Jacobian Nullwerte and Algebraic Equations: Jordi Guàrdia

Journal of Algebra 253 (2002) 112–132
www.academicpress.com
Jacobian nullwerte and algebraic equations ✩

Jordi Guàrdia
Escola Universitària Politècnica de Vilanova i la Geltrú, Av. Víctor Balaguer s/n,
E-08800 Vilanova i la Geltrú, Spain
Received 20 January 2001
Communicated by Michel Broué
Abstract
We present two applications of jacobian nullwerte, both related with the resolution of
algebraic equations of any degree. We give a very simple expression of the roots of a
polynomial of arbitrary degree in terms of derivatives of hyperelliptic theta functions. This
expression can be understood as an explicit proof of Torelli’s theorem in the hyperelliptic
case. We also give geometrical expressions of the discriminant of a polynomial. Both
applications are based on a jacobian version of Thomae’s formula.  2002 Elsevier Science
(USA). All rights reserved.
Keywords: Algebraic equations; Theta functions
1. Introduction
In the last decades of the 19th century, Rosenhain [1], Thomae [2], Rie-
mann [3] and Frobenius [4] began the study of determinants of derivatives of odd
theta functions, in order to find generalizations of Jacobi’s derivative formula [5]:

θ11 (0, τ ) = −πθ00(0, τ )θ01(0, τ )θ10(0, τ ).
They discovered fascinating formulae, expressing the value at zero of these
jacobians (“jacobian nullwerte”) as products of zero values of even Theta
functions (“Thetanullwerte”). For much of the twentieth century, these jacobian
✩
Partially supported by DGICYT Grant BFM2000-0627.
E-mail address: guardia@mat.upc.es.
0021-8693/02/$ – see front matter  2002 Elsevier Science (USA). All rights reserved.
PII: S 0 0 2 1 - 8 6 9 3 ( 0 2 ) 0 0 0 4 9 - 2
J. Guàrdia / Journal of Algebra 253 (2002) 112–132 113
nullwerte seemed to be forgotten, until Weil asked Igusa whether any jacobian
nullwerte could be expressed as a polynomial in the even Thetanullwerte, or at
least as a rational function of them. Igusa showed that the second assertion is true,
but not the first. He also used the jacobian nullwerte to show the non-integral
closeness of the ring of theta constants (cf. [6,7]).
In this paper we present two applications of jacobian nullwerte, both related
with the resolution of algebraic equations.
It is well known that it is possible to express the roots and the discriminant of
a polynomial in terms of hyperelliptic theta functions and hyperelliptic integrals
(cf. [8,9]). These formulae are rather intricate, and the geometry behind them is
not evident. Their proofs rely on the classical formula of Thomae, which relates
some products of differences between the roots of a polynomial with a given
Thetanullwerte.
We present here a more geometrical approach to the resolution of algebraic
equations. Using Riemann’s theorems on the theta function, we are able to give a
very simple expression of the roots of a polynomial in terms of derivatives of odd
theta functions. Geometrically speaking, we give an effective method to recover
a hyperelliptic curve from its jacobian, that is, we provide an explicit proof of
Torelli’s theorem in the hyperelliptic case.
As a second application of the jacobian Thetanullwerte, we discuss the
expression of the discriminant of a polynomial as a value of a Siegel modular
form. We propose formulas which give a geometrical content to the classical
definition of the discriminant as a modular form.
2. Preliminaries
2.1. Notations
Throughout the paper, C will be a smooth curve of genus g > 1 defined

over the complex field C. Given a divisor D on C, [D] will denote its linear
equivalence class; we will write l(D) = h0 (C, O(D)), and the sign ≡ will be
used to express the linear equivalence of divisors on C.
Given a separable polynomial f (X) ∈ C[X] of degree n, we will denote by Cf
the genus g = (n − 2)/2 hyperelliptic curve given by
Cf : Y 2 = f (X) = an Xn + · · · + a0 = an (X − α1 ) · · · (X − αn ).
In fact, the hyperelliptic curves that we will work with will always be of this
type, so that the symbol Cf will always stand for a hyperelliptic curve associated
with certain polynomial f (X). We will denote by Wi = (αi , 0), i = 1, . . . , n its
Weierstrass points.
114 J. Guàrdia / Journal of Algebra 253 (2002) 112–132
For α1 , . . . , αg ∈ C, let V (α1 , . . . , αg ) denote the Vandermonde matrix:

 
1 1 ··· 1
 α1 α2 · · · αg 
 
 .. .. ..  ,
 . . . 
g−1 g−1 g−1
α1 α2 · · · αg
whose determinant is

V (α1 , . . . , αg ) = (−1)g(g−1)/2 (αj − αk ).
j <k
The symbol diag(x1 , . . . , xn ) will stand for the matrix whose diagonal entries are
x1 , . . . , xn and remaining entries are zero.
We begin by recalling the main classical results that will be used in the
paper. There are many excellent books [5,10–12] which can be used as general
references for the rest of the paper.
2.2. The Abel–Jacobi map
Let C be a complex smooth curve of genus g > 1. We pick a symplectic basis

γ1 , . . . , γ2g of the singular homology H1 (C, Z) of C, and a basis ω1 , . . . , ωg of
H 0 (C, Ω 1 ), the space of holomorphic differentials on C. The period matrix of C
with respect to these bases is the g × 2g matrix
 
γ1 ω1 · · · γ2g ω1
 .. 
Ω =  ... .
.
γ1 ωg · · · γ2g ωg
We write Ω = (Ω1 |Ω2 ) to denote by Ω1 the first half of Ω (that is, the matrix
formed by the first g columns of Ω), and by Ω2 the second half. With these data,
we can consider a second basis of H 0 (C, Ω 1 ), given by
   
η1 ω1
 ..  = Ω −1  ...  . (1)
. 1
ηg ωg
The period matrix of C with respect to this new basis is (1g |Z), where 1g denotes
the g × g identity matrix and Z belongs to the Siegel upper half space Hg , i.e., it
is a symmetric matrix whose imaginary part is positive definite (these are called
the Riemann conditions). A matrix satisfying the Riemann conditions is called
a normalized period matrix. The jacobian variety J (C) of the curve C is the
complex torus J (C) := Cg /(1g |Z); the Riemann conditions ensure that this torus
is projective, which implies that J (C) is an abelian variety.
Let us denote by Cd the d-fold symmetric product of C, which can be identified
with the set of effective divisors of degree d on C. We fix once for all a Weierstrass
point W ∈ C as base point for the Abel–Jacobi map in degree d, which is defined
by
ud : Cd → J (C),

