Under the assumption,
observe that P I J i s meaningful, and defines a representation of
x
G ( & ~ k/ r (y)) into Al- . Now i t i s important t o investigate the b e -
where the product i s extended over a l l p r i m e ideals
I
havior of the Frobenius automorphisms of .gywith r e s p e c t to the p r i m e t o W , and n i s the degree of f
8 in k z ( y )

I -adic representations" pI o JI . F i r s t we notice: 7 '

Conjecture 11. 6(s; diy; P ) can be analytically continued f_o the


Theorem 12. There exists a n integral ideal p v ~ k (y) s u c h
whole s-plane and satisfies 5 functional equation.
-
that e v e r y p r i m e ideal of k (y), p r i m e to I M , unramified &
F u r t h e r , a s for the nature of & one may make
---
the subfield of Ay which c o r r e s p o n d s t o the kernel of p I J I ' 0
Y'

T h e r e f o r e , if i s such a p r i m e ideal of k (y), is a prime


Conjecture 111. The image of J i s a n open subgroup of 'aT/
tr ,
divisor of
Y
dividing 8,
and
is a Frobenius automorphism of -
unless y = yZ(z) with a point z such that

& over k (y) for f2 , then the conjugacy c l a s s of p


Y
determined only by J2 . In this setting, we have the following t h e o r e m ,
which m a y be r e g a r d e d a s the main r e s u l t of this section:

T h e o r e m 13. The roots of the principal polynomial of p J (U


-
for -
s o m e totally imaginary quadratic extension M f F and a n
0
I % F - l i n e a r isomorphism f of M into B.
over QI are algebraic n u m b e r s fabsolute value
- N(
It should be r e m a r k e d that the above representation p i s analo-
This fact leads u s t o a temptation of making the following conjec-
gous t o the G r k s e n - c h a r a c t e r in Hecke1 s s e n s e , o r m o r e precisely,
tures.
to the Grijssen-character of type (A ) in the s e n s e of Taniyama-Weil.
Conjecture I. T h e principal polynomial of p
1
0 J
1
(U
2
) has If y = yr(z) with a point z fixed by f(M - (0)) a s excluded in
rational coefficients, and i s independent of I . Conjecture 111, and if M C k (y), then we can show that c(s; ; p) &
Y
i s a product of s e v e r a l zeta functions of k 7 (y) with Grossen-characters.
One m a y a l s o notice that b%r/Et. i s analogous to the idsle c l a s s
group of a number field modulo the connected component. Further,
if B = M (Q), 0 = M (Z), Z =N Z with a positive integer N, then
2 2
r (7 ) is the principal congruence subgroup of S L2 (Z) of level N, One can show that a
* t B if a E B, and a H a
I
i s an involution
and the choice of a point y on Vr i s a l m o s t equivalent t o the choice of B in the s e n s e of 46, which i s not n e c e s s a r i l y positive. F o r an
element U = (a..) c M (B) with a.. r B, put U = (a .. ).
t :$ t * Then
of an elliptic curve. Thus, if p is the identity mapping of B' = GL2(Q)
1J n 1J 1J
to itself. ((a, gYy,
p) i s the Hasse zeta function of the c u r v e s o r r e s - t *
U w U defines a n involution of M (B). This involution c a n be
n
ponding to y. A s i m i l a r fact holds a l s o when F = Q and B i s a divi- R-linearly extended to M (B ), where B = B @ R. Define a Lie
sion algebra. It is a n open question whether such a n interpretation
n R R a
group G by
exists for c(s,
Y
, P ) in the c a s e where F is of degree >1 . In
this connection, i t should be mentioned that the extension &Y is
r a t h e r different f r o m the extension obtained f r o m the points of finite
o r d e r on a n abelian v a r i e t y , if [ F : Q] > 1. In view of (8. l ) , we have

11. F u r t h e r generalization and concluding r e m a r k s

We have obtained two different types of r e s u l t s : one i s r e p r e - According to this d i r e c t s u m decomposition, G can be decomposed into
sented by Th. 5. Th. 6, and Th. 6'; the other by Th. 9 and Th. 10. a d i r e c t product:
They a r e however two special c a s e s of a m o r e g e n e r a l theorem. To
s e e t h i s , l e t us introduce discontinuous groups which include Sp(n, Z) G = G X...
1
XGrXGr+lX... XG
g
.
and F ( 7 ) a s s p e c i a l c a s e s .
L e t F, B and r be a s in S8. Since B BFT = MZ(F) for the One can easily show that
algebraic c l o s u r e
-F of F, we c a n r e g a r d the elements of B a s
- b
m a t r i c e s of d e g r e e 2 with e n t r i e s in F . F o r every a = d) c M ~ F ) ,

put
by a maximal compact subgroup i s a bounded s y m m e t r i c domain. Then
one can speak of (meromorphic) automorphic functions and f o r m s on

where the b a r m e a n s the quaternion conjugate in D. Therefore, this domain with r e s p e c t to the group I? = qz formed by integral

