Journal of Geometry and Physics 58 (2008) 1063–1079

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 Chouaïb 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 Hénon–Heiles system, the Kowalewski rigid body motion and
Kirchhoff’s 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 Hénon–Heiles system........................................................................................................................................................... 1073
3.3. The Kowalewski rigid body motion............................................................................................................................................ 1076
3.4. Kirchhoff’s 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 σ is the involution on C interchanging sheets, then σ 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) 1063–1079
linearity to a map σ : 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 Kadomtsev–Petviashvili equation (KP equation), in the deformation theory of
two-dimensional Schrödinger operators [19], in relation to Calabi–Yau 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 Arnold–Liouville 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 Adler–van 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 Mumford’s 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 Hénon–Heiles system, the Kowalewski rigid body motion and Kirchhoff’s equations of motion of a solid in an
ideal fluid. The motivation was the excellent Haine’s paper [10] on the integration of the Euler–Arnold equations associated
to a class of geodesic flow on SO(4). The Kowalewski’s top and the Clebsch’s case of Kirchhoff’s 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 Kötter [12] as a result of complicated and mysterious computations. The concept of algebraic complete integrability
(Adler–vanMoerbeke) throws a completely newlight onthese two systems. Namely, inbothcases (see [14] for Kowalewski’s
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 ϕ : C −→ C
0
be a double covering with 2n branch points where C and C
0
are nonsingular complete curves. Let
σ : C −→C be the involution exchanging sheets of C over C
0
= C/σ. By Hurwitz’s formula, the genus of C is g = 2g
0
+n−1,
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
ω =
_
γ
ω, ∀ω ∈ Ω
1
C
,
where Ω
1
C
is the sheaf of holomorphic 1-forms on C and γ 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, Ω
1
C
)

/H
1
(C, Z) C
g
/Z
2g
,
via the Serre’s duality. Consider the following mapping
S
g
(C) −→ C
g
/L

,
g

k=1
µ
k
−→
g

k=1
_
µ
k
(t)
_
ω
1
, . . . , ω
g
_
= t
_
k
1
, . . . , k
g
_
,
where
_
ω
1
, . . . , ω
g
_
is a basis of Ω
1
C
, L

is the lattice associated to the period matrix Ω and µ
1
, . . . , µ
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 σ (a
1
) = a
g
0
+n
, . . . , σ
_
a
g
0
_
= a
g
, σ
_
a
g
0
+1
_
= −a
g
0
+1
, . . . , σ
_
a
g
0
+n−1
_
=
−a
g
0
+n−1
, σ (b
1
) = b
g
0
+n
, . . . , σ
_
b
g
0
_
= b
g
, σ
_
b
g
0
+1
_
= −b
g
0
+1
, . . . , σ
_
b
g
0
+n−1
_
= −b
g
0
+n−1
, for the involution σ. Notice
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1065
that ϕ
_
a
g
0
+1
_
, . . . , ϕ
_
a
g
0
+n−1
_
, ϕ
_
b
g
0
+1
_
, . . . , ϕ
_
b
g
0
+n−1
_
are homologous to zero on C
0
. Let
_
ω
1
, . . . , ω
g
_
be a basis of
holomorphic differentials on C where ω
g
0
+n
, . . . , ω
g
, are holomorphic differentials on C
0
and
σ