Pi
D= mi Pi → ud (D) := mi (η1 , . . . , ηg ).
i i W
Of course, the values of the integrals depend on the path chosen to integrate, but
the class of ud (D) on Cg /(1g |Z) is well-defined. Unless explicitly stated, the
choice of the base point will not affect any of the results, so we will make no
further reference to it.
It is usual to extend the Abel–Jacobi map ud to non-effective divisors: if

d+r
r
D= Pj − Qk ,
j =1 k=1
we take
d+r

Pi r

Qk
ud (D) = (η1 , . . . , ηg ) − (η1 , . . . , ηg ).

j =1 W k=1 W
Abel’s theorem tell us that the Abel–Jacobi map is invariant through linear
equivalence:
D1 ≡ D2 ⇒ ud (D1 ) = ud (D2 ).
2.3. Riemann theorems
For z ∈ Cg , we define the Riemann theta function of J (C) by

θ (z) = θ (z; Z) = exp πi t n.Z.n +t n.z .
n∈Zg
This function is quasi-periodic with respect to the columns of the period matrix
(1g |Z). Although it cannot be considered as a function on J (C), its zero locus is a
well-defined divisor Θ in J (C), called the theta divisor. Two fundamental results
of Riemann relate the geometry of Θ with the geometry of C.
Theorem 2.1 (Riemann theorem). The image of the Abel–Jacobi map in degree
g − 1 is a translate of the theta divisor: for a suitable κ ∈ J (C) we have
ug−1 (Cg−1 ) = Θ − κ.
The constant κ, which is usually called the Riemann vector, of course depends
on the base point W . We define a new map Π : Cg−1 → J (C), given by Π(D) =
ug−1 (D) + κ. This map satisfies the symmetry condition: Π(K − D) = −Π(D),
where K denotes a canonical divisor. The map Π is extended to non-effective
divisors of degree g − 1 in the same way as the map ug−1 .
Abel’s theorem assures that around an effective divisor D of degree g − 1 such
that l(D) = 0, the Abel–Jacobi map ug−1 is one-to-one, so by Riemann theorem,
the theta divisor is smooth at the point ug−1 (D). The Riemann singularity theorem
explains what happens for exceptional divisors of degree g − 1.
Theorem 2.2 (Riemann singularity theorem). For every effective divisor D of

degree g − 1, the multiplicity of the theta divisor Θ at the point Π(D) is l(D):
multΠ(D) Θ = l(D).
We are mainly interested in the case of effective divisors D with l(D) = 1. For
such a divisor, Riemann singularity theorem says that the function θ vanishes at
point Π(D), and that at least one of the first derivatives is non-zero at Π(D).
3. The canonical map
The choice of the basis ω1 , . . . , ωg of the holomorphic differential forms on C

provides a map from C to Pg−1 = PH 0 (C, Ω 1 )∗ , given by
φ : C → Pg−1 ,

P → φ(P ) = ω1 (P ), . . . , ωg (P ) .
This is the canonical map of C. Note that if the curve C (and the differential
forms ωj ) is defined over a number field K, the canonical map is also defined
over this field.
such that the divisor D = P1 + · · · +

Proposition 3.1. Let P1 , . . . , Pg−1 ∈ C(K)
Pg−1 satisfies l(D) = 1. The equation:
 
X1
∂θ ∂θ .
HD (X1 , . . . , Xg ) := Π(D) , . . . , Π(D) Ω1−1  ..  = 0
∂z1 ∂zg
Xg
determines a hyperplane HD of Pg−1 , which cuts the divisor φ(D) on the
curve φ(C).
Proof. In [10, p. 228] is seen that the points Pi satisfy the equation

g
∂θ
Π(D) ηk (P ) = 0,
∂zk
k=1
and the proposition follows from the relation (1). ✷
Corollary 3.1. If the curve C is defined over a number field K, any quotient
of the coefficients of the equation in Proposition 3.1 is algebraic over the field
K(P1 , . . . , Pg−1 ).
Proof. The hyperplane HD ⊂ Pg−1 determined by the points P1 , . . . , Pg−1 is

defined over the field K(P1 , . . . , Pg−1 ), so that any equation having a coefficient
equal to one must have all its coefficients in this field. ✷
4. Theta characteristics on hyperelliptic curves. Two-torsion points
A semicanonical divisor on a curve of genus g is a divisor D of degree

g − 1 such that 2D is linearly equivalent to a canonical divisor. A geometric
theta characteristic is the class of a semicanonical divisor. It is well known that
there are 4g different theta characteristics. A semicanonical divisor D (and the
corresponding geometric theta characteristic) is called even or odd according to
the parity of l(D).
For a hyperelliptic curve, all these classes can be described with divisors
supported on the Weierstrass points.
Lemma 4.1 [10]. Let Cf be a hyperelliptic curve of genus g defined over an

algebraically closed field, with Weierstrass points W1 , . . . , W2g+2 . Given a set of
indices I = {k, i1 , . . . , ig−1−2m }, denote by DI the divisor 2mWk + Wi1 + · · · +
Wig−1−2m .
(a) Any semicanonical divisor on Cf is linearly equivalent to one of the form DI

for some set of indices I with −1 m (g − 1)/2 and ir = is for r = s.
(b) Two semicanonical divisors 2mWk + Wi1 + · · · + Wig−1−2m and 2nWk +
Wj1 + · · · + Wjg−1−2n are linearly equivalent if and only if m = n = −1 and
{i1 , . . . , ig+1, j1 , . . . , jg+1 } = {1, 2, . . . , 2g + 2}.
(c) l(DI ) = m + 1.
(d) If l(DI ) = 1 then DI is the sum of g − 1 Weierstrass points on Cf .
We introduce the following definition.
Definition 4.1. A Weierstrass divisor on Cf is a semicanonical divisor satisfying

condition (d) of Lemma 4.1.
The set J (Cf )[2] of 2-torsion points in J (Cf ) = Cg /(1g |Z) can be described
as

1 1 m
J (Cf )[2] = zm = Z.m + m m = , m , m ∈ {0, 1} .
g
2 2 m
The point zm is said to be even (respectively odd) if m .m ≡ 0 (mod 2)

(respectively ≡ 1 (mod 2)). The set of 2-torsion points of J (C) splits into two
disjoint subsets Σ + and Σ − of even and odd 2-torsion points, respectively.
The image of a geometric theta characteristic [D] through the Abel–Jacobi
map Π introduced in Section 2.3 is a 2-torsion point of the jacobian of the curve,
since −Π(D) = Π(K − D) = Π(D) for any canonical divisor K. In fact, the
Abel–Jacobi map establishes a bijection between the set of theta characteristics
and the set of 2-torsion points of the jacobian, sending even (respectively odd)
geometric theta characteristics to even (respectively odd) 2-torsion points.
Definition 4.2. A Weierstrass 2-torsion point is an element of J (Cf )[2] which is

the image through the Abel–Jacobi map of the class of a Weierstrass divisor.
Riemann theorem, combined with part (d) of Lemma 4.1, gives a complete
characterization of the Weierstrass 2-torsion points: they are the odd 2-torsion
points w ∈ J (Cf )[2] such that