Gr+l,. .. , Gg a r e compact. Since Sp(n, R) a c t s on the Siege1 points on 9. T h e r e a r e many interesting arithmetical problems
in this field, which I d a r e not enumerate h e r e . But I should a t least
space fn, we c a n l e t e v e r y element U of G a c t on the product
mention that almost a l l important questions a r e r e l a t e d to automorphic
$,n
of r copies of ,
being that of the projection of
the action of U on the u-th factor
U t o Gv. As in 58, take a maximal
h -
f o r m s and zeta functions explicitly o r implicitly, on which I have talked
only in $9.
order (p in B , and put

12. Bibliography
I n this way we obtain a discontinuous group acting on #nr .
If B = M2(Q) and 0 = M2(Z), r i s Sp(n, 2). If n = I , the
Among a v a s t multitude of l i t e r a t u r e , I shall t r y to l i s t standard
present r i s a subgroup of finite index of the group considered in

$8. The quotient gnr / T i s compact in the following two c a s e s :


r e f e r e n c e books f r o m the view point of accessibility and (probable)
comprehensibility, along with a few r e c e n t p a p e r s relevant t o the topics
(i) r < g; (ii) r = g, n = 1, and B i s a division algebra. The group
discussed in these lectures.
r was introduced by Siegel in his paper on symplectic geometry (under
the r e s t r i c t i o n r = 1). Now, for this quotient kr/r, we c a n con-
The r e a d e r with / standard knowledge of algebraic groups or
Lie theory may find the following volume useful:
s t r u c t a couple (V, y) with the properties analogous to those in Th. 9,
[I] Algebraic groups and discontinuous subgroups, Proceedings
thus unifying the above mentioned two types of r e s u l t s . A p a r t of the
of Symposia in P u r e Mathematics, vol. 9, Amer. Math. Soc. 1966.
r e s u l t s of $10 can a l s o be extended to such a g e n e r a l c a s e .
This contains many interesting surveys of r e c e n t investigations,
One c a n investigate automorphic functions with r e s p e c t to a
m o s t of which have abundant references. F o r those who a r e m o r e
m o r e g e n e r a l type of group. Namely one takes a semi-simple algebraic

group defined over Q and consider a Lie group gRconsisting interested in the classical modular functions o r modular f o r m s , many
important papers in
of the points with coefficients in R. Suppose that the quotient of
[2] E. Hecke, Mathematische Werke, Gbttingen, 1959
will s e r v e a s standard references. A m o r e systematic and somewhat
e a s i e r t r e a t m e n t i s presented in
[3] M. E i c h l e r , Einfiihrung in die Theorie d e r algebraischen The l a t t e r will s e r v e a l s o a s an introduction to the theory of

Zahlen und Funktionen, B i r k h l u s e r , 1963 (the English r e v i s e d v e r s i o n automorphic functions of s e v e r a l variables. On this topic and other

i s available). r e l a t e d subjects, one can not m i s s

At t h e end of each chapter of t h i s book, t h e r e a r e plenty of [ll] C. L. Siegel, Gesammelte Abhandlungen, 3 v o l . , Springer,

references. As a textbook on the c l a s s i c a l theory of elliptic functions, 1966.

the following m a y b e recommended: Especially f o r Siege1 modular functions, the standard knowledge

[4] C. J o r d a n , C o u r s dl analyse de 1' i c o l e polytitchnique. P a r i s , can be obtained f r o m

vol. 11, Ch. VII. [12] H. Maass, L e c t u r e s on Siegel' s modular functions, Tata

As f o r the theory of complex multiplication of elliptic functions, Institute, 1954-55,

I pick h e r e only two, old and (relatively) new: [13] SBminaire H. C a r t a n , 19571 58, Fonctions automorphes.