j
) =
_
−ω
j
, 1 ≤ j ≤ g
0
+n −1,
ω
j
, g
0
+n ≤ j ≤ g,
the pullback of ω
j
. The period matrix Ω of Jac(C) is explicitly given by
Ω =
_
A B C D E F
G H I J K L
_
,
where A, . . . , L denote the following matrices:
A =
_
_
_
_
_
_
_
_
_
a
1
ω
1
. . .
_
a
g
0
ω
1
.
.
.
.
.
.
_
a
1
ω
g
0
+n−1
. . .
_
a
g
0
ω
g
0
+n−1
_
_
_
_
_
_
_
_
, (1)
B =
_
_
_
_
_
_
_
_
_
a
g
0
+1
ω
1
. . .
_
a
g
0
+n−1
ω
1
.
.
.
.
.
.
_
a
g
0
+1
ω
g
0
+n−1
. . .
_
a
g
0
+n−1
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
C =
_
_
_
_
_
_
_
_
_
a
g
0
+n
ω
1
. . .
_
a
g
ω
1
.
.
.
.
.
.
_
a
g
0
+n
ω
g
0
+n−1
. . .
_
a
g
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
D =
_
_
_
_
_
_
_
_
_
b
1
ω
1
. . .
_
b
g
0
ω
1
.
.
.
.
.
.
_
b
1
ω
g
0
+n−1
. . .
_
b
g
0
ω
g
0+n−1
_
_
_
_
_
_
_
_
,
E =
_
_
_
_
_
_
_
_
_
b
g
0
+1
ω
1
. . .
_
b
g
0
+n−1
ω
1
.
.
.
.
.
.
_
b
g
0
+1
ω
g
0
+n−1
. . .
_
b
g
0
+n−1
ω
g
0+n−1
_
_
_
_
_
_
_
_
,
F =
_
_
_
_
_
_
_
_
_
b
g
0
+n
ω
1
. . .
_
b
g
ω
1
.
.
.
.
.
.
_
b
g
0
+n
ω
g
0
+n−1
. . .
_
b
g
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
G =
_
_
_
_
_
_
_
_
_
a
1
ω
g
0
+n
. . .
_
a
g
0
ω
g
0
+n
.
.
.
.
.
.
_
a
1
ω
g
. . .
_
a
g
0
ω
g
_
_
_
_
_
_
_
_
,
H =
_
_
_
_
_
_
_
_
_
a
g
0
+1
ω
g
0
+n
. . .
_
a
g
0
+n−1
ω
g
0
+n
.
.
.
.
.
.
_
a
g
0
+1
ω
g
. . .
_
a
g
0
+n−1
ω
g
_
_
_
_
_
_
_
_
,
1066 A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079
I =
_
_
_
_
_
_
_
_
_
a
g
0
+n
ω
g
0
+n
. . .
_
a
g
ω
g
0
+n
.
.
.
.
.
.
_
a
g
0
+n
ω
g
. . .
_
a
g
ω
g
_
_
_
_
_
_
_
_
,
J =
_
_
_
_
_
_
_
_
_
b
1
ω
g
0
+n
. . .
_
b
g
0
ω
g
0
+n
.
.
.
.
.
.
_
b
1
ω
g
. . .
_
b
g
0
ω
g
_
_
_
_
_
_
_
_
,
K =
_
_
_
_
_
_
_
_
_
b
g
0
+1
ω
g
0
+n
. . .
_
b
g
0
+n−1
ω
g
0
+n
.
.
.
.
.
.
_
b
g
0
+1
ω
g
. . .
_
b
g
0
+n−1
ω
g
_
_
_
_
_
_
_
_
,
and
L =
_
_
_
_
_
_
_
_
_
b
g
0
+n
ω
g
0
+n
. . .
_
b
g
ω
g
0
+n
.
.
.
.
.
.
_
b
g
0
+n
ω
g
. . .
_
b
g
ω
g
_
_
_
_
_
_
_
_
.
Notice that
_
a
g
0
+1
ω
j
= −
_
σ(a
g
0
+1
)
ω
j
,
= −
_
a
g
0
+1
σ


j
),
=
_
¸
¸
_
¸
¸
_
_
a
g
0
+1
ω
j
, 1 ≤ j ≤ g
0
+n −1,

_
a
g
0
+1
ω
j
, g
0
+n ≤ j ≤ g,
.
.
.
_
a
g
0
+n−1
ω
j
=
_
¸
¸
¸
_
¸
¸
¸
_
_
a
g
0
+n−1
ω
j
, 1 ≤ j ≤ g
0
+n −1,

_
a
g
0
+n−1
ω
j
, g
0
+n ≤ j ≤ g,
_
b
g
0
+1
ω
j
= −
_
σ(b
g
0
+1)
ω
j
,
= −
_
b
g
0
+1
σ

_
ω
j
_
,
=
_
¸
¸
_
¸
¸
_
_
b
g
0
+1
ω
j
, 1 ≤ j ≤ g
0
+n −1,

_
b
g
0
+1
ω
j
, g
0
+n ≤ j ≤ g,
.
.
.
_
b
g
0
+1
ω
j
=
_
¸
¸
_
¸
¸
_
_
b
g
0
+1
ω
j
, 1 ≤ j ≤ g
0
+n −1,

_
b
g
0
+1
ω
j
, g
0
+n ≤ j ≤ g,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1067
_
a
g
0
+n
ω
j
=
_
σ(a
1
)
ω
j
,
=
_
a
1
σ

_
ω
j
_
,
=
_
¸
¸
_
¸
¸
_

_
a
1
ω
j
, 1 ≤ j ≤ g
0
+n −1,
_
a
1
ω
j
, g
0
+n ≤ j ≤ g,
.
.
.
_
a
g
0
+n
ω
j
=
_
¸
¸
¸
_
¸
¸
¸
_

_
a
g
0
ω
j
, 1 ≤ j ≤ g
0
+n −1,
_
a
g
0
ω
j
, g
0
+n ≤ j ≤ g,
and
_
b
g
0
+n
ω
j
=
_
σ(b
1
)
ω
j
,
=
_
b
1
σ