∂θ ∂θ
(w), . . . , (w) = (0, . . . , 0).
∂z1 ∂zg
The bijection between geometric theta characteristics and 2-torsion points of

the jacobian can be made completely explicit, once we have fixed an ordering
of the roots of the polynomial f (X) which defines the hyperelliptic curve Cf .
This is explained in [12, pp. 3.75–3.88]. For the sake of completeness, we will
briefly outline the process. We take the standard basis for H1 (Cf , Z) described
in loc. cit., p. 3.76. We compute the period matrix Ω = (Ω1 |Ω2 ) of some basis
of H 0 (Cf , Ω) with respect to this homology basis, and take the corresponding
normalized period matrix Z = Ω1−1 Ω2 (observe the change of notation with
respect to Mumford). For i = 1, . . . , g + 1, let
t
0 ··· 0 1 0 ··· 0 η2i−1
η2i−1 = = ,
1 ··· 1 0 0 ··· 0 t η
2i−1
t
0 ··· 0 1 0 ··· 0 η2i
η2i = = ,
1 ··· 1 1 0 ··· 0 t η
2i
where the non-zero

1 · · · entry in the first row is at the ith column in both cases
1
2g+2 = 1 · · · 1 ). Now, for any set S ⊂ {1, 2, . . . , 2g + 2} we define
(and η
ηS := i∈S ηi (mod 2) and
1 1
zi := Z.ηi + ηi , zS := zi .
2 2
i∈S
We take the point W2g+2 as base point for the Abel–Jacobi map. With these
choices, the image of a semicanonical divisor DI = 2mWk + Wi1 + · · · + Wig−1−2m
through the Abel–Jacobi map is
ug−1 (DI ) = zI ,
and the Riemann vector is

1  
g
..
1 . 1 g − 1
κ = z{1,3,...,2g+1} = Z. 
 .. 
 +  . ,
2 . 2  .. 
1 1
so that
wI := Π(DI ) = zI ◦U .
Finally, we have the following relation:
Π(DI ) + Π(DJ ) = Π(DI ◦J ◦U ), (2)
where U = {1, 3, . . . , 2g + 1} and S ◦ T denotes the symmetric difference of two

sets S, T .
For the rest of the paper, we will assume that we are working in this frame, in
particular with the choice of the standard basis for H1 (Cf , Z).
5. Jacobian nullwerte. Fundamental systems

t
We will call analytic theta characteristic a 2×g matrix η = t ηη with entries
in {0, 1}. We say that η is even (respectively odd) if η .η ≡ 0 (mod 2))
(respectively ≡ 1 (mod 2). We denote by TCg (respectively TC+ −
g , TCg ) the set of
all (respectively even, odd) analytic theta characteristics of dimension g.
Every analytic theta characteristic η defines a theta function with characteris-
tic:

1 t 1
θ [η](z; Z) := exp πi η .Z.η + πiη .(z + η )
4 2

1 1
× θ z + Z.η + η ; Z .
2 2
The values of these functions at z = 0 are called Thetanullwerte. It is common
to write θ [η](Z) instead of θ [η](0; Z). Thetanullwerte have been widely studied,
and there are many excellent references for them. But we are now interested on
the values at zero of the derivatives of the theta functions with characteristics:
Definition 5.1 [7]. A jacobian nullwerte is a determinant of the form:

 ∂θ [η1 ] ∂θ [η1 ] 
(0; Z) · · · (0; Z)
 ∂z1 ∂zg 
 
[η1 , . . . , ηg ](Z) := det  ,
.
. ..
 . . 
 ∂θ [η ] ∂θ [ηg ] 
g
(0; Z) · · · (0; Z)
∂z1 ∂zg
where η1 , . . . , ηg are different odd analytic theta characteristics.
Jacobian nullwerte can be thought of as (modular) functions over the Siegel

upper half space. Rosenhain, Thomae, Frobenius and Riemann obtained beautiful
results for jacobian nullwerte in low genera. Most of these results can be
understood as generalizations of Jacobi’s derivative formula. Igusa [6,7] studied
the Jacobian nullwerte for general genus, in connection with a problem raised by
Weil about the integral closeness of the ring of the Thetanullwerte. We recall one
of Igusa’s results about Jacobian nullwerte.
For an analytic theta characteristic η, we put e(η) = 1 if η is even, and
e(η) = −1 if η is odd. Given three analytic theta characteristics η1 , η2 , η3 we
define
e(η1 , η2 ) = e(η1 )e(η2 )e(η1 + η2 ),
e(η1 , η2 , η3 ) = e(η1 )e(η2 )e(η3 )e(η1 + η2 + η3 ).
We say that the triplet {η1 , η2 , η3 } is azygetic (respectively zygetic) if e(η1 , η2 , η3 )
is −1 (respectively 1). A fundamental system of analytic theta characteristics
is a set S = {η1 , . . . , η2g+2 } of 2g + 2 analytic theta characteristics such that
η1 , . . . , ηg are odd, ηg+1 , . . . , η2g+2 are even and every triplet {ηi , ηj , ηk } ⊂ S
is azygetic. The last condition is equivalent to e(η1 + ηi , η1 + ηj ) = −1 for every
pair i = j . The analytic theta characteristics in a fundamental system are always
essentially independent, which means that every sum of an even number of them
is non-zero.
Let C[θ ] be the “ring of Thetanullwerte”, which is the C-algebra generated
by the functions θ [η](Z), where η runs over the set of even analytic theta
characteristics.
Theorem 5.1 [6]. Let η1 , . . . , ηg be odd analytic theta characteristics such that
the function [η1 , . . . , ηg ](Z) is not identically zero and is contained in the ring of
Thetanullwerte C[θ ]. Then η1 , . . . , ηg can be completed to form a fundamental
system, and