[5] H. Weber, Lehrbuch d e r Algebra 111, 2nd e d . , 1908, A detailed account of the r e s u l t s discussed ,in 5 7 on the nature

[ 6 ] M. Deuring, Die klassenkijrper d e r komplexen Multiplication. of the number field K' ( (2)) etc. i s presented in

Enzyclopadie d. math. Wiss. neue Aufl. Bd. 12, Heft loII, Stuttgart, 1141 G. Shimura and Y. Taniyama, Complex multiplication of

1958. abelian v a r i e t i e s and i t s applications to number theory, Publ. Math.

The fundamental m a t e r i a l of abelian v a r i e t i e s i s p r e s e n t e d by Soc. Japan, No. 6, Tokyo, 1961.

[7] A. Weil, V a r i b t e s abeliennes e t courbes alg'ebriques , A r e c e n t volume

Hermann, P a r i s , 1948. [15] A. Weil, B a s i c number theory, Springer, 1967

[a] S. Lang, Abelian v a r i e t i e s , Interscience, New York, 1959. contains a m o d e r n t r e a t m e n t of c l a s s field theory, a s well a s t h e

The analytic t h e o r y of theta functions and abelian v a r i e t i e s i s s t r u c t u r e t h e o r e m s of simple a l g e b r a s over number fields, which

systematically treated in generalize (8. 2-4). The l a t t e r subject, in a concise style, c a n be


found in
[9] A.. Weil, Introduction a 1' btude d e s variittes kahlhriennes,
Hermann, P a r i s , 1958, [16] M. Deuring, Algebren, Ergebn, d e r Math., B e r l i n , 1935.

[lo] C. L. Siegel, Analytic functions of s e v e r a l complex As f o r the general theory of arithmetically defined discontinu-

variables, lecture notes, Institute f o r Advanced Study, 1948, ous groups, I mention h e r e only t h r e e papers:

reprinted with c o r r e c t i o n s , 1962. [17] A. Bore1 and Harish-Chandra, Arithmetic subgroups of


algebraic groups, Ann. of Math. 75 (1962), 485-535.
[18] G. D. Mostow and T. Tamagawa, On the compactness of connection of the H a s s e zeta function of an elliptic curve with the
arithmetically defined homogeneous s p a c e s , Ann. of Math. 76 (1962), Diophantine problems on the curve i s discussed in a survey a r t i c l e
446-463. (of c o u r s e with numerous r e f e r e n c e s )
[19] W. L. Baily and A. B o r e l , Compactification of a r i t h m e t i c [24] J. W. S. C a s s e l s , Diophantine equations with s p e c i a l
quotients of bounded s y m m e t r i c domains, Ann. of Math. 84 (19661, r e f e r e n c e t o elliptic c u r v e s , J. London Math. Soc., 41 Q966), 193-291.
442-528.
The compactness c r i t e r i o n , which g e n e r a l i z e s that f o r
#nrlr
of s11, i s given in [17] and [18]. The l a s t paper 1191 proves the existence
of a Z a r i s k i open s u b s e t of a projective v a r i e t y isomorphic t o a given
quotient like in g e n e r a l . F o r t h e s e topics, s e e a l s o the
$ 1
a r t i c l e s in [I].
T h e o r e m s 5, 9, 10, 11 and t h e i r generalizations a r e proved in
my papers:
[20] G. S h i m u r a , Moduli and f i b r e s y s t e m s of abelian v a r i e t i e s .
Ann. of Math. 83(1966), 294-338.
[21] G. Shimura, Construction of c l a s s fields and zeta functions
of algebraic c u r v e s , Ann. of Math. 85 (1967), 58-159.
[22] G. S h i m u r a , Algebraic number fields and symplectic dis-
continuous groups, Ann. of Math. 86 0967), 503-592.
The l a s t section of [22] i s a weaker v e r s i o n of the r e s u l t s s t a t e d
in §lo, for which a full account will b e d i s c u s s e d in a forthcoming paper.
Some b a s i c concepts of I -adic r e p r e s e n t a t i o n s can b e found in
[23] Y. Taniyama, L-functions of number fields and zeta functions
of abelian v a r i e t i e s , J . Math. Soc. Japan, 9(1957), 330-366.
This v e r i f i e s a l s o the H a s s e conjecture f o r abelian v a r i e t i e s
with sufficiently many complex multiplications. F o r this topic, s e e
a l s o [14], and l e c t u r e notes by J. -P. S e r r e to b e published soon. The

You might also like