A NNALI DELLA
S CUOLA N ORMALE S UPERIORE DI P ISA Classe di Scienze
K EIJI M ATSUMOTO
On modular functions in 2 variables attached to a family of hyperelliptic curves of genus 3
Annali della Scuola Normale Superiore di Pisa, Classe di Scienze 4
esérie, tome 16, n
o4 (1989), p. 557-578
<http://www.numdam.org/item?id=ASNSP_1989_4_16_4_557_0>
© Scuola Normale Superiore, Pisa, 1989, tous droits réservés.
L’accès aux archives de la revue « Annali della Scuola Normale Superiore di Pisa, Classe di Scienze » (http://www.sns.it/it/edizioni/riviste/annaliscienze/) implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/conditions). Toute utilisa- tion commerciale ou impression systématique est constitutive d’une infraction pénale.
Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
of Hyperelliptic Curves of Genus 3
KEIJI MATSUMOTO
0. - Introduction
Let us consider a
family
F ofhyperelliptic
curvesof genus
3,
on the space ofparameters
For each curve
C (x, y),
we take a system{ B3 , Ak },1
3, ofbases of the
homology
group so that thecorresponding
6 x 6intersection matrix takes the canonical
form,
i.e. J =j o
-/3B
.
I3
07
Then we take three
linearly independent holomorphic
1-forms onC(x, y)
such that the
period
matrix takes the form(fl, 13).
This isalways possible
andwe get a of the
Siegel
upper half spaceH3
={ 3
x 3complex matrix QI
III =fl, Im fl
>0 }
of
degree
3. Now let(X, Y)
vary in A and let the basisdepend continuously
on(x, y).
Then thecorrespondence (x, y)
1 )y) gives
amulti-valued
map T :
A -->H3.
For a closedloop
6 in A with a fixedterminal
point Ao,
theanalytic
continuation of the restriction of T to asimply
connected
neighbourhood
ofAo along 6 gives
rise to asymplectic
transformationN ( ~ ) : ~ -; N ( S ) ~. In
this way we have ahomomorphism
of the fundamental group into the group ofsymplectic
transformations. Theimage
F is called themonodromy
group of the multivaluedmap T.
The purpose of this paper is as follows: to present the
image
as adomain of an
algebraic
set ofH3,
to describe the discrete group rarithmetically
Pervenuto alla Redazione il 7 Novembre 1988.
and to express the inverse
map V-1 : ~(A) 2013~
Aexplicitly
in terms of thetaconstants.
More
precisely,
we show in Section 1 that theimage
is an open dense subset of asubvariety
V inH3
which isbiholomorphically equivalent
tothe domain
where
An
explicit equivalence ,u
: D - V isgiven by ( 1.10).
Westudy
thecompound
system of generators of themonodromy
group G
of ~
isgiven by (2.5).
G is characterized as a congruencesubgroup
of the
unitary
groupGo - U(H, Z [i]), (see
Section2). By making
use of theembedding
and theta constantsdefined on
H3,
we express the inversemap BÍ1-1 :
D 2013~ A as rations ofproducts
of theta constants
(main theorem).
When we restrict the
parameters
on thecomplex
liney),
ourexpression
reduces to the classical Jacobi formulaconcerning
the so calledlambda function:
Let us
speak
about a relation withAppell’s system
of differentialequations
withparameters (a, (3, (3’ , 7).
This system is defined on A and admits threelinearly independent holomorphic
solutions at eachpoint
in A.Let us call the ratio of three
linearly independent
solutions aprojective
solution.It is known
([1], [12])
that there are 27quadruples
of parameters whichsatisfy
the condition:The
image
of A under theprojective
solution ofF1 (a, (3, ~Q’, 7)
is an opendense subset of a domain D’ c
C 2
which isprojectively equivalent
to the2-dimensional
complex
ballD,
and the inverse map of theprojective
solutionextends to a
single-valued holomorphic
map D --~ A.Among
the 27 cases, arithmetic characterization of themonodromy
group and anexpression
of the inverse map: D’ -~ A in terms of theta constants areknown
only
in two cases; one is studiedby
Picard[5], Holzapfel [3]
andShiga
[10],
and the other ispresented
in this paper, i.e. entries of the3-vector BÍ1(x, y)
are
linearly independent
solutions of the systemP’l 23 2 2 4 and T :
A - Dgives
aprojective
solution.The autor expresses his
gratitude
to Professors HironoriShiga
and MasaakiYoshida for their advices
during
thepreparation
of this paper.1. - The
periodic
map of thefamily
FLet us consider an
algebraic
curvewhere PI = C U
{oo}
and(x, y)
is apair
of parametersrunning through
Let be the
non-singular
model ofC’(x, y).