2g+2
[η1 , . . . , ηg ](Z) = π g ± θ [ηi ](0; Z),
{ηg+1 ,...,η2g+2 }∈S i=g+1
where S is the set of all (g + 2)-tuples {ηg+1 , . . . , η2g+2 } of even theta

characteristics such that {η1 , . . . , ηg , ηg+1 , . . . , η2g+2 } form a fundamental
system. If Z is the period matrix of a hyperelliptic curve, there is exactly one
non-zero term in the sum of the right hand side of the equality.
We are especially interested in the values of jacobian nullwerte at hyperelliptic

period matrices, i.e., at matrices Zf ∈ Hg which are the period matrix of a
hyperelliptic genus g curve Cf . The complex abelian variety Cg /(1g |Z) is then
the jacobian J (Cf ) of Cf , and the analytic theta characteristics can be naturally
tied with geometric theta characteristics, following the process explained in
Section 4. When we work in this context, we apply the terminology introduced
for analytic theta characteristics to semicanonical divisors and to 2-torsion
points on J (Cf ). For instance, the parity of a 2-torsion point w ∈ J (Cf ) is
denoted by e(w), and e(w1 , w2 , w3 ) is defined in the obvious way for three
2-torsion points w1 , w2 , w3 ∈ J (Cf ). A fundamental system of 2-torsion points
in J (Cf ) (respectively of semicanonical divisors in Cf ) is a set of 2-torsion
points (respectively of semicanonical divisors) whose corresponding analytic
theta characteristic form a fundamental system. With this in mind, we define the
following specific jacobian nullwerte:
Definition 5.2. Let D1 , . . . , Dg be g Weierstrass divisors on a hyperelliptic

genus g curve Cf , with period matrix Ω = (Ω1 |Ω2 ). Let Zf = Ω1−1 Ω2 and write
wi = Π(Di ).
 ∂θ ∂θ 
(w1 ; Zf ) · · · (w1 ; Zf )
 ∂z1 ∂zg 
 
J (D1 , . . . , Dg ) = J (w1 , . . . , wg ) := 

.. .. .

 ∂θ . ∂θ
. 
(wg ; Zf ) · · · (wg ; Zf )
∂z1 ∂zg
6. Fundamental systems for hyperelliptic curves
Let Cf be a hyperelliptic curve of genus g, with Weierstrass points W1 , . . . ,

W2g+2 . We will show a geometric way of constructing fundamental systems
on J (Cf ). We pick a set of g Weierstrass points, which we assume to be
W1 , . . . , Wg to simplify. We define

g
D = Wi ,
i=1
Di = D − Wi , i = 1, . . . , g,
Di = D + Wi − 2W2g+2 , i = g + 1, . . . , 2g + 1,
wi = Π(Di ).
Proposition 6.1. The set w1 , . . . , w2g+2 is a fundamental system of 2-torsion

points in J (Cf ).
Proof. By Proposition 4.1, the divisors Di have the right parity, so we have
only to check that e(w1 + wi , w1 + wj ) = −1 for every pair i = j . Let U =
{1, 3, . . . , 2g + 1}. Applying formula (2) we find that wi + wj = Π(Di ) +
Π(Dj ) = Π(D{i,j }◦U ). Hence, e(wi + wj ) = (−1)i+j +1 , and e(w1 + wi , w1 +
wj ) = (−1)2i+2j +1 = −1. ✷
7. Jacobian nullwerte on hyperelliptic curves
Let
Cf : Y 2 = f (X) = an (X − α1 ) · · · (X − αn ) ∈ C
be a complex hyperelliptic curve of genus g = (n − 2)/2 . We fix
dx x dx x g−1 dx
ω1 = , ω2 = , ..., ωg =
y y y
as the basis of the holomorphic differential forms and denote by Ω = (Ω1 |Ω2 )
the period matrix of Cf with respect to this basis of H 0 (Cf , Ω 1 ) and some
symplectic basis of H1 (Cf , Z). The canonical map of Cf with respect to
ω1 , . . . , ωg has a very simple expression:
φ : Cf → Pg−1 ,

P = (x, y) → φ(P ) = 1, x, . . . , x g−1 .
Theorem 7.1. Let W1 = (α1 , 0), . . . , Wg = (αg , 0) ∈ Cf (C) be g different

Weierstrass points on Cf . Take Di = j =i Wj , wi = Π(Di ) and let HDi be the
hyperplane associated to Di by Proposition 3.1. Then

(a) J (w1 , . . . , wg )Ω1−1 V (α1 , . . . , αg ) = diag HD1 (α1 ), . . . , HDg (αg ) . (3)
(b) For any Weierstrass point P = (α, 0) ∈ i , . . . , Wg }, the quotients
/ {W1 , . . . , W

1 ∂θ ∂θ
(α) Π(Di ) · · · Π(Di ) Ω1−1
HDi ∂z1 ∂zg
belong to the field Q(α1 , . . . , α̂i , . . . , αg , α).
Proof. The divisors Dk satisfy l(Dk ) = 1, so that the corresponding hyper-

planes HDi are well-defined, and cut the divisor φ(Dk ) on φ(Cf ), by Proposi-
tion 3.1. This is the content of (a). As for (b), it is true for P = Wi since

diag HD1 (α1 )−1 , . . . , HDg (αg )−1 J (w1 , . . . , wg )Ω1−1 = V (α1 , . . . , αg )−1 ,
which is a matrix with entries in the field Q(α1 , . . . , αg ). If we replace Wi by any

Weierstrass point different from the remaining Wk , the divisor Di will not change,
and hence the hyperplane HDi will be exactly the same. ✷
Taking determinants in formula (3), we obtain the following corollary, which

could be considered as a jacobian version of Thomae’s formula (cf. Section 11).
Corollary 7.1.
g

V (α1 , . . . , αg ) = det Ω1 g−1
HDj 1, αj , . . . , αj J (w1 , . . . , wg ).
j =1
8. Solving algebraic equations with jacobian nullwerte
Let f (X) = an (X − α1 ) · · · (X − αn ) ∈ C[X] be a polynomial of arbitrary

degree n > 4 with different roots. We will express the roots α1 , . . . , αn of f in
terms of jacobian nullwerte on the jacobian of the hyperelliptic curve
Cf : Y 2 = f (X).
As in the previous section, we fix
dx x dx x g−1 dx
ω1 = , ω2 = , ..., ωg =
y y y
as basis of H 0 (Cf , Ω 1 ). Let Ω = (Ω1 |Ω2 ) the period matrix of ω1 , . . . , ωg

with respect to the standard symplectic basis of H1 (Cf , Z) (cf. [12, p. 3.76]),
and let Zf = Ω1−1 Ω2 be the corresponding normalized period matrix, so that
J (Cf ) = Cg /(1g |Zf ).
−1
Note that we have 2g g−2 Weierstrass D divisors which contain a given
Weierstrass point Wk = (αk , 0). For any of these divisors, the hyperplane HD
given by Proposition 3.1 passes through the point φ(W ) = (1 : αk : αk2 : · · · :
g−1
αk ). Now, if we take g − 1 different Weierstrass divisors containing W and
intersect the corresponding hyperplanes (which amounts to solving a linear
system) we recover the coordinates of φ(W ), and hence we find the root αk of
the polynomial f (x). We thus obtain the following theorem, which can be seen
as a completely explicit version of Torelli’s theorem for hyperelliptic curves:
Theorem 8.1. Let W1 = (α1 , 0), . . . , Wn = (αn , 0) be g different Weierstrass

points on Cf . Let Di = Σj =i Wj , and take wi = Π(Di ). If (x1 , . . . , xg ) is
a solution of the linear system of equations:

 ∂θ (w ) · · · ∂θ
(w ) 
∂z1 1 ∂zg 1
 .. .. 
     
