Journal of Geometry and Physics 58 (2008) 10631079

Contents lists available at ScienceDirect

Journal of Geometry and Physics
journal homepage: www.elsevier.com/locate/jgp
Review
Prym varieties and applications
A. Lesfari

Department of Mathematics, Faculty of Sciences, University of Chouab Doukkali, B.P. 20, El-Jadida, Morocco
a r t i c l e i n f o
Article history:
Received 7 March 2007
Received in revised form 5 April 2008
Accepted 7 April 2008
Available online 15 April 2008
JGP SC:
Algebraic geometry
Classical integrable systems
MSC:
14Q05
14H40
70H06
Keywords:
Curves
Prym varieties
Integrable systems
a b s t r a c t
The classical definition of Prym varieties deals with the unramified covers of curves. The
aim of this article is to give explicit algebraic descriptions of the Prym varieties associated
with ramified double covers of algebraic curves. We make a careful study of the connection
with the concept of algebraic completely integrable systems and we apply the methods to
some problems such as the HnonHeiles system, the Kowalewski rigid body motion and
Kirchhoffs equations of motion of a solid in an ideal fluid.
2008 Elsevier B.V. All rights reserved.
Contents
1. Introduction .............................................................................................................................................................................................. 1063
2. Prym varieties ........................................................................................................................................................................................... 1064
3. Algebraic complete integrability ............................................................................................................................................................. 1071
3.1. A survey on abelian varieties and algebraic integrability ......................................................................................................... 1071
3.2. The HnonHeiles system........................................................................................................................................................... 1073
3.3. The Kowalewski rigid body motion............................................................................................................................................ 1076
3.4. Kirchhoffs equations of motion of a solid in an ideal fluid ...................................................................................................... 1077
References ................................................................................................................................................................................................. 1079
1. Introduction
During the last decades, algebraic geometry has become a tool for solving differential equations and spectral questions of
mechanics and mathematical physics. This paper consists of two separate but related topics: the first part purely algebraic-
geometric, the second one on Prym varieties in algebraic integrability.
Prym variety Prym(C/C
0
) is a subabelian variety of the jacobian variety Jac(C) = Pic
0
(C) = H
1
(O
C
)/H
1
(C, Z)
constructed from a double cover C of a curve C
0
: if is the involution on C interchanging sheets, then extends by

Corresponding address: B.P. 271, El-Jadida, Morocco.

E-mail addresses: Lesfariahmed@yahoo.fr, lesfari@ucd.ac.ma.
0393-0440/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomphys.2008.04.005
1064 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
linearity to a map : Jac(C) Jac(C). Up to some points of order two, Jac(C) splits into an even part and an odd
part: the even part is Jac(C
0
) and the odd part is a Prym(C/C
0
). The classical definition of Prym varieties deals with the
unramified double covering of curves and was introduced by W. Schottky and H. W. E. Jung in relation with the Schottky
problem [6] of characterizing jacobian varieties among all principally polarized abelian varieties (an abelian variety is a
complex torus that can be embedded into projective space). The theory of Prym varieties was dormant for a long time,
until revived by D. Mumford around 1970. It now plays a substantial role in some contemporary theories, for example
integrable systems [1,2,5,8,10,14,15,17], the KadomtsevPetviashvili equation (KP equation), in the deformation theory of
two-dimensional Schrdinger operators [19], in relation to CalabiYau three-folds and string theory, in the study of the
generalized theta divisors on the moduli spaces of stable vector bundles over an algebraic curve [7,11],. . .
Integrable hamiltonian systems are nonlinear ordinary differential equations described by a hamiltonian function and
possessing sufficiently many independent constants of motion in involution. By the ArnoldLiouville theorem [4,16], the
regular compact level manifolds defined by the intersection of the constants of motion are diffeomorphic to a real torus
on which the motion is quasi-periodic as a consequence of the following differential geometric fact; a compact and
connected n-dimensional manifold on which there exist n vector fields which commute and are independent at every point
is diffeomorphic to an n-dimensional real torus and each vector field will define a linear flow there. A dynamical system is
algebraic completely integrable (in the sense of Adlervan Moerbeke [1]) if it can be linearized on a complex algebraic torus
C
n
/lattice (=abelian variety). The invariants (often called first integrals or constants) of the motion are polynomials and the
phase space coordinates (or some algebraic functions of these) restricted to a complex invariant variety defined by putting
these invariants equals to generic constants, are meromorphic functions on an abelian variety. Moreover, in the coordinates
of this abelian variety, the flows (run with complex time) generated by the constants of the motion are straight lines.
One of the remarkable developments in recent mathematics is the interplay between algebraic completely integrable
systems and Prym varieties. The period of these Prym varieties provide the exact periods of the motion in terms of explicit
abelian integrals. The aim of the first part of the present paper is to give explicit algebraic descriptions of the Prym varieties
associated to ramified double covers of algebraic curves. The basic algebraic tools are known and can be found in the book by
Arbarello, Cornalba, Griffiths, Harris [3] and in Mumfords paper [18]. In the second part of the paper, we make a careful study
of the connection with the concept of algebraic completely integrable systems and we apply the methods to some problems
such as the HnonHeiles system, the Kowalewski rigid body motion and Kirchhoffs equations of motion of a solid in an
ideal fluid. The motivation was the excellent Haines paper [10] on the integration of the EulerArnold equations associated
to a class of geodesic flow on SO(4). The Kowalewskis top and the Clebschs case of Kirchhoffs equations describing the
motion of a solid body in a perfect fluid were integrated in terms of genus two hyperelliptic functions by Kowalewski [13]
and Ktter [12] as a result of complicated and mysterious computations. The concept of algebraic complete integrability
(AdlervanMoerbeke) throws a completely newlight onthese two systems. Namely, inbothcases (see [14] for Kowalewskis
top and [10] for geodesic flowon SO(4)), the affine varieties in C
6
obtained by intersecting the four polynomial invariants of
the floware affine parts of Prymvarieties of genus 3 curves which are double cover of elliptic curves. Such Prymvarieties are
not principally polarized and so they are not isomorphic but only isogenous to Jacobi varieties of genus two hyperelliptic
curves. This was a total surprise as it was generically believed that only jacobians would appear as invariants manifolds
of such systems. The method that is used for reveiling the Pryms is due to Haine [10]. By now many algebraic completely
integrable systems are known to linearize on Prym varieties.
2. Prym varieties
Let : C C
0
be a double covering with 2n branch points where C and C
0
are nonsingular complete curves. Let
: C C be the involution exchanging sheets of C over C
0
= C/. By Hurwitzs formula, the genus of C is g = 2g
0
+n1,
where g
0
is the genus of C
0
. Let O
C
be the sheaf of holomorphic functions on C. Let S
g
(C) be the g-th symmetric power of C
(the totality of unordered sets of points of C). On S
g
(C), two divisors D
1
and D
2
are linearly equivalent (in short, D
1
D
2
)
if their difference D
1
D
2
is the divisor of a meromorphic function or equivalently if and only if
_
D
2
D
1
=
_

,
1
C
,
where
1
C
is the sheaf of holomorphic 1-forms on C and is a closed path on C. We define the jacobian (Jacobi variety) of C,
to be Jac(C) = S
g
(C)/ . To be precise, the jacobian Jac(C) = Pic
0
(C) of C is the connected component of its Picard group
parametrizing degree 0 invertible sheaves. It is a compact commutative algebraic group, i.e., a complex torus. Indeed, from
the fundamental exponential sequence, we get an isomorphism
Jac(C) = Pic
0
(C) = H
1
(C, O
C
)/H
1
(C, Z) H
0
(C,
1
C
)