Westudy
thefamily
One
readily
knows from the Riemann-Hurwitz formula thatC(x, y)
is acurve of genus 3. We choose a basis of
holomorphic
1-forms as followsLet
P~ , Po2 ~, P12}
be thepreimages,
underthe
projection C(x, y)
-iC’(x, y),
of(x, 0), (y, 0)
and the threesingular points
~0, 0), ( 1, 0)
and(00,00), respectively.
The divisors of the
holomorphic
1-forms aregiven
as followswhere
(h)
stands for the divisor of a form or a function h.PROPOSITION 1.1 The curve
C(x, y), (x, y)
E A, ishyperelliptic.
PROOF. The divisor of a
meromorphic
functionf
=771/(173 - YrJ2)
onC(x, y)
isgiven by
which means that
f
is a map ofdegree
2. 0In the
following
we choose a basis of onCo ==
where we assume and 1 xo yo. We
regard Co
asa four sheeted cover over the
z-sphere;
let 1ro be theprojection Cho -
P1defined
by (z, w) --+
z. Letto (Im
to0)
be a fixedpoint
on thez-plane,
let71, ~ ~ ~ , 74 and 75 be line segments
connecting
to and z =0,1,
xo, yo and oo,respectively.
Let al, ~2, ~3 and ~4 be the four connectedcomponents
of1r-1
5
(z-sphere - U 1j).
j=l
Let p be the
automorphism
ofCo
definedby p ( z, w )
=( z, i w ) ,
wherei2 = -1. Here the
(J j , s
aresupposed
tosatisfy
= =1, 2, 3
andP(0’4)
= In order to recoverCo,
one has toglue
(7j andaj+2,j = 1, 2, along
11,12 and ~5, as well as (Jj andj
4,along
~3 and ~4, because the ramification indices of 7ro at =0,1, oo, j
=1, 2,
= x, y, 9respectively)
areequal
to2(4, respectively).
Let denote an orientedarc in cr~ from .P to
Q. Using
the abovenotations,
we define1-cycles Aj, Bk
on
Co
as followswhich are
displayed
inFigure
1. Intersection numbers of thecycles
form thefollowing
intersection matrix:Figure
1In order to have a basis of
HI (C, Z)
for ageneral
member C =C ( x, y )
ofthe
family F,
we take apath s joining Ao
=(xo, yo)
and( x, y )
in A and definea of
by
the continuationalong
s ; it ispossible
since thefamily
F is alocally
trivial fibre space over A. Notice that this choice of basesdepends
on thepath
s. Notice also that theautomorphism p
ofCo
is defined also ongeneral
C is an obvious manner, andp
operates
on the as followsNow we
integrate
the1-forms nj
=along
thecycles
the values will be denoted as follows
which reads for
example, c~2~~, y) = f (x, y).
SetB(x,y)
Since the
Ai’S
and theBk’s satisfy (1.5),
belongs
to theSiegel
upper half spaceH3
ofdegree
3, i.e. f1 issymmetric
andIm fl > 0. Hence we obtain the multi-valued map
PROPOSITION 1.2. The
map T:
A --~H3 is given by
where t Moreover we have
PROOF.
By
the relation(1.6)
we haveThese identities and the symmetry of fl lead to the first assertion
(1.7).
The second assertion
(1.8)
comes from theinequality
1m 0 > 0. 0 Notice that theinequality (1.8)
isequivalent
towhere
Let us define an
embedding
ofinto
H3 by
where u
= e1/0/eo,v
o= E2/E0 .
Eo SinceProposition
1.2says T(A)
cp(D),
we can definethe
by
Here we
briefly
recall thehypergeometric
systemF1 (a, {3, (3’ , ’1)
of lineardifferential
equations:
defined on A. The
integral representations
are known to be the Euler
integral representations
whichgive linearly independent
solutions ofF, (-! .1 -1 ’) (see [13]).
Thereforeby
the results21 23 23 4
obtained in
[1] ]
and[12],
we conclude that theimage ifi (A)
is an open dense in D(cf. [12])
andthat - I
can be extended to D as asingle-valued holomorphic
map onto P1 x
{ (0, o), ( 1,1),
Let us use the samenotation
for the extension
of T
onP 1
x P~ -{(O, 0), ( 1, 1 ) ,
Then theimage
x
P I - 1 (0, 0), ( 1,1), (oo, oo) })
isexactly
D.2. - The
monodromy
groupAny
element s of11’1 (A, Ào)
induces anautomorphism
~* ofas it is
explained
in Section 0. Let be the matrixrepresentation
of 6*relative to the basis
Ak ),
i.e.Because the transformation
N(b)
preseves the intersection matrix(1.5)
ofthe system it
belongs
toSp (3, Z ).