 ∂θ . .  X1 0
 (wi−1 ) ··· ∂θ
(w ) 
 ∂z1 ∂zg i−1   X 2   0
  Ω −1  .  =  .  ,
 ∂θ  1  .   .. 
 ∂z (wi+1 ) ··· ∂θ
∂zg (wi+1 ) 
 .
 1
 .. ..  Xg 0
 . . 
∂θ
∂z1 (wg ) ··· ∂θ
∂zg (wg )
then αi = x2 /x1 .
In some situations, one has a period matrix of Cf with respect to a symplectic

basis which is not the standard one. As will be seen below, this has no effect for
genus 2 and 3. For higher genus, this may present a problem, since we do not
know a priori which odd 2-torsion points we must take to find the roots of f . If
we take a 2-torsion point which does not come from a Weierstrass divisor we will
detect it immediately, since we will obtain

∂θ ∂θ
(w1 ), . . . , (w1 ) = (0, . . . , 0).
∂z1 ∂zg
It is more difficult to know which combinations of Weierstrass torsion points we
must take. In this case, we can use the following approach.
Theorem 8.2. Let w ∈ J (Cf ) be a Weierstrass 2-torsion point. Let

 
X1

g
∂θ ∂θ .
Hw (X1 , . . . , Xg ) := (w) · · · (w) Ω1−1  ..  = hj Xj .
∂z1 ∂zg
Xg j =1
The polynomial fw (X) = h1 + h2 X + · · · + hg Xg−1 has g − 1 common roots with

the polynomial f .
Proof. Let W1 = (α1 , 0), . . . , Wg−1 = (αg−1 , 0) be the Weierstrass points on Cf

such that w = Π(W1 + · · · + Wg−1 ). The αj satisfy the equality
g−1
Hw 1, αj , αj2 , . . . , αj = 0,
so they are roots of fw . ✷
In the following two sections we give specific versions of Theorems 8.1 and 8.2
for genus 2 and 3, where they are especially simple.
9. The quintic and sextic equations
Let f (x) = a6 X6 + a5 X5 + a4 X4 + a3 X3 + a2 X2 + a1 X + a0 ∈ C[X] be

a separable polynomial of degree 5 or 6. Consider the associated genus 2
hyperelliptic curve Cf defined by Y 2 = f (X), which has 6 Weierstrass points
W1 = (α1 , 0), . . . , W6 = (α6 , 0) given by the roots αk of the polynomial f (we
take α6 = ∞ if deg f = 5). The Weierstrass divisors are precisely these six points,
Dk = Wk , k = 1, . . . , 6. The hyperplanes Hk := HDk intersect the canonical
curve φ(Cf ) exactly on the images φ(Wk ) of the Weierstrass points through the
canonical map. The jacobian has exactly 6 odd 2-torsion points. Hence, any odd
2-torsion point is a Weierstrass 2-torsion point. In this situation, Theorem 8.1
reads:
Theorem 9.1. Let

f (x) = a6 X6 + a5 X5 + a4 X4 + a3 X3 + a2 X2 + a1 X + a0 ∈ C[X]
be a separable polynomial of degree 5 or 6. The six roots of the polynomial f
are the ratios xk,2 /xk,1 , given by the solutions (xk,1 , xk,2 ) of the homogeneous
equations:

∂θ ∂θ −1 X1
Hk (X1 , X2 ) := (wk ) (wk ) Ω1 = 0, 1 k 6.
∂z1 ∂z2 X2
Here Ω = (Ω1 |Ω2 ) is the period matrix of the curve Cf with respect to the basis
ω1 = dx/y, ω2 = x dx/y of H 0 (Cf, Ω 1 ) and any symplectic basis of H1 (Cf , Z),
Zf = Ω1−1 Ω2 , and

1 0 1 0 1 0 1 1
w1 = Zf + , w2 = Zf + ,
2 1 2 1 2 1 2 1

1 1 1 1 1 1 1 1
w3 = Zf + , w4 = Zf + ,
2 0 2 0 2 0 2 1

1 1 1 0 1 1 1 1
w5 = Zf + , w6 = Zf + ,
2 1 2 1 2 1 2 0
are the six odd 2-torsion points of J (Cf ).
Note that we do not need to use any particular homology basis, since the only
effect of a symplectic change of basis will be a reordering of the values of the
ratios xk,1 /xk,2 .
10. The septic and octic equations
Let us consider now the case of a hyperelliptic curve Cf of genus 3, or

equivalently, the problem of solving a separable equation of degree 7 or 8, f (x) =
a8 X8 + · · · + a0 . We have now 8 Weierstrass points W1 = (α1 , 0), . . . , W8 =

(α8 , 0) (one of the αi may be ∞) which give rise to 28 Weierstrass divisors
Dij = Wi + Wj , 1 i < j 28. The images wij = Π(Dij ) of these divisors
through the Abel–Jacobi map are exactly all the 28 odd 2-torsion points in J (Cf ),
so any odd 2-torsion point is a Weierstrass 2-torsion point. The specific version of
Theorem 8.2 for this case is the following corollary.
Corollary 10.1. Let f (x) = a8 X8 + · · · + a0 be a separable polynomial of

degree 7 or 8. Let Ω = (Ω1 |Ω2 ) be the period matrix of the curve Cf with respect
to the basis
dx x dx x 2 dx
ω1 = , ω2 = , ω3 = of H 0 Cf, Ω 1
y y y
and any symplectic basis of H1 (Cf , Z), and let Zf = Ω1−1 Ω2 . Let w ∈ J (Cf )[2]
be any odd 2-torsion point, and denote by Hw (X1 , X2 , X3 ) = MX1 + NX2 +
P X3 = 0 the line of P2 given by