/H
1
(C, Z) C
g
/Z
2g
,
via the Serres duality. Consider the following mapping
S
g
(C) C
g
/L

,
g

k=1

k

g

k=1
_

k
(t)
_

1
, . . . ,
g
_
= t
_
k
1
, . . . , k
g
_
,
where
_

1
, . . . ,
g
_
is a basis of
1
C
, L

is the lattice associated to the period matrix and

1
, . . . ,
g
some
appropriate variables defined on a non empty Zariski open set. Let (a
1
, . . . , a
g
0
, . . . , a
g
, b
1
, . . . , b
g
0
, . . . , b
g
), be a canonical
homology basis of H
1
(C, Z) such that (a
1
) = a
g
0
+n
, . . . ,
_
a
g
0
_
= a
g
,
_
a
g
0
+1
_
= a
g
0
+1
, . . . ,
_
a
g
0
+n1
_
=
a
g
0
+n1
, (b
1
) = b
g
0
+n
, . . . ,
_
b
g
0
_
= b
g
,
_
b
g
0
+1
_
= b
g
0
+1
, . . . ,
_
b
g
0
+n1
_
= b
g
0
+n1
, for the involution . Notice
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1065
that
_
a
g
0
+1
_
, . . . ,
_
a
g
0
+n1
_
,
_
b
g
0
+1
_
, . . . ,
_
b
g
0
+n1
_
are homologous to zero on C
0
. Let
_

1
, . . . ,
g
_
be a basis of
holomorphic differentials on C where
g
0
+n
, . . . ,
g
, are holomorphic differentials on C
0
and

(
j
) =
_

j
, 1 j g
0
+n 1,

j
, g
0
+n j g,
the pullback of
j
. The period matrix of Jac(C) is explicitly given by
=
_
A B C D E F
G H I J K L
_
,
where A, . . . , L denote the following matrices:
A =
_
_
_
_
_
_
_
_
_
a
1

1
. . .
_
a
g
0

1
.
.
.
.
.
.
_
a
1

g
0
+n1
. . .
_
a
g
0

g
0
+n1
_
_
_
_
_
_
_
_
, (1)
B =
_
_
_
_
_
_
_
_
_
a
g
0
+1

1
. . .
_
a
g
0
+n1

1
.
.
.
.
.
.
_
a
g
0
+1

g
0
+n1
. . .
_
a
g
0
+n1

g
0
+n1
_
_
_
_
_
_
_
_
,
C =
_
_
_
_
_
_
_
_
_
a
g
0
+n

1
. . .
_
a
g

1
.
.
.
.
.
.
_
a
g
0
+n

g
0
+n1
. . .
_
a
g

g
0
+n1
_
_
_
_
_
_
_
_
,
D =
_
_
_
_
_
_
_
_
_
b
1

1
. . .
_
b
g
0

1
.
.
.
.
.
.
_
b
1

g
0
+n1
. . .
_
b
g
0

g
0+n1
_
_
_
_
_
_
_
_
,
E =
_
_
_
_
_
_
_
_
_
b
g
0
+1

1
. . .
_
b
g
0
+n1

1
.
.
.
.
.
.
_
b
g
0
+1

g
0
+n1
. . .
_
b
g
0
+n1

g
0+n1
_
_
_
_
_
_
_
_
,
F =
_
_
_
_
_
_
_
_
_
b
g
0
+n

1
. . .
_
b
g

1
.
.
.
.
.
.
_
b
g
0
+n

g
0
+n1
. . .
_
b
g

g
0
+n1
_
_
_
_
_
_
_
_
,
G =
_
_
_
_
_
_
_
_
_
a
1

g
0
+n
. . .
_
a
g
0

g
0
+n
.
.
.
.
.
.
_
a
1

g
. . .
_
a
g
0

g
_
_
_
_
_
_
_
_
,
H =
_
_
_
_
_
_
_
_
_
a
g
0
+1

g
0
+n
. . .
_
a
g
0
+n1

g
0
+n
.
.
.
.
.
.
_
a
g
0
+1

g
. . .
_
a
g
0
+n1

g
_
_
_
_
_
_
_
_
,
1066 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
I =
_
_
_
_
_
_
_
_
_
a
g
0
+n

g
0
+n
. . .
_
a
g

g
0
+n
.
.
.
.
.
.
_
a
g
0
+n

g
. . .
_
a
g

g
_
_
_
_
_
_
_
_
,
J =
_
_
_
_
_
_
_
_
_
b
1

g
0
+n
. . .
_
b
g
0

g
0
+n
.
.
.
.
.
.
_
b
1

g
. . .
_
b
g
0

g
_
_
_
_
_
_
_
_
,
K =
_
_
_
_
_
_
_
_
_
b
g
0
+1

g
0
+n
. . .
_
b
g
0
+n1

g
0
+n
.
.
.
.
.
.
_
b
g
0
+1

g
. . .
_
b
g
0
+n1

g
_
_
_
_
_
_
_
_
,
and
L =
_
_
_
_
_
_
_
_
_
b
g
0
+n

g
0
+n
. . .
_
b
g

g
0
+n
.
.
.
.
.
.
_
b
g
0
+n

g
. . .
_
b
g

g
_
_
_
_
_
_
_
_
.
Notice that
_
a
g
0
+1

j
=
_
(a
g
0
+1
)

j
,
=
_
a
g
0
+1

(
j
),
=
_

_
_
a
g
0
+1

j
, 1 j g
0
+n 1,

_
a
g
0
+1

j
, g
0
+n j g,
.
.
.
_
a
g
0
+n1

j
=
_

_
_
a
g
0
+n1

j
, 1 j g
0
+n 1,

_
a
g
0
+n1

j
, g
0
+n j g,
_
b
g
0
+1

j
=
_
(b
g
0
+1)

j
,
=
_
b
g
0
+1

j
_
,
=
_

_
_
b
g
0
+1

j
, 1 j g
0
+n 1,

_
b
g
0
+1

j
, g
0
+n j g,
.
.
.
_
b
g
0
+1

j
=
_

_
_
b
g
0
+1

j
, 1 j g
0
+n 1,

_
b
g
0
+1

j
, g
0
+n j g,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1067
_
a
g
0
+n

j
=
_
(a
1
)

j
,
=
_
a
1

j
_
,
=
_

_
a
1

j
, 1 j g
0
+n 1,
_
a
1

j
, g
0
+n j g,
.
.
.
_
a
g
0
+n

j
=
_

_
a
g
0

j
, 1 j g
0
+n 1,
_
a
g
0

j
, g
0
+n j g,
and
_
b
g
0
+n

j
=
_
(b
1
)

j
,
=
_
b
1

j
_
,
=
_

_
b
1

j
, 1 j g
0
+n 1,
_
b
1

j
, g
0
+n j g,
.
.
.
_
b
g

j
=
_

_
b
g
0

j
, 1 j g
0
+n 1,
_
b
g
0

j
, g
0
+n j g.
Then, C = A, F = D, H = O, I = G, K = O, L = J, where O is the null matrix and therefore
=
_
A B A D E D
G O G J O J
_
,

_
C
1
C
2
C
3
C
4
C
5
C
6
_
.
By elementary column operation, we obtain the following matrices:

1
=
_
C
1
C
2
C
1
+C
3
C
4
C
5
C
4
+C
6
_
,
=
_
A B O D E O
G O 2G J O 2J
_
,

2
=
_
C
1
C
2
C
1
C
3
C
4
C
5
C
4
C
6
_
,
=
_
A B 2A D E 2D
G O O J O O
_
,

3
=
_
C
1
C
3
C
2
C
4
C
6
C
5
C
1
+C
3
C
4
+C
6
_
,
=
_
2A B 2D E O O
O O O O 2G 2J
_
,
=
_
O
O 2
_
,
where
=
_
G J
_
,
=
_
_
_
_
_
_
_
_
_
a
1

g
0
+n
. . .
_
a
g
0

g
0
+n
_
b
1

g
0
+n
. . .
_
b
g
0

g
0
+n
.
.
.
.
.
.
.
.
.
.
.
.
_
a
1

g
. . .
_
a
g
0

g
_
b
1

g
. . .
_
b
g
0

g
_
_
_
_
_
_
_
_
, (2)
1068 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
and
=
_
2A B 2D E
_
,
=
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
2
_
a
1

1
. . . 2
_
a
1

g
0
+n1
.
.
.
.
.
.
2
_
a
g
0

1
. . . 2
_
a
g
0

g
0
+n1
_
a
g
0
+1

1
. . .
_
a
g
0
+1

g
0
+n1
.
.
.
.
.
.
_
a
g
0
+n1

1
. . .
_
a
g
0
+n1

g
0
+n1
2
_
b
1

1
. . . 2
_
b
1

g
0
+n1
.
.
.
.
.
.
2
_
b
g
0

1
. . . 2
_
b
g
0

g
0
+n1
_
b
g
0
+1

1
. . .
_
b
g
0
+1

g
0
+n1
.
.
.
.
.
.
_
b
g
0
+n1

1
. . .
_
b
g
0
+n1

g
0
+n1
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_

. (3)
Let
L

=
_

_
g

i=1
m
i
_
a
i
_
_
_
_

1
.
.
.

g
_
_
_
_
+n
i
_
b
i
_
_
_
_

1
.
.
.

g
_
_
_
_
: m
i
, n
i
Z
_

_
,
be the period lattice associated to . Let us denote also by L

1
, L

2
, L

3
and L

the period lattices associated respectively to

1
,
2
,
3
and . Since L

= L

1
= L

2
= L

3
, it follows that the maps
C
g
0
+n1
/L

:
_
_
_
_
t
1
.
.
.
t
g
0
+n1
_
_
_
_
mod L

C
g
/L

:
_
_
_
_
_
_
_
_
_
_
_
t
1
.
.
.
t
g
0
+n1
0
.
.
.
0
_
_
_
_
_
_
_
_
_
_
_
mod L

,
C
g
0
/2L

:
_
_
_
_
t
1
.
.
.
t
g
0
_
_
_
_
mod 2L

C
g
/L

:
_
_
_
_
_
_
_
_
_
_
_
0
.
.
.
0
t
1
.
.
.
t
g
0
_
_
_
_
_
_
_
_
_
_
_
mod L

,
are injectives. Therefore, the tori C
g
0
+n1
/L

and C
g
0
/2L

can be embedded into C

g
/L

and the map

C
g
/L

3
= C
g
0
+n1
/L

C
g
0
/2L

C
g
/L

_
_
_
_
t
1
.
.
.
t
g
_
_
_
_
mod L

3

_
_
_
_
t
1
.
.
.
t
g
_
_
_
_
mod L

,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1069
shows that the jacobian variety Jac (C
0
) intersects the torus C
g
0
+n1
/L

in 2
2g
0
points. We have then the following diagram
C Ker N

C
u
Jac(C)
u

Jac

(C)
N

C
0
u
0
Jac(C
0
)
u

0
Jac

(C
0
)

Jac(C) H
1
(C, O

C
) H
1
(C
0
, (

O
C
)

)
N

Jac(C)
N

Jac(C
0
) H
1
(C
0
, O

C
0
),
where u : z divisor class(z p), p C, fixed, u
0
: z
0
divisor class(z
0
p
0
), p
0
C
0
, fixed with p
0
= (p),
N

: Jac(C) Jac(C
0
),

m
i
p
i


m
i
(p
i
) is the norm mapping and Jac

(C) = dual of Jac(C). The norm map N

is
surjective. Moreover, the dual Jac

(C) of Jac(C) is isomorphic to Jac(C). The Prym variety denoted Prym(C/C

0
) is defined
by
Prym(C/C
0
) =
_
H
0
_
C,
1
C
_

/H
1
(C, Z)

,
where denote the 1 eigenspace for a vector space on which the involution acts. Let D Prym(C/C
0
), i.e.,
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
0
.
.
.
0
2
_
D
0

g
0
+n
.
.
.
2
_
D
0

g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
L

_
_
_
_
_
_
_
_
D
0

g
0
+n
.
.
.
_
D
0

g
_
_
_
_
_
_
_
L

_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
D
0

1
.
.
.
_
D
0

g
0
+n1
_
D
0

g
0
+n
.
.
.
_
D
0

g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
t
1
.
.
.
t
g
0
+n1
0
.
.
.
0
_
_
_
_
_
_
_
_
_
_
_
.
Consequently, the AbelJacobi map
Jac (C) C
g
/L

, D
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
D
0

1
.
.
.
_
D
0

g
0
+n1
_
D
0

g
0
+n
.
.
.
_
D
0

g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
,
maps Prym(C/C
0
) bi-analtically onto C
g
0
+n1
/L

. By Chows theorem[9], Prym(C/C

0
) is thus a subabelian variety of Jac (C).
More precisely, we have
Prym(C/C
0
) = (Ker N

)
0
,
= Ker(1
Jac(C)
+)
0
,
= Im(1
Jac(C)
) Jac(C).
1070 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
Equivalently, the involution induces an involution
: Jac(C) Jac(C), class of D class of D,
and up to some points of order two, Jac(C) splits into an even part Jac (C
0
) and an odd part Prym(C/C
0
):
Jac(C) = Prym(C/C
0
) Jac (C
0
) ,
with
dimJac(C
0
) = g
0
,
dimJac(C) = g = 2g
0
+n 1,
dimPrym(C/C
0
) = g g
0
= g
0
+n 1.
Observe that (2) (resp. (3)) is the period matrix of Jac (C
0
) (resp. Prym(C/C
0
)). Write = (U, V), with
U =
_
_
_
_
_
_
_
_
2
_
a
1

1
. . . 2
_
a
g
0

1
_
a
g
0
+1

1
. . .
_
a
g
0
+n1

1
.
.
.
.
.
.
.
.
.
.
.
.
2
_
a
1

g
0
+n1
. . . 2
_
a
g
0

g
0
+n1
_
a
g
0
+1

g
0
+n1
. . .
_
a
g
0
+n1

g
0
+n1
_
_
_
_
_
_
_
_
,
V =
_
_
_
_
_
_
_
_
2
_
b
1

1
. . . 2
_
b
g
0

1
_
b
g
0
+1

1
. . .
_
b
g
0
+n1

1
.
.
.
.
.
.
.
.
.
.
.
.
2
_
b
1

g
0
+n1
. . . 2
_
b
g
0

g
0
+n1
_
b
g
0
+1

g
0
+n1
. . .
_
b
g
0
+n1

g
0
+n1
_
_
_
_
_
_
_
_
,
and let us call
e
1
=
_
_
_
_
_
_
_
2
_
a
1

1
.
.
.
2
_
a
1

g
0
+n1
_
_
_
_
_
_
_
, . . . , e
g
0
=
_
_
_
_
_
_
_
_
2
_
a
g
0

1
.
.
.
2
_
a
g
0

g
0
+n1
_
_
_
_
_
_
_
_
,
e
g
0
+1
=
_
_
_
_
_
_
_
_
_
a
g
0
+1

1
.
.
.
_
a
g
0
+1
_
_
_
_
_
_
_
_
, . . . , e
g
0
+n1
=
_
_
_
_
_
_
_
_
_
a
g
0
+n1