_
ω
j
_
,
=
_
¸
¸
_
¸
¸
_

_
b
1
ω
j
, 1 ≤ j ≤ g
0
+n −1,
_
b
1
ω
j
, g
0
+n ≤ j ≤ g,
.
.
.
_
b
g
ω
j
=
_
¸
¸
_
¸
¸
_

_
b
g
0
ω
j
, 1 ≤ j ≤ g
0
+n −1,
_
b
g
0
ω
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
Ω =
_
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
_
,
=
_
Γ O
O 2∆
_
,
where
∆ =
_
G J
_
,
=
_
_
_
_
_
_
_
_
_
a
1
ω
g
0
+n
. . .
_
a
g
0
ω
g
0
+n
_
b
1
ω
g
0
+n
. . .
_
b
g
0
ω
g
0
+n
.
.
.
.
.
.
.
.
.
.
.
.
_
a
1
ω
g
. . .
_
a
g
0
ω
g
_
b
1
ω
g
. . .
_
b
g
0
ω
g
_
_
_
_
_
_
_
_
, (2)
1068 A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079
and
Γ =
_
2A B 2D E
_
,
=
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
2
_
a
1
ω
1
. . . 2
_
a
1
ω
g
0
+n−1
.
.
.
.
.
.
2
_
a
g
0
ω
1
. . . 2
_
a
g
0
ω
g
0
+n−1
_
a
g
0
+1
ω
1
. . .
_
a
g
0
+1
ω
g
0
+n−1
.
.
.
.
.
.
_
a
g
0
+n−1
ω
1
. . .
_
a
g
0
+n−1
ω
g
0
+n−1
2
_
b
1
ω
1
. . . 2
_
b
1
ω
g
0
+n−1
.
.
.
.
.
.
2
_
b
g
0
ω
1
. . . 2
_
b
g
0
ω
g
0
+n−1
_
b
g
0
+1
ω
1
. . .
_
b
g
0
+1
ω
g
0
+n−1
.
.
.
.
.
.
_
b
g
0
+n−1
ω
1
. . .
_
b
g
0
+n−1
ω
g
0
+n−1
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_

. (3)
Let
L

=
_
¸
¸
_
¸
¸
_
g

i=1
m
i
_
a
i
_
_
_
_
ω
1
.
.
.
ω
g
_
_
_
_
+n
i
_
b
i
_
_
_
_
ω
1
.
.
.
ω
g
_
_
_
_
: m
i
, n
i
∈ Z
_
¸
¸
_
¸
¸
_
,
be the period lattice associated to Ω. Let us denote also by L

1
, L

2
, L

3
and L

the period lattices associated respectively to

1
, Ω
2
, Ω
3
and ∆. Since L

= L

1
= L

2
= L

3
, it follows that the maps
C
g
0
+n−1
/L
Γ
:
_
_
_
_
t
1
.
.
.
t
g
0
+n−1
_
_
_
_
mod L
Γ
→ C
g
/L

:
_
_
_
_
_
_
_
_
_
_
_
t
1
.
.
.
t
g
0
+n−1
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
+n−1
/L
Γ
and C
g
0
/2L

can be embedded into C
g
/L

and the map
C
g
/L

3
= C
g
0
+n−1
/L
Γ
⊕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) 1063–1079 1069
shows that the jacobian variety Jac (C
0
) intersects the torus C
g
0
+n−1
/L
Γ
in 2
2g
0
points. We have then the following diagram
C Ker N
ϕ
↓ σ ↓
C
u
−→ Jac(C)
u

←− Jac

(C)
↓ ϕ ↓ N
ϕ
↑ N

ϕ
C
0
u
0
−→ Jac(C
0
)
u

0
←− Jac

(C
0
)
↓ ϕ

Jac(C) ⊂ H
1
(C, O

C
) H
1
(C
0
, (ϕ

O
C
)

)
↓ σ ↓ N
ϕ
Jac(C)
N
ϕ
−→ 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
= ϕ(p),
N
ϕ
: Jac(C) → Jac(C
0
),

m
i
p
i


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

(C) = dual of Jac(C). The norm map N
ϕ
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, Ω
1
C
_

_

/H
1
(C, Z)

,
where −denote the −1 eigenspace for a vector space on which the involution σ acts. Let D ∈ Prym(C/C
0
), i.e.,
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
0
.
.
.
0
2
_
D
0
ω
g
0
+n
.
.
.
2
_
D
0
ω
g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
∈ L