∂θ ∂θ ∂θ X1
−1
Hw (X1 , X2 , X3 ) := (w) (w) (w) Ω1 X2 = 0.
∂z1 ∂z2 ∂z3 X3

The x-coordinates of the Weierstrass points W = (α, 0), W = (β, 0) such that
w = Π(W + W ) are the roots of P x 2 + Nx + M = 0.
11. The discriminant
For a hyperelliptic curve of genus g

Cf : Y 2 = f (X) = a2g+2X2g+2 + · · · + a0
= a2g+2(X − α1 ) · · · (X − α2g+2 ),
there are two objects that deserve to be called discriminant of Cf . First we have
the algebraic discriminant of Cf , ∆alg (Cf ), which is the discriminant of the
polynomial f defining the curve:
4g+2

∆alg (Cf ) := ∆(f ) = a2g+2 (αj − αk )2 .
j <k
On the other side, we have the analytic discriminant, whose definition is slightly
more intricate (cf. [13]). In order to simplify, we are assuming that deg f = 2g + 2
(unlike in loc. cit.; we give the adapted versions of formulas appearing there). We
denote by Tk be the collection of subsets T ⊂ {1, . . . , 2g + 2} of cardinality k.
Definition 11.1 [13]. The discriminant modular form on Hg is

δ(Z) = θ [ηT ◦U ](Z)8 .
T ∈Tg+1
Here U = {1, 3, . . ., 2g + 1} and the analytic theta characteristic ηT ◦U has been

introduced in Section 4.
This product is never zero on hyperelliptic period matrices, since the

analytic theta characteristics ηT ◦U are always associated with geometric theta
characteristics given by non-effective semicanonical divisors.
It is well known that these two discriminants agree up to a multiplicative
constant for genus 1 curves. Grant [14] proved this relation for genus 2. In the
general case we have the following statement.
Proposition 11.1 (Lockhart [13]). Let Cf : Y 2 = f (X) be a genus g hyperelliptic

curve given by a polynomial f of degree 2g + 2 (not necessarily monic), and let
Zf be a normalized period matrix for it. Then
∆alg (Cf )2n = (2π)4rg (det Ω1 )−4r δ(Zf )2 ,

2g + 2 2g
where r = , n= .
g+1 g+1
(Note that the result that appears in [13] is for monic polynomials of degree
2g + 1.) The proof of this proposition is based on the classical formula of Thomae,
which we recall in the general case of a non-monic polynomial.
Theorem 11.1 (Thomae’s formula [12]). For any set S ⊂ {1, . . . , 2g + 2} of even
cardinality,
g
θ [ηS ](Z)4 = ± 4π 2 a2g+2 det Ω12 (αi − αj ) (αi − αj ).
i<j i<j
i,j ∈S◦U i,j ∈S◦U
/
We will now give another modular approach to the discriminant, using our
jacobian version of Thomae’s formula (Theorem 7.1).
So far, we have been using a slightly different definition of the Thetanullwerte
and the jacobian nullwerte: we have overpassed the exponential factor, since it
was completely unnecessary for the resolution of algebraic equations. But now
we want to combine our results with the classical ones, so we introduce some
notations to make all the definitions compatible.
Given a semicanonical divisor D, we will write θ [D] to denote the Thetanull-
werte θ [η](Zf ) given by the analytic theta characteristic η corresponding to D.
For a Weierstrass divisor D, we define H [D] to be the hyperplane given by the
equation:
 
X1
∂θ [η] ∂θ [η] .
H [D](X1 , . . . , Xg ) := (0, Zf ) · · · (0, Zf ) Ω1−1  ..  = 0.
∂z1 ∂zg
Xg
Finally, given g semicanonical divisors D1 , . . . , Dg , the symbol J [D1 , . . . , Dg ]

will stand for [η1 , . . . , ηg ](Zf ), where the ηi are the analytic theta characteristics
associated with the Di .
In Section 6 we have explained how to associate with every I = {i1 , . . . , ig }
∈ Tg a fundamental system DI = {Di1 , . . . , Dig , Dig+1 , . . . , Di2g+2 } of semi-
canonical divisors (where I ∪ {ig+1 , . . . , i2g+2} = {1, . . . , 2g + 2}). This fun-
damental system of semicanonical divisors gives in turn a fundamental system
wi1 , . . . , wi2g+2 of 2-torsion points in J (Cf ). We will denote by J [DI ] the jaco-
bian nullwerte J [Di1 , . . . , Dig ] of this fundamental system. Note that J [DI ] and
J (wi1 , . . . , wig ) differ only by an exponential factor.
Each subset J = {j1 , . . . , jg−1 } ∈ Tg−1 gives a Weierstrass divisor DJ =
j ∈J Wj . We define

H (J ) := H [DJ ](Wj ),
j ∈J
/
g−1
where H [DJ ](Wj ) = H [DJ ](1, αj , . . . , αj ).
2g 2g + 2
Theorem 11.2. Let s = g−2 , m= g = |Tg |. Then

∆alg (Cf )s = a2g+2
(4g+2)s
det Ω12m H (J )2 J [DI ]−2 .
J ∈Tg−1 I ∈ Tg
Proof. We first rewrite Theorem 7.1 with the notations just introduced. It looks
exactly as before, since the exponential factors that appear in H [Dj ] and
|J [D1 , . . . , Dg ]| are cancelled:
g

V (α1 , . . . , αg ) = det Ω H [Dj ](Wj ) J [D1 , . . . , Dg ].
1
j =1
Now, we take the product of these equalities over all fundamental systems of
the form DI , and raise it to the square. Every factor (αi − αj ) appearing in the
definition of ∆alg (Cf ) appears once for every DI such that {i, j } ⊂ I , and hence

it appears 2 g 2g
− 2 times in the final product. For every fundamental system DI
we also obtain the product
g
H [DJ ] 1, αj , . . . , αj ,
J ∈Tg−1 ,
J ∪{j }=I
and thus, in the final product every form H [DJ ](X1 , . . . , Xg ) appears evaluated
at every αj with j ∈/ J. ✷
We can now combine Lockhart’s result with our theorem, to find relations
between jacobian nullwerte and products of theta nullwerte. We will make this for
low genera, where a geometrical interpretation of these relations becomes clear.
12. Geometric Jacobi identity and discriminant for genus 2
For curves of genus 2, there is a classical relation between jacobian null-

werte and products of theta nullwerte: Rosenhain’s identity (cf. [1,4,7]). Let us
denote by η1 , . . . , η6 the six odd analytic theta characteristics of dimension 2.
Rosenhain’s identity states that for every pair ηj , ηk :

[ηj , ηk ](Zf ) = ±π 2 θ [ηj + ηk − ηr ](Zf ),
r=j,k
where the sum ηj + ηk − ηr is taken modulus 2. This formula has a geometrical

counterpart: let Cf a hyperelliptic curve of genus 2, with Weierstrass points
W1 = (α1 , 0), . . . , W6 = (α6 , 0). For every pair Wj , Wk ,