1
.
.
.
_
a
g
0
+n1

g
0
+n1
_
_
_
_
_
_
_
_
.
Then, in the new basis
_

1
, . . . ,
g
0
+n1
_
where

j
=
e
j

j
,
j
=
_
1 pour 1 j g
0
,
2 pour g
0
+1 j g
0
+n 1,
the period matrix takes the canonical form (

, Z), with

= diag(
1
, . . . ,
n
) and Z =

U
1
V, symmetric and Im > 0.
Then

=
_

g
0
+n1

,
g
0
+n1

Z
1

_
,
= (
g
0
+n1

,
g
0
+n1
U
1
V
1

),
=
_

g
0
+n1

,
g
0
+n1

(U

)
1
V

_
,
and

= (U

, V

) = (A B D E),
is the period matrix of the dual abelian variety Prym

(C/C
0
). The above discussion is summed up in the following statement:
Theorem 1. Let : C C
0
be a double covering where C and C
0
are nonsingular algebraic curves with jacobians Jac(C)
and Jac(C
0
). Let : C C be the involution exchanging sheets of C over C
0
= C/. This involution extends by linearity to
a map (which will again be denoted by ) : Jac(C) Jac(C) and up some points of order two, Jac(C) splits into an even
part,i.e., Jac(C
0
) and an odd part (called Prym variety) denoted Prym(C/C
0
) and defined by
Prym(C/C
0
) =
_
H
0
_
C,
1
C
_

/H
1
(C, Z)

,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1071
where
1
C
is the sheaf of holomorphic 1-forms on C and denote the 1 eigenspace for a vector space on which j acts. To be
precise, we have
Jac(C) = Prym(C/C
0
) Jac (C
0
) ,
with dimJac (C
0
) = genus g
0
of C
0
, dimJac (C) = genus g of C = 2g
0
+n 1 and dimPrym(C/C
0
) = g g
0
= g
0
+n +1,
with 2n branch points. Moreover, if
=
_
A B C D E F
G H I J K L
_
,
is the period matrix of Jac(C) where A, . . . , L denote the matrices (1), then the period matrices of Jac (C
0
) , Prym(C/C
0
) and the
dual Prym

(C/C
0
) of Prym(C/C
0
), are respectively = (G H), = (2A B 2D E) and

= (A B D E).
3. Algebraic complete integrability
3.1. A survey on abelian varieties and algebraic integrability
We give some basic facts about integrable hamiltonian systems and some results about abelian surfaces which will be
used, as well as the basic techniques to study two-dimensional algebraic completely integrable systems. Let T = C
n
/ be
a n-dimensional abelian variety where is the lattice generated by the 2n columns
1
, . . . ,
2n
of the n 2n period matrix
and let D be a divisor on T. Define L(D) = {f meromorphic on T : (f ) D}, i.e., for D =

k
j
D
j
a function f L(D)
has at worst a k
j
-fold pole along D
j
. The divisor D is called ample when a basis (f
0
, . . . , f
N
) of L(kD) embeds T smoothly
into P
N
for some k, via the map T P
N
, p [1 : f
1
(p) : ... : f
N
(p)], then kD is called very ample. It is known that every
positive divisor D on an irreducible abelian variety is ample and thus some multiple of D embeds M into P
N
. By a theorem
of Lefschetz, any k 3 will work. Moreover, there exists a complex basis of C
n
such that the lattice expressed in that basis
is generated by the columns of the n 2n period matrix
_
_
_
_

1
0 |
.
.
.
| Z
0
n
|
_
_
_
_
,
with Z

= Z, ImZ > 0,
j
N

and
j
|
j+1
. The integers
j
which provide the so-called polarization of the abelian variety T
are then related to the divisor as follows:
dimL(D) =
1
. . .
n
. (4)
In the case of a 2-dimensional abelian varieties (surfaces), even more can be stated: the geometric genus g of a positive
divisor D (containing possibly one or several curves) on a surface T is given by the adjunction formula
g(D) =
K
T
.D +D.D
2
+1, (5)
where K
T
is the canonical divisor on T, i.e., the zero-locus of a holomorphic 2-form, D.D denote the number of intersection
points of D with a + D (where a + D is a small translation by a of D on T), where as the RiemannRoch theorem for line
bundles on a surface tells you that
(D) = p
a
(T) +1 +
1
2
(D.D DK
T
), (6)
where p
a
(T) is the arithmetic genus of T and (D) the Euler characteristic of D. To study abelian surfaces using Riemann
surfaces on these surfaces, we recall that
(D) = dimH
0
(T, O
T
(D)) dimH
1
(T, O
T
(D)),
= dimL(D) dimH
1
(T,
2
(D K

T
)), (KodairaSerre duality),
= dimL(D), (Kodaira vanishing theorem), (7)
whenever D K

T
defines a positive line bundle. However for abelian surfaces, K
T
is trivial and p
a
(T) = 1; therefore
combining relations (4)(7), we obtain
(D) = dimL(D) =
D.D
2
= g (D) 1 =
1

2
. (8)
Let M be a 2n-dimensional differentiable manifold and a closed non-degenerate differential 2-form. The pair (M, ) is
called a symplectic manifold. Let H : M R be a smooth function. A hamiltonian system on (M, ) with hamiltonian H can
be written in the form
q
1
=
H
p
1
, . . . , q
n
=
H
p
n
, p
1
=
H
q
1
, . . . , p
n
=
H
q
n
, (9)
1072 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
where (q
1
, . . . , q
n
, p
1
, . . . , p
n
) are coordinates in M. Thus the hamiltonian vector field X
H
is defined by X
H
=

n
k=1
_
H
p
k

q
k

H
q
k

p
k
_
. If F is a smooth function on the manifold M, the Poisson bracket {F, H} of F and H is defined by
X
H
F =
n

k=1
_
H
p
k
F
q
k

H
q
k
F
p
k
_
= {F, H} . (10)
A function F is an invariant (first integral) of the hamiltonian system (9) if and only if the Lie derivative of F with respect X
H
is identically zero. The functions F and H are said to be in involution or to commute, if {F, H} = 0. Note that Eqs. (9) and (10)
can be written in more compact form
x = J
H
x
, x = (q
1
, . . . , q
n
, p
1
, . . . , p
n
)

,
{F, H} =
_
F
x
, J
H
x
_
=
n

k,l=1
J
kl
F
x
k
H
x
l
,
with J =
_
O I
I O
_
, a skew-symmetric matrix where I is the n n unit matrix and O the n n zero matrix. A hamiltonian
systemis completely integrable in the sense of Liouville if there exist n invariants H
1
= H, H
2
, . . . , H
n
in involution (i.e., such
that the associated Poisson bracket {H
k
, H
l
} all vanish) with linearly independent gradients (i.e., dH
1
dH
n
= 0). For
generic c = (c
1
, . . . , c
n
) the level set M
c
= {H
1
= c
1
, . . . , H
n
= c
n
} will be an n-manifold, and since X
H
k
H
l
= {H
k
, H
l
} = 0,
the integral curves of each X
H
k
will lie in M
c
and the vector fields X
H
k
span the tangent space of M
c
. By a theorem of
Arnold [4,15], if M
c
is compact and connected, it is diffeomorphic to an n-dimensional real torus and each vector field will
define a linear flow there. To be precise, in some open neighbourhood of the torus one can introduce regular symplectic
coordinates s
1
, . . . , s
n
,
1
, . . . ,
n
in which takes the canonical form =