PutAccordingly,
al, a2 and a5 are transformed as follows:where
(in
view of(1.9))
Put
We take a system
1, ... , 5,
of generators of7r 1 (A, ao) represented
by
thefollowing loops:
a
loop
contained inL+
except for a smallpositively
oriented semi-circle inLyo
around x = 1,(x
= yo and 0,respectively);
d2(and d4 , respectively) :
a
loop
contained inL)
except for a smallpositively
oriented semi-circle inLao around y
= 0,(y
= oo,respectively),
whereAlong 0"
the branchpoints x and y
vary as are shown inFigure
2. Oncethe movement of branch
points
areknown,
a routine work leads to matricxesN(s~ ) and~),~=l, ,5):
The matrices and generate r and
G, respectively.
Figure
2In the
family
F of curvesy),
there are curves which areisomorphic.
In
fact,
if we defineautomorphisms
andl~3
of A as follows:and denote K the group
generated by them,
curvesC(x, y)
andC ( x’ , y’ )
are
isomorphic
if andonly
if(x_’, y’)
isequivalent
to(x, y)
under K. Letus meanwhile
consider
afamily
F ofisomorphic classes
with theparameter space
A
=A /K
andthe period map W : ¡ ~ H3 .
Themonodromy
group of the multi-valued
map T
is obtained as follows. In order T to bewell-defined,
we choose bases{ Bk , A~ ~
ofC ( x’, y’ ) ,
which isK-equivalent
toso that
Since we
have the exact sequence 1 -~1r(A, Ao) c 1r(A, ¡o)
--+ K --+ 1, the group1r(A, ¡oJ
isgenerated by 1r(A, Ao)
andloops
inA
of which lifts are arcsin A
joining Ao
and itsK-equivalent points.
Let66,67
and68
be arcsjoining Ao
andk3 (Ao), respectively.
Then themonodromy
group of~ is
generated by
that of T and matricesN ( ~3 ) , j
=6, 7, 8,
which are definedby
Accordingly,
al, a2 and a5 are transformed as follows:Let us take the
8j’ s
as follows:Then
N (6j)
and =6, ?, 8,
are known to beREMARK 2.1. Matrices
N ( ~3 ) , j
=1, ~ ~ ~ , 6, belong
to the group{ N
ES p ( 3, ~ ) : diagonal
elements of t AC and t B D areeven},
studiedby
J.Igusa;
so that r cr 12 .
We set
The
single-valued map :
A -DIG
extends to the map P, x P, -{(O, 0), (1, 1),
->D/G,
which is known to bebiholomorphic (cf. [12]).
The transformation group G has three cups which are
represented by
PROPOSITION 2.2.
If
7r denotes theprojection of Go
ontoGo
modulo itscenter, then
,, ,
PROOF.
(1).
Since we haveGI c G 1,
there is a naturalprojection
p :D/G~
to
D /Gi .
LetAut(D)
be the group ofholomorphic automorphisms
and let Mbe the
isotropic subgroup
ofAut(D)
relative to[ço, Ç1, Ç2] = [0,1,0].
Then Mis
given (cf. [6]) by
Hence
GIn M
is thetotality
of transformations of thefollowing
type:where m, n,
b, v
m - n = b mod 2. It turns out thatG1
n M isgenerated by g(0, 0, 0, 1)
andg(1, 1, 1, 0).
Sinceg(0, 0, 0, 1)
andg(1, 1, 1, 0) belong
toGi,
we have
GI
n M =G1
n M. Therefore theprojection
p is atopological
cover.By
astraightforward
calculation one knowsthat 1, 0, 0~, [0,1,0]
and~0,1,1~
arenot
G1-equivalent.
Hence p is a cover ofdegree
1.(2).
Since we haveGo
cGo,
there is a naturalprojection p’ :
D / Go . Go
n M is thetotality
of transformations of thefollowing
type:g(m,
n,b, v)
where m, n,b, v E Z,
m = r~ mod 2. It turns out thatG1
n M isgenerated by g (0, 0, 0, 1) , g ( 1, 1, 1, 0)
andg (0, 0, 1, 0) .
Sincethey belong
toGo,
we haveGI n M = Go n M.
Therefore the
projection p’
is atopological
cover. Hencep’
is a cover ofdegree
1.(3).