J [Wj , Wk ] = ±π 2 θ [Wj + Wk − Wr ]. (4)
r=i,j
This relation, combined with Theorem 7.1, gives the following formula, which
can be seen as a geometric Jacobi identity for genus 2.
Proposition 12.1.
H [Wj ](1, αk )H [Wk ](1, αj )

= ±π 2 det Ω1−1 (αk − αj ) θ [Wj + Wk − Wr ].
r=i,j
For genus 2 we also have a specific formula for the discriminant [14]:

∆alg (Cf ) = ±(2π)20 det Ω1−10 θ [η](Zf )2 (5)
η∈TC+
2

= ±(2π)20 det Ω1−10 θ [Wi + Wj − Wk ]. (6)
i<j <k
If we take the product of |J [Wj , Wk ]| over all pairs j < k and apply Rosenhain’s
identity in each factor we obtain:

J [Wj , Wk ]2 = π 60 θ [Wi + Wj − Wk ]6
j <k i<j <k
−120 −60
= 2 π det Ω160 ∆alg (Cf )6 .
Finally, Theorem 11.2 for genus 2 states that

30 j <k
H [Wj ](Wk )2
∆alg (Cf ) = a6 det Ω1
10
.
j <k |J [Wj , Wk ]|
2
Combining the last two equalities we arrive at the following proposition.

Proposition 12.2.

∆alg (Cf )7 = 2120a610 π 60 det Ω1−30 H [Wj ](Wk )2 .
j <k
13. Geometric Jacobi identity and discriminant for genus 3
In order to obtain an analogue of Proposition 12.1 for genus 3, we need

a version of Rosenhain’s identity for genus 3. We find this identity in [4]. We
provide a geometric version of Frobenius’ formula.
Proposition 13.1. Let Cf be a hyperelliptic curve of genus 3, with Weierstrass

points W1 , . . . , W8 . The following equality holds for every triplet Wi , Wj , Wk :

J [Wi + Wk , Wi + Wj , Wj + Wk ] = π 3 θ [Wi + Wj + Wk − Wr ].
r=i,j,k
Proof. Applying formula (4.2), we see that the sum
Π(Wi + Wk ) + Π(Wi + Wj ) + Π(Wj + Wk )

= Π(D{j,k}◦{1,3,5,7} ) + Π(Wj + Wk ) = Π(D{1,3,5,7} ) = κ
is always even, and hence, by [7, Corollary to Theorem 2], the jacobian
Thetanullwerte |J [Wi + Wk , Wi + Wj , Wj + Wk ]| is a product of five even
Thetanullwerte θ [D 1 ] · · · θ [Dz
5 ]. By Theorem 5.1, the divisors Dr are the only
ones which form a fundamental system with Wi + Wk , Wi + Wj , Wj + Wk and
r ] = 0. We have seen in Proposition 6.1 that the divisors D
θ [D r = Wi + Wj +
Wk −Wr , r = i, j, k form a fundamental system with Wi +Wk , Wi +Wj , Wj +Wk
and by Riemann’s singularity theorem, θ [D r ] = 0, because l(Wi + Wj + Wk −
Wr ) = 0 by Lemma 4.1. ✷
With this in hand, we can now proceed exactly as in the previous section. We
obtain the following results.
Proposition 13.2 (Geometric Jacobi identity for genus 3). For every triplet of
Weierstrass points Wi , Wj , Wk ,
H [Wi + Wj ](Wk )H [Wi + Wk ](Wj )H [Wj + Wk ](Wi )

= ±π 3 det Ω1−1 V (αi , αj , αk ) θ [Wi + Wj + Wk − Wr ].
r=i,j,k
Proposition 13.3.
(a) ∆alg (Cf )30 = 2840 π 504 a8420 det Ω1−280

× J [Wi + Wk , Wi + Wj , Wj + Wk ].
i<j <k

(b) ∆alg (Cf )36 = a884 π 504 2840 det Ω1−168 H [Wj + Wk ](Wr ).
j <k r=j,k
14. Some conjectural formulas
At this point, it is natural to ask whether it is possible to generalize the

formulas in the previous sections for arbitrary genus g. The necessary tool is a
generalization of Rosenhain and Frobenius formulae (that is, a generalization of
Jacobi’s triple product identity). Igusa’s Theorem 5.1 shows that in genus g 6
this generalization is not possible for a general matrix Z ∈ Hg . But Frobenius [4,
p. 254] showed that for any hyperelliptic matrix Zf and any fundamental system
η1 , . . . , η2g+2 , there is a constant ε ∈ C such that

2g+2
[η1 , . . . , ηg ](Zf ) = επ g θ [ηi ](0; Zf ).
i=g+1
Taking into account Igusa’s theorem, it seems reasonable to conjecture that this
constant ε is equal to ±1. This would imply the following (conjectural) formulas.
Conjecture 14.1. Let Cf be a hyperelliptic curve of genus g g, with Weierstrass

points W1 = (α1 , 0), . . . , W2g+2 = (α2g+2 , 0). Let D = i=1 Wi , Di = D − Wi ,
for i = 1, . . . , g, and Di = D + Wi − 2W2g+2 , for i = g + 1, . . . , 2g + 1. Then

2g
(a) J [D1 , . . . , Dg ] = ±π g θ [Dk ].
k=g+1

g

(b) H [Di ](Wi ) = ±π g det Ω1−1 V (α1 , . . . , αg ) θ [D − Wr ].
i=1 r∈
/ {1,...,g}
+2
(c) Let m = 2gg + 1 . Then