n
k=1
ds
k
d
k
. Here the functions s
k
(called
action-variables) give coordinates in the direction transverse to the torus and can be expressed functionally in terms of
the first integrals H
k
. The functions
k
(called angle-variables) give standard angular coordinates on the torus, and every
vector field X
H
k
can be written in the form
k
= h
k
(s
1
, . . . , s
n
), that is, its integral trajectories define a conditionally-periodic
motion on the torus. Consequently, in a neighbourhood of the torus the hamiltonian vector field X
H
k
take the following form
s
k
= 0,
k
= h
k
(s
1
, . . . , s
n
), and can be solved by quadratures.
Consider now hamiltonian problems of the form
X
H
: x = J
H
x
f (x), x R
m
, (11)
where H is the hamiltonian and J = J(x) is a skew-symmetric matrix with polynomial entries in x, for which the
corresponding Poisson bracket {H
i
, H
j
} =
H
i
x
, J
H
j
x
, satisfies the Jacobi identities. The system (11) with polynomial right
hand side will be called algebraically completely integrable (a.c.i.) in the sense of Adlervan Moerbeke [1] when:
(i) The system possesses n + k independent polynomial invariants H
1
, . . . , H
n+k
of which k lead to zero vector fields
J
H
n+i
x
(x) = 0, 1 i k, the n remaining ones are in involution (i.e.,
_
H
i
, H
j
_
= 0) and m = 2n +k. For most values of c
i
R,
the invariant varieties
_
n+k
i=1
{x R
m
: H
i
= c
i
} are assumed compact and connected. Then, according to the ArnoldLiouville
theorem, there exists a diffeomorphism
n+k
_
i=1
_
x R
m
: H
i
= c
i
_
R
n
/Lattice,
and the solutions of the system (11) are straight lines motions on these tori.
(ii) The invariant varieties, thought of as affine varieties
A =
n+k
_
i=1
{H
i
= c
i
, x C
m
},
in C
m
can be completed into complex algebraic tori, i.e., AD = C
n
/Lattice, where C
n
/Lattice is a complex algebraic torus
(i.e., abelian variety) and D a divisor.
Remark 3.1. Algebraic means that the torus can be defined as an intersection
_
M
i=1
{P
i
(X
0
, . . . , X
N
) = 0} involving a large
number of homogeneous polynomials P
i
. In the natural coordinates (t
1
, . . . , t
n
) of C
n
/Lattice coming fromC
n
, the functions
x
i
= x
i
(t
1
, . . . , t
n
) are meromorphic and (11) defines straight line motion on C
n
/Lattice. Condition (i) means, in particular,
there is an algebraic map (x
1
(t), . . . , x
m
(t)) (
1
(t), . . . ,
n
(t)) making the following sums linear in t:
n

i=1
_

i
(t)

i
(0)

j
= d
j
t, 1 j n, d
j
C,
where
1
, . . . ,
n
denote holomorphic differentials on some algebraic curves.
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1073
Adler and van Moerbeke [1] have shown that the existence of a coherent set of Laurent solutions:
x
i
=

j=0
x
(j)
i
t
jk
i
, k
i
Z, some k
i
> 0, (12)
depending on dim (phase space) 1 = m 1 free parameters is necessary and sufficient for a hamiltonian system with
the right number of constants of motion to be a.c.i. So, if the hamiltonian flow (11) is a.c.i., it means that the variables
x
i
are meromorphic on the torus C
n
/Lattice and by compactness they must blow up along a codimension one subvariety (a
divisor) D C
n
/Lattice. By the a.c.i. definition, the flow (11) is a straight line motion in C
n
/Lattice and thus it must hit the
divisor D in at least one place. Moreover through every point of D, there is a straight line motion and therefore a Laurent
expansion around that point of intersection. Hence the differential equations must admit Laurent expansions which depend
on the n 1 parameters defining D and the n + k constants c
i
defining the torus C
n
/Lattice , the total count is therefore
m1 = dim (phase space) 1 parameters.
Assume now hamiltonian flows to be (weight)-homogeneous with a weight
i
N, going with each variable x
i
, i.e.,
f
i
(

1
x
1
, . . . ,

m
x
m
) =

i
+1
f
i
(x
1
, . . . , x
m
) , C.
Observe that then the constants of the motion H can be chosen to be (weight)-homogeneous:
H(

1
x
1
, . . . ,

m
x
m
) =
k
H(x
1
, . . . , x
m
) , k Z.
If the flow is algebraically completely integrable, the differential equations (11) must admits Laurent series solutions (12)
depending onm1 free parameters. We must have k
i
=
i
andcoefficients inthe series must satisfy at the 0thstepnon-linear
equations,
f
i
_
x
(0)
1
, . . . , x
(0)
m
_
+g
i
x
(0)
i
= 0, 1 i m, (13)
and at the kth step, linear systems of equations:
(L kI) z
(k)
=
_
0 for k = 1
some polynomial in x
(1)
, . . . , x
(k1)
for k > 1,
(14)
where
L = Jacobian map of (13) =
f
z
+gI |
z=z
(0) .
If m 1 free parameters are to appear in the Laurent series, they must either come from the non-linear equations (13) or
from the eigenvalue problem (14), i.e., L must have at least m 1 integer eigenvalues. These are much less conditions than
expected, because of the fact that the homogeneity k of the constant H must be an eigenvalue of L. Moreover the formal series
solutions are convergent as a consequence of the majorant method [1]. So, the question is how does one prove directly that
the system is effectively a.c.i. with abelian space coordinates? The idea of the direct proof used by Adler and van Moerbeke
is closely related to the geometric spirit of the (real) ArnoldLiouville theorem discuted above. Namely, a compact complex
n-dimensional variety on which there exist n holomorphic commuting vector fields which are independent at every point is
analytically isomorphic to a n-dimensional complex torus C
n
/Lattice and the complex flows generated by the vector fields
are straight lines on this complex torus. Now, the affine variety Ais not compact and the main problem will be to complete
A into a non singular compact complex algebraic variety

A = A D in such a way that the vector fields X
H
1
, . . . , X
H
m
extend holomorphically along the divisor D and remain independent there. If this is possible,

A is an algebraic complex
torus, i.e., an abelian variety and the coordinates x
i
restricted to Aare abelian functions. A naive guess would be to take the
natural compactificationAof AinP
m
(C). Indeed, this cannever work for a general reason: Anabelianvariety

Aof dimension
bigger or equal than two is never a complete intersection, that is it can never be descriped in some projective space P
m
(C)
by m-dim

Aglobal polynomial homogeneous equations. In other words, if Ais to be the affine part of an abelian variety, A
must have a singularity somewhere along the locus at infinity A{X
0
= 0}. In fact, Adler and van Moerbeke [1] showed that
the existence of meromorphic solutions to the differential equations (2) depending on m 1 free parameters can be used
to manufacture the tori, without ever going through the delicate procedure of blowing up and down. Information about the
tori can then be gathered from the divisor. An exposition of such methods and their applications can be found in [1,2].
3.2. The HnonHeiles system
The HnonHeiles system
q
1
=
H
p
1
, q
2
=
H
p
2
, p
1
=
H
q
1
, p
2
=
H
q
2
, (15)
with
H H
1
=
1
2
_
p
2
1
+p
2
2
+Aq
2
1
+Bq
2
2
_
+q
2
1
q
2
+6q
3
2
,
1074 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
has another constant of motion
H
2
= q
4
1
+4q
2
1
q
2
2
4p
1
(p
1
q
2
p
2
q
1
) +4Aq
2
1
q
2
+(4A B)
_
p
2
1
+Aq
2
1
_
,
where A, B are constant parameters and q
1
, q
2
, p
1
, p
2
are canonical coordinates and momenta, respectively. First studied
as a mathematical model to describe the chaotic motion of a test star in an axisymmetric galactic mean gravitational field
this system is widely explored in other branches of physics. It well-known from applications in stellar dynamics, statistical
mechanics and quantummechanics. It provides a model for the oscillations of atoms in a three-atomic molecule. The system
(15) possesses Laurent series solutions depending on 3 free parameters , , , namely
q
1
=