⇐⇒
_
_
_
_
_
_
_
_
D
0
ω
g
0
+n
.
.
.
_
D
0
ω
g
_
_
_
_
_
_
_
∈ L

,
⇐⇒
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
D
0
ω
1
.
.
.
_
D
0
ω
g
0
+n−1
_
D
0
ω
g
0
+n
.
.
.
_
D
0
ω
g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
_
_
t
1
.
.
.
t
g
0
+n−1
0
.
.
.
0
_
_
_
_
_
_
_
_
_
_
_
.
Consequently, the Abel–Jacobi map
Jac (C) −→C
g
/L

, D −→
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
D
0
ω
1
.
.
.
_
D
0
ω
g
0
+n−1
_
D
0
ω
g
0
+n
.
.
.
_
D
0
ω
g
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
,
maps Prym(C/C
0
) bi-analtically onto C
g
0
+n−1
/L
Γ
. By Chow’s theorem[9], Prym(C/C
0
) is thus a subabelian variety of Jac (C).
More precisely, we have
Prym(C/C
0
) = (Ker N
ϕ
)
0
,
= Ker(1
Jac(C)
+σ)
0
,
= Im(1
Jac(C)
−σ) ⊂ Jac(C).
1070 A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079
Equivalently, the involution σ induces an involution
σ : Jac(C) −→ Jac(C), class of D −→ class of σD,
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. Γ (3)) is the period matrix of Jac (C
0
) (resp. Prym(C/C
0
)). Write Γ = (U, V), with
U =
_
_
_
_
_
_
_
_
2
_
a
1
ω
1
. . . 2
_
a
g
0
ω
1
_
a
g
0
+1
ω
1
. . .
_
a
g
0
+n−1
ω
1
.
.
.
.
.
.
.
.
.
.
.
.
2
_
a
1
ω
g
0
+n−1
. . . 2
_
a
g
0
ω
g
0
+n−1
_
a
g
0
+1
ω
g
0
+n−1
. . .
_
a
g
0
+n−1
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
V =
_
_
_
_
_
_
_
_
2
_
b
1
ω
1
. . . 2
_
b
g
0
ω
1
_
b
g
0
+1
ω
1
. . .
_
b
g
0
+n−1
ω
1
.
.
.
.
.
.
.
.
.
.
.
.
2
_
b
1
ω
g
0
+n−1
. . . 2
_
b
g
0
ω
g
0
+n−1
_
b
g
0
+1
ω
g
0
+n−1
. . .
_
b
g
0
+n−1
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
and let us call
e
1
=
_
_
_
_
_
_
_
2
_
a
1
ω
1
.
.
.
2
_
a
1
ω
g
0
+n−1
_
_
_
_
_
_
_
, . . . , e
g
0
=
_
_
_
_
_
_
_
_
2
_
a
g
0
ω
1
.
.
.
2
_
a
g
0
ω
g
0
+n−1
_
_
_
_
_
_
_
_
,
e
g
0
+1
=
_
_
_
_
_
_
_
_
_
a
g
0
+1
ω
1
.
.
.
_
a
g
0
+1
_
_
_
_
_
_
_
_
, . . . , e
g
0
+n−1
=
_
_
_
_
_
_
_
_
_
a
g
0
+n−1
ω
1
.
.
.
_
a
g
0
+n−1
ω
g
0
+n−1
_
_
_
_
_
_
_
_
.
Then, in the new basis
_
λ
1
, . . . , λ
g
0
+n−1
_
where
λ
j
=
e
j
δ
j
, δ
j
=
_
1 pour 1 ≤ j ≤ g
0
,
2 pour g
0
+1 ≤ j ≤ g
0
+n −1,
the period matrix Γ takes the canonical form (∆
δ
, Z), with ∆
δ
= diag(δ
1
, . . . , δ
n
) and Z = ∆
δ
U
−1
V, symmetric and Im > 0.
Then
Γ

=
_
δ
g
0
+n−1

−1
δ
, δ
g
0
+n−1

−1
δ
Z∆
−1
δ
_
,
= (δ
g
0
+n−1

−1
δ
, δ
g
0
+n−1
U
−1
V∆
−1
δ
),
=
_
δ
g
0
+n−1

−1
δ
, δ
g
0
+n−1

−1
δ
(U

)
−1
V

_
,
and
Γ

= (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 ϕ : 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 σ : C −→ C be the involution exchanging sheets of C over C
0
= C/σ. This involution extends by linearity to
a map (which will again be denoted by σ) σ : 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, Ω
1
C
_

_

/H
1
(C, Z)

,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1071
where Ω
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
Ω =
_
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), Γ = (2A B 2D E) and Γ

= (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
/Λ be
a n-dimensional abelian variety where Λ is the lattice generated by the 2n columns λ
1
, . . . , λ
2n
of the n ×2n period matrix
Ω 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
_
_
_
_
δ
1
0 |
.
.
.
| Z
0 δ
n
|
_
_
_
_
,
with Z