It is easy to check thatg (66)
E G andg(66)2 = g ( ~3 ) E
G(cf.
[11]).
D3. - An
expression of iii by
theta constantsLet us recall some basic facts on the Riemann theta function:
where z =
t (zi, ... ,
EHg.
It isholomorphic
on C 9 xHg
andsatisfies
period
relations:where
For column vectors p and q of
(Z j2)g
the theta function with acharacteristic
[22 2 tq_
definedby
The function e
2 tP’ gj (£1)
:= e"2 tP (0, (1)
is called a theta constant.If m and n are increased
by
evenintegral
vectors, tm(z, fl) hardly changes:
where m,
m’, n, n’
REMARK 3.1. The function
[tt
n(z, n)
of z is even(odd),
if t mnis even
(odd), respectively.
Inparticular
if t mn isodd, then 6 tn (0, n)
vanishes.
vanishes.
Next let us consider a
compact
Riemann surface X of genus g. We taka basis of so that the
corresponding
intersection matrix takes the canonical form J. Then we takelinearly independent holomorphic
1-formswj j
=1, ~ ~ ~ ,
g, on X such that theperiod
matrix takes the form(0, Ig).
REMARK 3.2. If X is a
hyperelliptic
curve of genus 3 and itsperiod
matrix is
(0, 13),
then there isonly
one characteristic I mod 2Zsuch that t mn is even and In case : we will
see in
Proposition
4.5 thatWe set w
== t(W1"" c~9 ) .
For fixed z and fixed X we definea multi-valued function on X
by
Let us recall the celebrated Abel’s Theorem.
THEOREM.
Suppose
are divisors on Xof
samedegree.
If
we havethen there is a
meromorphic function f
on X withpoles
and zerosSince we need
explicit
form off
later under oursituation,
wegive
a way to constructf by using
the - function B.h(z; P).
/ We can choose B. e E so that=0 and that h
d,
do notvanish
identically. Consider
thefollowing function on .X:
where the
paths
ofintegration
are chosen so thatand that those
joining Po
and P in the numerator and in the denominator aresupposer to be the same. Then
f
issingle-valued
and hasrequired
zeros andpoles.
Applying
this constructionby taking C (x, y)
for X andPy
forPo,
wegive
anexpression
ofq,-1 by
theta constants. Letf
be theprojection
Because we have , two divisors
satisfy
the condition:Moreover we have
where
Pk3
= andPk 4
= =0, or.
Here thesymbol (j)
attached to thesign
ofintegral
stands for apath
ofintegration By applying
the aboveconstruction,
the functionf
has thefollowing expression:
where e and the
paths
ofintegration
are supposer tosatisfy
the conditions mentionedabove,
and r, is a constantdepending
on e.If we take
Px, Py, Pl1
andPl 2
for P in(3.5),
we obtain thefollowing
equalities:
LEMMA 3.3. We have
THEOREM 3.4. The
(u, v) ’2013~ (x, y)
E A has anexpression
in terms
of
theta constants asfollows:
where
1 0
PROOF. If we take
e 1 + -1 1
, then we haveO ( e 1, H)
= 0by
Remark 3, I . For this ei we2 2
neither the numerator nor the
0
U7
by
Remark 3.1. For this e 1 we have to show that neither the numerator nor the denominator of(3.5)
vanishesidentically. Using (3.1), (3.2), (3.3)
and Lemma3.3,
we can express the numerator and denominator of(3.6)
and(3.9) by
theproduct
of even theta constants and non-zerofactors. Then neither the numerator nor the denominator of(3.5)
vanishesidentically
in view of Remark 3.2. Ifwe eliminate r,, 1 from
(3.6)
and(3.9),
then we obtain the desiredpresentation
1 0
of
x ( u, v ) .
If we takee 2 = Q (
1+
1 for ei, then we obtain thepresentation
of y(u, V).
2
(111 )+ U7 2 (0) W D
presentation
ofy ( u, v ) .
DCOROLLARY 3.5. We have
PROOF. If we take e and use
(3.8)
and(3.9),
thenwe obtain the relation.
REMARK 3.6. In the next section we shall find more
precise
relationsamong the O
/t u v))’s,
where m, ne Z 3
and(u,v)
e D.L tn - ))
4. - Modular forms induced
from q,-1
Let
fjJ1
andQ2 (respectively Q3
andQ4)
stand for numerator anddenominator of
(3.10) (respectively (3.11 )).