t
+
_

3
12
+
A
2

B
12
_
t +t
2
+q
(4)
1
t
3
+q
(5)
1
t
4
+q
(6)
1
t
5
+ ,
q
2
=
1
t
2
+

2
12

B
12
+
_

4
48
+

2
A
10


2
B
60

B
2
240
_
t
2
+

3
t
3
+t
4
+ ,
where p
1
=
.
q
1
, p
2
=
.
q
2
and
q
(4)
1
=
AB
24


5
72
+
11
3
B
720

11
3
A
120

B
2
720

A
2
8
,
q
(5)
1
=

2
12
+
B
60

A
10
,
q
(6)
1
=

9


7
15552


5
A
2160
+

5
B
12960
+

3
B
2
25920
+

3
A
2
1440


3
AB
4320
+
AB
2
1440

B
3
19440

A
2
B
288
+
A
3
144
.
Let D be the pole solutions restricted to the surface
M
c
=
2
_
i=1
_
x (q
1
, q
2
, p
1
, p
2
) C
4
, H
i
(x) = c
i
_
,
to be precise D is the closure of the continuous components of the set of Laurent series solutions x (t) such that H
i
(x (t)) =
c
i
, 1 i 2, i.e., D = t
0
coefficient of M
c
. Thus we find an algebraic curve defined by
D :
2
= P
8
(), (16)
where
P
8
() =
7
15552

1
432
_
5A
13
18
B
_

1
36
_
671
15120
B
2
+
17
7
A
2

943
1260
BA
_

1
36
_
4A
3

1
2520
B
3

13
6
A
2
B +
2
9
AB
2

10
7
c
1
_

2
+
1
36
c
2
.
The curve D determined by an eight-order equation is smooth, hyperelliptic and its genus is 3. Moreover, the map
: D D, (, ) (, ), (17)
is an involution on D and the quotient E = D/ is an elliptic curve defined by
E :
2
= P
4
(), (18)
where P
4
() is the degree 4 polynomial in =
2
obtained from(16). The hyperelliptic curve D is thus a 2-sheeted ramified
covering of the elliptic curve E (18),
: D E, (, ) (, ), (19)
ramified at the four points covering = 0 and . Following the methods in Adlervan Moerbeke [1,2], the affine surface M
c
completes into an abelian surface

M
c
, by adjoining the divisor D. The latter defines on

M
c
a polarization (1, 2). The divisor
2D is very ample and the functions 1, y
1
, y
2
1
, y
2
, x
1
, x
2
1
+y
2
1
y
2
, x
2
y
1
2x
1
y
2
, x
1
x
2
+2Ay
1
y
2
+2y
1
y
2
2
, embed

M
c
smoothly into
CP
7
with polarization (2, 4), according to (8). Then the system (15) is algebraic complete integrable and the corresponding
flow evolves on an abelian surface

M
c
= C
2
/lattice, where the lattice is generated by the period matrix
_
2 0 a c
0 4 c b
_
,
Im
_
a c
c b
_
> 0.
Theorem 2. The abelian surface

M
c
which completes the affine surface M
c
is the dual Prym variety Prym

(D/E) of the
genus 3 hyperelliptic curve D (16) for the involution interchanging the sheets of the double covering (19) and the problem
linearizes on this variety.
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1075
Proof. Let (a
1
, a
2
, a
3
, b
1
, b
2
, b
3
) be a canonical homology basis of D such that (a
1
) = a
3
, (b
1
) = b
3
, (a
2
) =
a
2
, (b
2
) = b
2
, for the involution (17). As a basis of holomorphic differentials
0
,
1
,
2
on the curve D (16) we
take the differentials

1
=

2
d

,
2
=
d

,
3
=
d

,
and obviously

(
1
) =
1
,

(
2
) =
2
, .

(
3
) =
3
. We consider the period matrix of Jac(D)
=
_
_
_
_
_
_
_
_
a
1

1
_
a
2

1
_
a
3

1
_
b
1

1
_
b
2

1
_
b
3

1
_
a
1

2
_
a
2

2
_
a
3

2
_
b
1

2
_
b
2

2
_
b
3

2
_
a
1

3
_
a
2

3
_
a
3

3
_
b
1

3
_
b
2

3
_
b
3

3
_
_
_
_
_
_
_
.
By Theorem 1,
=
_
_
_
_
_
_
_
_
a
1

1
_
a
2

1

_
a
1

1
_
b
1

1
_
b
2

1

_
b
1

1
_
a
1

2
_
a
2

2

_
a
1

2
_
b
1

2
_
b
2

2

_
b
1

2
_
a
1

3
0
_
a
1

3
_
b
1

3
0
_
b
1

3
_
_
_
_
_
_
_
,
and therefore the period matrices of Jac(E) (i.e., E), Prym(D/E) and Prym

(D/E) are respectively = (

_
a
1

3
_
b
1

3
),
=
_
_
_
_
2
_
a
1

1
_
a
2

1
2
_
b
1

1
_
b
2

1
2
_
a
1

2
_
a
2

2
2
_
b
1

2
_
b
2

2
_
_
_
_
,
and

=
_
_
_
_
_
a
1

1
_
a
2

1
_
b
1

1
_
b
2

1
_
a
1

2
_
a
2

2
_
b
1

2
_
b
2

2
_
_
_
_
.
Let
L

=
_
_
_
3

i=1
m
i
_
a
i
_
_

3
_
_
+n
i
_
b
i
_
_

3
_
_
: m
i
, n
i
Z
_
_
_
,
be the period lattice associated to . Let us denote also by L

, the period lattice associated . According to Theorem 1, we

get the following diagram:
0

E D

0 ker N

Prym(D/E) E = Jac(D)
N

E 0

M
c
= M
c
2D C
2
/lattice

0
The polarization map : Prym(D/E)

M
c
= Prym

(D/E), has kernel (

E) Z
2
Z
2
and the induced polarization on
Prym(D/E) is of type (1,2). Let

M
c
C
2
/L

: p
_
p
p
0
_
dt
1
dt
2
_
,
be the uniformizing map, where dt
1
, dt
2
are two differentials on

M
c
corresponding to the flows generated respectively by
H
1
, H
2
such that: dt
1
|
D
=
1
and dt
2
|
D
=
2
,
L

=
_

_
4

k=1
n
k
_
_
_
_
_

k
dt
1
_

k
dt
2
_
_
_
_
: n
k
Z
_

_
,
1076 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
is the lattice associated to the period matrix
=
_
_
_
_
_

1
dt
1
_

2
dt
1
_

4
dt
1
_

4
dt
1
_

1
dt
2
_

2
dt
2
_

3
dt
2
_

4
dt
2
_
_
_
_
,
and (
1
,
2
,
3
,
4
) is a basis of H
1
(

M
c
, Z). By the Lefschetz theorem on hyperplane section [9], the map H
1
(D, Z)
H
1
(

M
c
, Z) induced by the inclusion D

M
c
is surjective and consequently we can find 4 cycles
1
,
2
,
3
,
4
on the curve
D such that
=
_
_
_
_
_

1
_

1
_

1
_

1
_

2
_

2
_

2
_

2
_
_
_
_
,
and
L

=
_

_
4

k=1
n
k
_
_
_
_
_

1
_

2
_
_
_
_
: n
k
Z
_

_
.
The cycles
1
,
2
,
3
,
4
in D which we look for are a
1
, b
1
, a
2
, b
2
and they generate H
1
(

M
c
, Z) such that
=
_
_
_
_
_
a
1

1
_
b
1

1
_
a
2

1
_
b
2

1
_
a
1

2
_
b
1

2
_
a
2

2
_
b
2

2
_
_
_
_
,
is a Riemann matrix. We show that =

,i.e., the period matrix of Prym

(D/E) dual of Prym(D/E). Consequently

M
c
and Prym

(D/E) are two abelian varieties analytically isomorphic to the same complex torus C
2
/L

. By Chows theorem,

A
c
and Prym

(D/E) are then algebraically isomorphic.