= Z, ImZ > 0, δ
j
∈ N

and δ
j

j+1
. The integers δ
j
which provide the so-called polarization of the abelian variety T
are then related to the divisor as follows:
dimL(D) = δ
1
. . . δ
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 Riemann–Roch theorem for line
bundles on a surface tells you that
χ(D) = p
a
(T) +1 +
1
2
(D.D −DK
T
), (6)
where p
a
(T) is the arithmetic genus of T and χ(D) the Euler characteristic of D. To study abelian surfaces using Riemann
surfaces on these surfaces, we recall that
χ(D) = dimH
0
(T, O
T
(D)) −dimH
1
(T, O
T
(D)),
= dimL(D) −dimH
1
(T, Ω
2
(D ⊗K

T
)), (Kodaira–Serre 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
χ(D) = dimL(D) =
D.D
2
= g (D) −1 = δ
1
δ
2
. (8)
Let M be a 2n-dimensional differentiable manifold and ω a closed non-degenerate differential 2-form. The pair (M, ω) is
called a symplectic manifold. Let H : M → R be a smooth function. A hamiltonian system on (M, ω) 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) 1063–1079
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
, ϕ
1
, . . . , ϕ
n
in which ω takes the canonical form ω =

n
k=1
ds
k
∧ dϕ
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 ϕ
k
(called angle-variables) give standard angular coordinates on the torus, and every
vector field X
H
k
can be written in the form ˙ ϕ
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, ˙ ϕ
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 Adler–van 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 Arnold–Liouville
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., A∪D = 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)) → (µ
1
(t), . . . , µ
n
(t)) making the following sums linear in t:
n

i=1
_
µ
i
(t)
µ
i
(0)
ω
j
= d
j
t, 1 ≤ j ≤ n, d
j
∈ C,
where ω
1
, . . . , ω
n
denote holomorphic differentials on some algebraic curves.
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 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
j−k
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
m−1 = dim (phase space) −1 parameters.
Assume now hamiltonian flows to be (weight)-homogeneous with a weight ν
i
∈ N, going with each variable x
i
, i.e.,
f
i

ν
1
x
1
, . . . , α
ν
m
x
m
) = α
ν
i
+1
f
i
(x
1
, . . . , x
m
) , ∀α ∈ C.
Observe that then the constants of the motion H can be chosen to be (weight)-homogeneous:
H(α
ν
1
x
1
, . . . , α
ν
m
x
m
) = α
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 onm−1 free parameters. We must have k
i
= ν
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
(k−1)
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) Arnold–Liouville 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 Hénon–Heiles system
The Hénon–Heiles 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) 1063–1079
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 α, β, γ, namely
q
1
=
α
t
+
_
α
3
12
+
αA
2

αB
12
_
t +βt
2
+q
(4)
1
t
3
+q
(5)
1
t
4
+q
(6)
1
t
5
+· · · ,
q
2
= −
1
t
2
+
α
2
12

B
12
+
_
α
4
48
+
α
2
A
10

α
2
B
60

B
2
240
_
t
2
+
αβ
3
t
3
+γt
4
+· · · ,
where p
1
=
.
q
1
, p
2
=
.
q
2
and
q
(4)
1
=
αAB
24

α
5
72
+
11α
3
B
720

11α
3
A
120

αB
2
720

αA
2
8
,
q
(5)
1
= −
βα
2
12
+
βB
60


10
,
q
(6)
1
= −
αγ
9

α
7
15552

α
5
A
2160
+
α
5
B
12960
+
α
3
B
2
25920
+
α
3
A
2
1440

α
3
AB
4320
+
αAB
2
1440

αB
3
19440

αA
2
B
288
+
αA
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 : β
2
= P
8
(α), (16)
where
P
8
(α) = −
7
15552
α
8

1
432
_
5A −
13
18
B
_
α
6

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

943
1260
BA
_
α
4

1
36
_
4A
3

1
2520
B
3

13
6
A
2
B +
2
9
AB
2

10
7
c
1
_
α
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
σ : D −→D, (β, α) −→ (β, −α), (17)
is an involution on D and the quotient E = D/σ is an elliptic curve defined by
E : β
2
= P
4
(ζ), (18)
where P
4
(ζ) is the degree 4 polynomial in ζ = α
2
obtained from(16). The hyperelliptic curve D is thus a 2-sheeted ramified
covering of the elliptic curve E (18),
ρ : D −→E, (β, α) −→ (β, ζ), (19)
ramified at the four points covering ζ = 0 and ∞. Following the methods in Adler–van 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 σ interchanging the sheets of the double covering ρ (19) and the problem
linearizes on this variety.
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1075
Proof. Let (a
1
, a
2
, a
3
, b
1
, b
2
, b
3
) be a canonical homology basis of D such that σ (a
1
) = a
3
, σ (b
1
) = b
3
, σ (a
2
) =
−a
2
, σ (b
2
) = −b
2
, for the involution σ (17). As a basis of holomorphic differentials ω
0
, ω
1
, ω
2
on the curve D (16) we
take the differentials
ω
1
=
α
2