J [DI ]8 = π mg δ(Zf )g+1 .
I ∈ Tg
2g + 2 2g
(d) Let r = g+1 ,n=2 g+1 . Then

∆alg (Cf )n(g+1) = (2π)4g(g+1)r π −2mg (det Ω1 )−4r(g+1) J [DI ]16 .
I ∈ Tg
We observe that formulas (a) and (b) are equivalent by Theorem 7.1 and they
imply formulas (c) and (d).
References
[1] G. Rosenhain, Mémoire sur les fonctions de deux variables et à quatre périodes qui sont les
inverses des intégrales ultra-elliptiques de la première classe, Mémoires des savants étrangers XI
(1851) 362–468.
[2] J. Thomae, Beitrag zur Bestimmung von θ (0, 0, . . . , 0) durch die Klassenmoduln algebraischer
Funktionen, J. Reine Angew. Math. 71 (1870) 201–222.
[3] B. Riemann, Gesammelten Mathematische Werke. Nachträge by M. Noether and W. Wirtinger,
Teubner, Leipzig, 1902, Dover, New York, 1953.
[4] F.G. Frobenius, Über die constanten Factoren der Thetareihen, J. Reine Angew. Math. 98 (1885)
241–260.
[5] D. Mumford, Tata Lectures on Theta, I, Progr. Math., Vol. 28, Birkhäuser, Boston, 1983.
[6] J.I. Igusa, On Jacobi’s derivative formula and its generalizations, Amer. J. Math. 102 (2) (1980)
409–446.
[7] J.I. Igusa, On the nullwerte of jacobians of odd theta functions, Symp. Math. 24 (1979) 83–95.
[8] H. Umemura, Solution of algebraic equations by theta constants, appendix to [12].
[9] K. Takase, A generalization of Rosenhain’s normal form for hyperelliptic curves with an
application, Proc. Japan Acad. Ser. A Math. Sci. 72 (7) (1996) 162–165.
[10] E. Arbarello, M. Cornalba, P.A. Griffiths, J. Harris, Geometry of Algebraic Curves, Grundlehren
Math. Wiss., Vol. 267, Springer-Verlag, New York, 1985.
[11] C. Birkenhake, H. Lange, Complex Abelian Varieties, Grundlehren Math. Wiss., Vol. 302,
Springer-Verlag, New York, 1992.
[12] D. Mumford, Tata Lectures on Theta, II, Progr. Math., Vol. 43, Birkhäuser, Boston, 1984.
[13] P. Lockhart, On the discriminant of a hyperelliptic curve, Trans. Amer. Math. Soc. 342 (2) (1994)
729–752.
[14] D. Grant, A generalization of Jacobi’s derivative formula, J. Reine Angew. Math. 392 (1988)
125–136.

Jacobian Nullwerte and Algebraic Equations: Jordi Guàrdia

Uploaded by

Copyright:

Available Formats

Jacobian Nullwerte and Algebraic Equations: Jordi Guàrdia

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jacobian Nullwerte and Algebraic Equations: Jordi Guàrdia

Uploaded by

Copyright:

Available Formats

Journal of Algebra 253 (2002) 112–132

Jacobian nullwerte and algebraic equations ✩

Throughout the paper, C will be a smooth curve of genus g > 1 defined

For α1 , . . . , αg ∈ C, let V (α1 , . . . , αg ) denote the Vandermonde matrix:

2.2. The Abel–Jacobi map

Let C be a complex smooth curve of genus g > 1. We pick a symplectic basis

ud (D) = (η1 , . . . , ηg ) − (η1 , . . . , ηg ).

2.3. Riemann theorems

For z ∈ Cg , we define the Riemann theta function of J (C) by

Theorem 2.2 (Riemann singularity theorem). For every effective divisor D of

3. The canonical map

The choice of the basis ω1 , . . . , ωg of the holomorphic differential forms on C

 such that the divisor D = P1 + · · · +

Proof. The hyperplane HD ⊂ Pg−1 determined by the points P1 , . . . , Pg−1 is

4. Theta characteristics on hyperelliptic curves. Two-torsion points

A semicanonical divisor on a curve of genus g is a divisor D of degree

Lemma 4.1 [10]. Let Cf be a hyperelliptic curve of genus g defined over an

(a) Any semicanonical divisor on Cf is linearly equivalent to one of the form DI

We introduce the following definition.

Definition 4.1. A Weierstrass divisor on Cf is a semicanonical divisor satisfying

The point zm is said to be even (respectively odd) if m .m ≡ 0 (mod 2)

Definition 4.2. A Weierstrass 2-torsion point is an element of J (Cf )[2] which is

The bijection between geometric theta characteristics and 2-torsion points of

where the non-zero

and the Riemann vector is

Finally, we have the following relation:

Π(DI ) + Π(DJ ) = Π(DI ◦J ◦U ), (2)

where U = {1, 3, . . . , 2g + 1} and S ◦ T denotes the symmetric difference of two

5. Jacobian nullwerte. Fundamental systems

Definition 5.1 [7]. A jacobian nullwerte is a determinant of the form:

Jacobian nullwerte can be thought of as (modular) functions over the Siegel

where S is the set of all (g + 2)-tuples {ηg+1 , . . . , η2g+2 } of even theta

We are especially interested in the values of jacobian nullwerte at hyperelliptic

Definition 5.2. Let D1 , . . . , Dg be g Weierstrass divisors on a hyperelliptic

6. Fundamental systems for hyperelliptic curves

Let Cf be a hyperelliptic curve of genus g, with Weierstrass points W1 , . . . ,

Proposition 6.1. The set w1 , . . . , w2g+2 is a fundamental system of 2-torsion

7. Jacobian nullwerte on hyperelliptic curves

Theorem 7.1. Let W1 = (α1 , 0), . . . , Wg = (αg , 0) ∈ Cf (C) be g different

Proof. The divisors Dk satisfy l(Dk ) = 1, so that the corresponding hyper-

which is a matrix with entries in the field Q(α1 , . . . , αg ). If we replace Wi by any

Taking determinants in formula (3), we obtain the following corollary, which

8. Solving algebraic equations with jacobian nullwerte

Let f (X) = an (X − α1 ) · · · (X − αn ) ∈ C[X] be a polynomial of arbitrary

As in the previous section, we fix

as basis of H 0 (Cf , Ω 1 ). Let Ω = (Ω1 |Ω2 ) the period matrix of ω1 , . . . , ωg

Theorem 8.1. Let W1 = (α1 , 0), . . . , Wn = (αn , 0) be g different Weierstrass

a solution of the linear system of equations:

In some situations, one has a period matrix of Cf with respect to a symplectic

Theorem 8.2. Let w ∈ J (Cf ) be a Weierstrass 2-torsion point. Let

The polynomial fw (X) = h1 + h2 X + · · · + hg Xg−1 has g − 1 common roots with

Proof. Let W1 = (α1 , 0), . . . , Wg−1 = (αg−1 , 0) be the Weierstrass points on Cf

so they are roots of fw . ✷

9. The quintic and sextic equations

Let f (x) = a6 X6 + a5 X5 + a4 X4 + a3 X3 + a2 X2 + a1 X + a0 ∈ C[X] be

Theorem 9.1. Let

10. The septic and octic equations

Let us consider now the case of a hyperelliptic curve Cf of genus 3, or

such that the divisor D = P1 + · · · +

Theorem 7.1. Let W1 = (α1 , 0), . . . , Wg = (αg , 0) ∈ Cf (C) be g different

Conjecture 14.1. Let Cf be a hyperelliptic curve of genus g g, with Weierstrass