In this section we show thatfjJ 1, fjJ2, fjJ3
andfjJ4
are modular forms relative to themonodromy
group G.A
holomorphic function Q
onis called a modular form of
weight
l~ relative toif it satisfies the condition
for any =
(g;k)
EG, 1 j, k
3. Aholomorphic function 0
onH3
is called a
Siegel
modular form ofweight
k relative toSp(3,Z)/(±I6)
if itsatisfies the condition
for any 0 E
H3
and N =(C D)
E Via theembedding
M: D --+ V c
H3
andAut(D)
cAu (H3)
weregard
modular forms ofweight
2k on D as those of
weight k
onH3.
Let us recall the transformation formula of theta constants(see [4]):
where
depends only
on j(not
on p, qand 0).
THEOREM 4.1. The
functions
=1, ... ,
4, are modularforms of weight
8 relative to G.
PROOF. We show that
Oj, ,1 j
4,satisfy (4.1 )
with respect tog ( ~~ )
EG,1
k 5. Since a direct calculation leads towhere
we show
Since
N ( sk ) 1
k 5,belong
toFi ,
we canapply (4.3)
toj
4.By
a routine argument it turns out that we haveonly
to show thefollowing
lemma to finish the
proof
of Theorem 4.1.LEMMA 4.2. We have
PROOF OF LEMMA 4.2.
Step
1.By using (3.2),
we obtain the Fourierexpansion
Step
2. Sublemma 4.3.Proof of
Sublemma 4.3.Step
3. Sublemma 4.4.Proof of
Sublemma 4.4.By using (4.3)
forN(Ó6)
we obtainwhere
By
Sublemma the aboveequality
reduces
to
Let us determine the factor
By (4.4)
we haveAs the constant term of the above series does not
vanish,
e
0 0 0 (u( u, v) )
does not vanishidentically.
If we put Pi = qi =0, .0
00-
j
=1, 2, 3,
in(4.6),
we obtain = 1. DSubstituting explicit
values in(4.5)
we obtain the formulae in Lemma4.2. D
The
following
fact which is announced in Remark 3.2 follows from Sublemma 4.4.PROPOSITION 4.5. We have
PROOF. If we put
p3 - q~ - 2 , j
=1, 2,
and p3 - q3 - 0 in Sublemma4.4,
we obtainREFERENCES
[1]
P. DELIGNE - G. D. MOSTOW, Monodromy of hypergeometric functions and non-latticeintegral monodromy, Publ. Math I.H.E.S., 63 (1986), pp. 5-88.
[2]
T. KIMURA, Hypergeometric Functions of Two Variables, Lecture Notes, Tokyo Univ., (1973).[3]
R. P. HOLZAPFEL, Zweidimensionale periodische Funktionentheorie der Kugel,Akademie der Wissenschaften der DDR, Berlin, (1983).
[4]
D. MUMFORD, Tata Lectures on Theta I-II, Birkhauser, Boston-Basel-Stuttgard, (1983).[5]
E. PICARD, Sur les fonctions de deux variables independentes analogues aux fonctions modulaires, Acta Math., 2 (1883), pp. 114-135.[6]
I. I. PJATECKII-SAPIRO, Geometry of Classical Domains and Automorphic Functions, Fizmatgiz, Moscow, (1961).[7]
H. E. RAUCH - H. M. FARKAS, Theta Functions with Applications to Reimann Surfaces, Williams and Wilkins, Baltimore, (1974).[8]
H. L. RESNIKOFF - Y. S. TAI, On the structure of a graded ring of automorphic forms on the 2-dimensional complex ball, Math. Ann. 238 (1978), pp. 97-117.[9]
H. SHIGA, One attempt to the K3 modular function I-II, Ann. Scuola Norm. Pisa, Serie IV-VI (1979), pp. 609-635, Serie IV-VIII (1981), pp. 157-182.[10]
H. SHIGA, On the representation of the Picard modular function by theta constants I-II, to appear in Pub. R.I.M.S. Kyoto Univ., 24 (1988).[11]
T. TERADA, Problème de Riemann et fonctions automorphes provenant des fonctions hypergeometriques de plusieurs variables, J. Math. Kyoto Univ., 13 (1973), pp.557-578.
[12]
T. TERADA, Fonctions HypergeometriquesF1
et fonctions automorphes I-II, J. Math.Soc. Japan, 35 (1983), pp. 451-475, 37 (1985), pp. 173-185.
[13]
M. YOSHIDA, Fuchsian Differential Equations, Vieweg, Verlag, Wiesbaden, (1987).[14]
I. WAKABAYASHI, Note on Picard’s modular function of two variables, private note.Kyushu University
Department of Mathematics
Faculty of Science
Fukuoka 812 Japan