β
, ω
2
=

β
, ω
3
=
αdα
β
,
and obviously σ


1
) = −ω
1
, σ


2
) = −ω
2
, .σ


3
) = ω
3
. We consider the period matrix Ω of Jac(D)
Ω =
_
_
_
_
_
_
_
_
a
1
ω
1
_
a
2
ω
1
_
a
3
ω
1
_
b
1
ω
1
_
b
2
ω
1
_
b
3
ω
1
_
a
1
ω
2
_
a
2
ω
2
_
a
3
ω
2
_
b
1
ω
2
_
b
2
ω
2
_
b
3
ω
2
_
a
1
ω
3
_
a
2
ω
3
_
a
3
ω
3
_
b
1
ω
3
_
b
2
ω
3
_
b
3
ω
3
_
_
_
_
_
_
_
.
By Theorem 1,
Ω =
_
_
_
_
_
_
_
_
a
1
ω
1
_
a
2
ω
1

_
a
1
ω
1
_
b
1
ω
1
_
b
2
ω
1

_
b
1
ω
1
_
a
1
ω
2
_
a
2
ω
2

_
a
1
ω
2
_
b
1
ω
2
_
b
2
ω
2

_
b
1
ω
2
_
a
1
ω
3
0
_
a
1
ω
3
_
b
1
ω
3
0
_
b
1
ω
3
_
_
_
_
_
_
_
,
and therefore the period matrices of Jac(E) (i.e., E), Prym(D/E) and Prym

(D/E) are respectively ∆ = (
_
a
1
ω
3
_
b
1
ω
3
),
Γ =
_
_
_
_
2
_
a
1
ω
1
_
a
2
ω
1
2
_
b
1
ω
1
_
b
2
ω
1
2
_
a
1
ω
2
_
a
2
ω
2
2
_
b
1
ω
2
_
b
2
ω
2
_
_
_
_
,
and
Γ

=
_
_
_
_
_
a
1
ω
1
_
a
2
ω
1
_
b
1
ω
1
_
b
2
ω
1
_
a
1
ω
2
_
a
2
ω
2
_
b
1
ω
2
_
b
2
ω
2
_
_
_
_
.
Let
L

=
_
_
_
3

i=1
m
i
_
a
i
_
_
ω
1
ω
2
ω
3
_
_
+n
i
_
b
i
_
_
ω
1
ω
2
ω
3
_
_
: m
i
, n
i
∈ Z
_
_
_
,
be the period lattice associated to Ω. Let us denote also by L

, the period lattice associated ∆. According to Theorem 1, we
get the following diagram:
0

E D
↓ ϕ

↓ ϕ
0 −→ ker N
ϕ
−→ Prym(D/E) ⊕E = Jac(D)
N
ϕ
−→ E −→ 0
τ ↓
¯
M
c
= M
c
∪ 2D C
2
/lattice

0
The polarization map τ : Prym(D/E) −→
¯
M
c
= Prym

(D/E), has kernel (ϕ

E) Z
2
×Z
2
and the induced polarization on
Prym(D/E) is of type (1,2). Let
¯
M
c
→ C
2
/L
Λ
: 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
= ω
1
and dt
2
|
D
= ω
2
,
L
Λ
=
_
¸
¸
_
¸
¸
_
4

k=1
n
k
_
_
_
_
_
ν
k
dt
1
_
ν
k
dt
2
_
_
_
_
: n
k
∈ Z
_
¸
¸
_
¸
¸
_
,
1076 A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079
is the lattice associated to the period matrix
Λ =
_
_
_
_
_
ν
1
dt
1
_
ν
2
dt
1
_
ν
4
dt
1
_
ν
4
dt
1
_
ν
1
dt
2
_
ν
2
dt
2
_
ν
3
dt
2
_
ν
4
dt
2
_
_
_
_
,
and (ν
1
, ν
2
, ν
3
, ν
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 ν
1
, ν
2
, ν
3
, ν
4
on the curve
D such that
Λ =
_
_
_
_
_
ν
1
ω
1
_
ν
2
ω
1
_
ν
4
ω
1
_
ν
4
ω
1
_
ν
1
ω
2
_
ν
2
ω
2
_
ν
3
ω
2
_
ν
4
ω
2
_
_
_
_
,
and
L
Λ
=
_
¸
¸
_
¸
¸
_
4

k=1
n
k
_
_
_
_
_
ν
k
ω
1
_
ν
k
ω
2
_
_
_
_
: n
k
∈ Z
_
¸
¸
_
¸
¸
_
.
The cycles ν
1
, ν
2
, ν
3
, ν
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
Λ =
_
_
_
_
_
a
1
ω
1
_
b
1
ω
1
_
a
2
ω
1
_
b
2
ω
1
_
a
1
ω
2
_
b
1
ω
2
_
a
2
ω
2
_
b
2
ω
2
_
_
_
_
,
is a Riemann matrix. We show that Λ = Γ