3.3. The Kowalewski rigid body motion
The motion for the Kowalewskis top is governed by the equations
.
m = m m+ l,
.

= m, (20)
where m, and l denote respectively the angular momentum, the directional cosine of the z-axis (fixed in space), the center
of gravity which after some rescaling and normalization may be taken as l = (1, 0, 0) and m = (m
1
/2, m
2
/2, m
3
/2). The
system (20) can be written
.
m
1
= m
2
m
3
,
.

1
= 2 m
3

2
m
2

3
,
.
m
2
= m
1
m
3
+2
3
,
.

2
= m
1

3
2m
3

1
,
.
m
3
= 2
2
,
.

3
= m
2

1
m
1

2
,
(21)
with constants of motion
H
1
=
1
2
_
m
2
1
+m
2
2
_
+m
2
3
+2
1
= c
1
,
H
2
= m
1

1
+m
2

2
+m
3

3
= c
2
, (22)
H
3
=
2
1
+
2
2
+
2
3
= c
3
= 1,
H
4
=
_
_
m
1
+im
2
2
_
2
(
1
+i
2
)
__
_
m
1
im
2
2
_
2
(
1
i
2
)
_
= c
4
.
The system (21) admits two distinct families of Laurent series solutions:
m
1
(t) =
_

1
t
+i
_

2
1
2
_

2
+ (t) ,

1
t
i
_

2
1
2
_

2
+ (t) ,

1
(t) =
_

_
1
2t
2
+ (t) ,
1
2t
2
+ (t) ,
m
2
(t) =
_

_
i
1
t

2
1

2
+ (t) ,
i
1
t

2
1

2
+ (t) ,

2
(t) =
_

_
i
2t
2
+ (t) ,
i
2t
2
+ (t) ,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1077
m
3
(t) =
_

_
i
t
+
1

2
+ (t) ,
i
t
+
1

2
+ (t) ,

3
(t) =
_

2
t
+ (t) ,

2
t
+ (t) ,
which depend on 5 free parameters
1
, . . . ,
5.
By substituting these series in the constants of the motion H
i
(22), one
eliminates three parameters linearly, leading to algebraic relation between the two remaining parameters, which is nothing
but the equation of the divisor D along which the m
i
,
i
blow up. Since the system (21) admits two families of Laurent
solutions, then D is a set of two isomorphic curves of genus 3, D = D
1
+D
1
:
D

: P (
1
,
2
) =
_

2
1
1
_ __

2
1
1
_

2
2
P (
2
)
_
+c
4
= 0, (23)
where P (
2
) = c
1

2
2
2c
2

2
1 and = 1. Each of the curve D

is a 2 1 ramified cover (
1
,
2
, ) of elliptic curves D
0

:
D
0

:
2
= P
2
(
2
) 4c
4

4
2
, (24)
ramified at the 4 points
1
= 0 covering the 4 roots of P (
2
) = 0. It was shown by the author [14] that each divisor D

is
ample and defines a polarization (1, 2), whereas the divisor D, of geometric genus 9, is very ample and defines a polarization
(2, 4), accordind to (8). The affine surface M
c
=
_
4
i=1
{H
i
= c
i
} C
6
, defined by putting the four invariants (22) of the
Kowalewski flow (21) equal to generic constants, is the affine part of an abelian surface

M
c
with

M
c
\ M
c
= D = one genus 9 curve consisting of two genus 3 curves D

(23) intersecting in 4 points. Each

D

is a double cover of an elliptic curve D

0

(24) ramified at 4 points.

Moreover, the Hamiltonian flows generated by the vector fields X
H
1
and X
H
4
are straight lines on

M
c
. The 8 functions 1
, f
1
= m
1
, f
2
= m
2
, f
3
= m
3
, f
4
=
3
, f
5
= f
2
1
+ f
2
2
, f
6
= 4f
1
f
4
f
3
f
5
, f
7
= (f
2

1
f
1

2
) f
3
+ 2f
4

2
, form a basis of the
vector space of meromorphic functions on

M
c
with at worst a simple pole along D Moreover, the map

M
c
C
2
/Lattice CP
7
, (t
1
, t
2
) [(1, f
1
(t
1
, t
2
) , . . . , f
7
(t
1
, t
2
))] ,
is an embedding of

M
c
into CP
7
. Following the method (Section 3.1), we obtain the following theorem:
Theorem 3. The tori

M
c
can be identified as

M
c
= Prym

(D

/D
0

), i.e., dual of Prym(D

/D
0

) and the problem linearizes on this

Prym variety.
3.4. Kirchhoffs equations of motion of a solid in an ideal fluid
The Kirchhoffs equations of motion of a solid in an ideal fluid have the form
p
1
= p
2
H
l
3
p
3
H
l
2
,

l
1
= p
2
H
p
3
p
3
H
p
2
+l
2
H
l
3
l
3
H
l
2
,
p
2
= p
3
H
l
1
p
1
H
l
3
,

l
2
= p
3
H
p
1
p
1
H
p
3
+l
3
H
l
1
l
1
H
l
3
,
p
3
= p
1
H
l
2
p
2
H
l
1
,

l
3
= p
1
H
p
2
p
2
H
p
1
+l
1
H
l
2
l
2
H
l
1
,
(25)
where (p
1
, p
2
, p
3
) is the velocity of a point fixed relatively to the solid, (l
1
, l
2
, l
3
) the angular velocity of the body expressed
with regard to a frame of reference also fixed relatively to the solid and H is the hamiltonian. These equations can be regarded
as the equations of the geodesics of the right-invariant metric on the group E (3) = SO(3) R
3
of motions of 3-dimensional
euclidean space R
3
, generated by rotations and translations. Hence the motion has the trivial coadjoint orbit invariants p, p
and p, l. As it turns out, this is a special case of a more general system of equations written as
x = x
H
x
+y
H
y
, y = y
H
x
+x
H
y
,
where x = (x
1
, x
2
, x
3
) R
3
et y = (y
1
, y
2
, y
3
) R
3
. The first set can be obtained from the second by putting (x, y) = (l, p/)
and letting 0. The latter set of equations is the geodesic flowon SO(4) for a left invariant metric defined by the quadratic
form H. In Clebschs case, Eqs. (25) have the four invariants:
H
1
= H =
1
2
_
a
1
p
2
1
+a
2
p
2
2
+a
3
p
2
3
+b
1
l
2
1
+b
2
l
2
2
+b
3
l
2
3
_
,
H
2
= p
2
1
+p
2
2
+p
2
3
, (26)
H
3
= p
1
l
1
+p
2
l
2
+p
3
l
3
,
H
4
=
1
2
_
b
1
p
2
1
+b
2
p
2
2
+b
3
p
2
3
+
_
l
2
1
+l
2
2
+l
2
3
__
,
1078 A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079
with
a
2
a
3
b
1
+
a
3
a
1
b
2
+
a
1
a
2
b
3
= 0, and the constant satisfies the conditions =
b
1
(b
2
b
3
)
a
2
a
3
=
b
2
(b
3
b
1
)
a
3
a
1
=
b
3
(b
1
b
2
)
a
1
a
2
. The system
(25) can be written in the form (11) with m = 6; to be precise
x = f (x) J
H
x
, x = (p
1
, p
2
, p
3
, l
1
, l
2
, l
3
)