,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
Λ
. By Chow’s theorem,
¯
A
c
and Prym

(D/E) are then algebraically isomorphic.
3.3. The Kowalewski rigid body motion
The motion for the Kowalewski’s top is governed by the equations
.
m = m∧ λm+γ ∧ l,
.
γ
= γ ∧ λm, (20)
where m, γ 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 λm = (m
1
/2, m
2
/2, m
3
/2). The
system (20) can be written
.
m
1
= m
2
m
3
,
.
γ
1
= 2 m
3
γ
2
−m
2
γ
3
,
.
m
2
= −m
1
m
3
+2γ
3
,
.
γ
2
= m
1
γ
3
−2m
3
γ
1
,
.
m
3
= −2γ
2
,
.
γ
3
= m
2
γ
1
−m
1
γ
2
,
(21)
with constants of motion
H
1
=
1
2
_
m
2
1
+m
2
2
_
+m
2
3
+2γ
1
= c
1
,
H
2
= m
1
γ
1
+m
2
γ
2
+m
3
γ
3
= c
2
, (22)
H
3
= γ
2
1

2
2

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

1
t
−α
2
1
α
2
+◦ (t) ,
−iα
1
t
−α
2
1
α
2
+◦ (t) ,
γ
2
(t) =
_
¸
¸
_
¸
¸
_
i
2t
2
+◦ (t) ,
−i
2t
2
+◦ (t) ,
A. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1077
m
3
(t) =
_
¸
¸
_
¸
¸
_
i
t

1
α
2
+◦ (t) ,
−i
t

1
α
2
+◦ (t) ,
γ
3
(t) =
_
¸
_
¸
_
α
2
t
+◦ (t) ,
α
2
t
+◦ (t) ,
which depend on 5 free parameters α
1
, . . . , α
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
, γ
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
ε
: P (α
1
, α
2
) =
_
α
2
1
−1
_ __
α
2
1
−1
_
α
2
2
−P (α
2
)
_
+c
4
= 0, (23)
where P (α
2
) = c
1
α
2
2
−2εc
2
α
2
−1 and ε = ±1. Each of the curve D
ε
is a 2 −1 ramified cover (α
1
, α
2
, β) of elliptic curves D
0
ε
:
D
0
ε
: β
2
= P
2

2
) −4c
4
α
4
2
, (24)
ramified at the 4 points α
1
= 0 covering the 4 roots of P (α
2
) = 0. It was shown by the author [14] that each divisor D
ε
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
ε
(23) intersecting in 4 points. Each
D
ε
is a double cover of an elliptic curve D
0
ε
(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
= γ
3
, f
5
= f
2
1
+ f
2
2
, f
6
= 4f
1
f
4
− f
3
f
5
, f
7
= (f
2
γ
1
−f
1
γ
2
) f
3
+ 2f
4
γ
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
ε
/D
0
ε
), i.e., dual of Prym(D
ε
/D
0
ε
) and the problem linearizes on this
Prym variety.
3.4. Kirchhoff’s equations of motion of a solid in an ideal fluid
The Kirchhoff’s 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/ε)
and letting ε → 0. The latter set of equations is the geodesic flowon SO(4) for a left invariant metric defined by the quadratic
form H. In Clebsch’s 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) 1063–1079
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 θ
2
+c
1
β
2
γ
2
+c
2
α
2
γ
2
+c
3
α
2
β
2
+c
4
αβγ = 0, (29)
where θ is an arbitrary parameter and where α = x
(0)
4
, β = x
(0)
5
, γ = x
(0)
6
parameterizes the elliptic curve
E : β
2
= d
2
1
α
2
−1, γ
2
= d
2
2
α
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
β
2
γ
2
+c
2
α
2
γ
2
+c
3
α
2
β
2
+c
4
αβγ on E. The curve D is unramified at infinity and by Hurwitz’s
formula, the genus of D is 9. Upon putting ζ ≡ α
2
, the curve D can also be seen as a 4-sheeted unramified covering of the
following curve of genus 3:
C :
_
θ
2
+c
1
β
2
γ
2
+
_
c
2
γ
2
+c
3
β
2
_
ζ
_
2
−c
2
4
ζβ
2
γ
2
= 0.
Moreover, the map τ : C → C, (θ, ζ) → (−θ, ζ), is an involution on C and the quotient C
0
= C/τ is an elliptic curve defined
by
C
0
: η
2
= c
2
4
ζ
_
d
2
1
d
2
2
ζ
2
+
_
d
2
1
−d
2
2
_
ζ −1
_
.
The curve C is a double ramified covering of C
0
, C → C
0
, (θ, η, ζ) → (η, ζ),
C :
_
_
_
θ
2
= −c
1
β
2
γ
2