, (27)
where
J =
_
O P
P L
_
, P =
_
_
0 p
3
p
2
p
3
0 p
1
p
2
p
1
0
_
_
, L =
_
_
0 l
3
l
2
l
3
0 l
1
l
2
p
1
0
_
_
.
Consider points at infinity which are limit points of trajectories of the flow. In fact, there is a Laurent decomposition of such
asymptotic solutions,
x (t) = t
1
_
x
(0)
+x
(1)
t +x
(2)
t
2
+
_
, (28)
which depend on dim(phase space)1 = 5 free parameters. Putting (28) into (27), solving inductively for the x
(k)
, one finds
at the 0th step a non-linear equation, x
(0)
+f (x
(0)
) = 0, and at the kth step, a linear system of equations,
(L kI)x
(k)
=
_
0 for k = 1
quadratic polynomial in x
(1)
, . . . , x
(k)
for k 1,
where L denotes the jacobian map of the non-linear equation above. One parameter appear at the 0th step, i.e., in the
resolution of the non-linear equation and the 4 remaining ones at the kth step, k = 1, . . . , 4. Taking into account only
solutions trajectories lying on the invariant surface M
c
=
_
4
i=1
{H
i
(x) = c
i
} C
6
, we obtain one-parameter families which
are parameterized by a curve. To be precise we search for the set of Laurent solutions which remain confined to a fixed affine
invariant surface, related to specific values of c
1
, c
2
, c
3
, c
4
, i.e.,
D =
4
_
i=1
_
t
0
coefficient of H
i
(x (t)) = c
i
_
,
= an algebraic curve defined by
2
+c
1

2
+c
2

2
+c
3

2
+c
4
= 0, (29)
where is an arbitrary parameter and where = x
(0)
4
, = x
(0)
5
, = x
(0)
6
parameterizes the elliptic curve
E :
2
= d
2
1

2
1,
2
= d
2
2

2
+1, (30)
with d
1
, d
2
such that: d
2
1
+d
2
2
+1 = 0. The curve D is a 2-sheeted ramified covering of the elliptic curve E . The branch points
are defined by the 16 zeroes of c
1

2
+c
2

2
+c
3

2
+c
4
on E. The curve D is unramified at infinity and by Hurwitzs
formula, the genus of D is 9. Upon putting
2
, the curve D can also be seen as a 4-sheeted unramified covering of the
following curve of genus 3:
C :
_

2
+c
1

2
+
_
c
2

2
+c
3

2
_

_
2
c
2
4

2
= 0.
Moreover, the map : C C, (, ) (, ), is an involution on C and the quotient C
0
= C/ is an elliptic curve defined
by
C
0
:
2
= c
2
4

_
d
2
1
d
2
2

2
+
_
d
2
1
d
2
2
_
1
_
.
The curve C is a double ramified covering of C
0
, C C
0
, (, , ) (, ),
C :
_
_
_

2
= c
1

_
c
2

2
+c
3

2
_
+

2
= c
2
4

_
d
2
1
d
2
2

2
+
_
d
2
1
d
2
2
_
1
_
.
Let (a
1
, a
2
, a
3
, b
1
, b
2
, b
3
) be a canonical homology basis of C such that (a
1
) = a
3
, (b
1
) = b
3
, (a
2
) = a
2
and
(b
2
) = b
2
for the involution . Using the Poincar residue map [9], we show that

0
=
d

,
1
=
d

,
2
=
d

,
form a basis of holomorphic differentials on C and

(
0
) =
0
,

(
k
) =
k
(k = 1, 2). Haine [10] shows that the flow
evolves on an abelian surface

M
c
CP
7
of period matrix
_
2 0 a c
0 4 c b
_
, Im
_
a c
c b
_
> 0 and also identified

M
c
as Prymvariety
Prym(C/C
0
).
Theorem 4. The abelian surface

M
c
can be identified as Prym(C/C
0
). More precisely
4
_
i=1
_
x C
6
, H
i
(x) = c
i
_
= Prym(C/C
0
) \ D,
where D is a genus 9 curve (29), which is a ramified cover of an elliptic curve E (30) with 16 branch points.
A. Lesfari / Journal of Geometry and Physics 58 (2008) 10631079 1079
References
[1] M. Adler, P. van Moerbeke, The complex geometry of the KowalewskiPainlev analysis, Invent. Math. 97 (1989) 351.
[2] M. Adler, P. van Moerbeke, P. Vanhaecke, Algebraic integrability, Painlev geometry and Lie algebras, in: A Series of Modern Surveys in Mathematics,
vol. 47, Springer-Verlag, 2004.
[3] E. Arbarello, M. Cornalba, P.A. Griffiths, J. Harris, Geometry of Algebraic Curves I, Springer-Verlag, 1994.
[4] V.I. Arnold, Mathematical Methods in Classical Mechanics, Springer-Verlag, 1978.
[5] W. Barth, Abelian surfaces with (1, 2)-polarization, in: Conf. on Alg. Geom., Sendai, 1985, Adv. Stud. Pure Math. 10 (1987) 4184.
[6] A. Beauville, Prym varieties and the Schottky problem, Invent. Math. 41 (1977) 149196.
[7] A. Beauville, M.S. Narasimhan, S. Ramanan, Spectral curves and the generalized theta divisor, J. Reine Angew. Math. 398 (1989) 169179.
[8] E.D. Belokolos, A.I. Bobenko, V.Z. Enolskii, A.R. Its, V.B. Matveev, Algebro-Geometric Approach to Nonlinear Integrable Equations, Springer-Verlag,
1994.
[9] P.A. Griffiths, J. Harris, Principles of Algebraic Geometry, Wiley-Interscience, 1978.
[10] L. Haine, Geodesic flow on SO(4) and Abelian surfaces, Math. Ann. 263 (1983) 435472.
[11] N. Hitchin, Stable bundles and integrable sytems, Duke Math. J. 54 (1987) 91114.
[12] F. Ktter, Uber die Bewegung eines festen Krpers in einer Flssigkeit I, II, J. Reine Angew. Math. 109 (1892) 5181. 89111.
[13] S. Kowalewski, Sur le problme de la rotation dun corps solide autour dun point fixe, Acta Math. 12 (1989) 177232.
[14] A. Lesfari, Abelian surfaces and Kowalewskis top, Ann. Scient. cole Norm. Sup., Paris sr. 4 21 (1988) 193223.
[15] A. Lesfari, Completely integrable systems: Jacobis heritage, J. Geom. Phys. 31 (1999) 265286.
[16] A. Lesfari, Le thorme dArnoldLiouville et ses consquences, Elem. Math. 58 (1) (2003) 620.
[17] A. Lesfari, Le systme diffrentiel de HnonHeiles et les varits Prym, Pacific J. Math. 212 (1) (2003) 125132.
[18] D. Mumford, Prym varieties I, in: L.V. Ahlfors, I. Kra, B. Maskit, L. Niremberg (Eds.), Contributions to Analysis, Academic Press, New-York, 1974,
pp. 325350.
[19] S.P. Novikov, Two-dimensional Schrdinger operator and solitons: Three-dimensional integrable systems, in: Proceedings of the VIIIth International
Congress on Mathematical Physics, World Scientific, 1987, pp. 226241.