_
c
2
γ
2
+c
3
β
2
_
ζ +η
η
2
= c
2
4
ζ
_
d
2
1
d
2
2
ζ
2
+
_
d
2
1
−d
2
2
_
ζ −1
_
.
Let (a
1
, a
2
, a
3
, b
1
, b
2
, b
3
) be a canonical homology basis of C such that τ (a
1
) = a
3
, τ (b
1
) = b
3
, τ (a
2
) = −a
2
and
τ (b
2
) = −b
2
for the involution τ. Using the Poincaré residue map [9], we show that
ω
0
=

η
, ω
1
=
ζdζ
θη
, ω
2
=

θη
,
form a basis of holomorphic differentials on C and τ


0
) = ω
0
, τ


k
) = −ω
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) 1063–1079 1079
References
[1] M. Adler, P. van Moerbeke, The complex geometry of the Kowalewski–Painlevé analysis, Invent. Math. 97 (1989) 3–51.
[2] M. Adler, P. van Moerbeke, P. Vanhaecke, Algebraic integrability, Painlevé 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) 41–84.
[6] A. Beauville, Prym varieties and the Schottky problem, Invent. Math. 41 (1977) 149–196.
[7] A. Beauville, M.S. Narasimhan, S. Ramanan, Spectral curves and the generalized theta divisor, J. Reine Angew. Math. 398 (1989) 169–179.
[8] E.D. Belokolos, A.I. Bobenko, V.Z. Enol’skii, 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) 435–472.
[11] N. Hitchin, Stable bundles and integrable sytems, Duke Math. J. 54 (1987) 91–114.
[12] F. Kötter, Uber die Bewegung eines festen Körpers in einer Flüssigkeit I, II, J. Reine Angew. Math. 109 (1892) 51–81. 89–111.
[13] S. Kowalewski, Sur le problème de la rotation d’un corps solide autour d’un point fixe, Acta Math. 12 (1989) 177–232.
[14] A. Lesfari, Abelian surfaces and Kowalewski’s top, Ann. Scient. École Norm. Sup., Paris sér. 4 21 (1988) 193–223.
[15] A. Lesfari, Completely integrable systems: Jacobi’s heritage, J. Geom. Phys. 31 (1999) 265–286.
[16] A. Lesfari, Le théorème d’Arnold–Liouville et ses conséquences, Elem. Math. 58 (1) (2003) 6–20.
[17] A. Lesfari, Le système différentiel de Hénon–Heiles et les variétés Prym, Pacific J. Math. 212 (1) (2003) 125–132.
[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. 325–350.
[19] S.P. Novikov, Two-dimensional Schrödinger operator and solitons: Three-dimensional integrable systems, in: Proceedings of the VIIIth International
Congress on Mathematical Physics, World Scientific, 1987, pp. 226–241.

ωg = t k1 . in the study of the generalized theta divisors on the moduli spaces of stable vector bundles over an algebraic curve [7. By now many algebraic completely integrable systems are known to linearize on Prym varieties. the flows (run with complex time) generated by the constants of the motion are straight lines. i. 1 where ΩC is the sheaf of holomorphic 1-forms on C and γ is a closed path on C . The classical definition of Prym varieties deals with the unramified double covering of curves and was introduced by W. . It now plays a substantial role in some contemporary theories. Schottky and H. The method that is used for reveiling the Pryms is due to Haine [10]. Integrable hamiltonian systems are nonlinear ordinary differential equations described by a hamiltonian function and possessing sufficiently many independent constants of motion in involution. in both cases (see [14] for Kowalewski’s top and [10] for geodesic flow on SO(4)). .8. until revived by D. Let σ : C −→ C be the involution exchanging sheets of C over C0 = C /σ . . σ ag0 = ag . . Such Prym varieties are not principally polarized and so they are not isomorphic but only isogenous to Jacobi varieties of genus two hyperelliptic curves. . . The motivation was the excellent Haine’s paper [10] on the integration of the Euler–Arnold equations associated to a class of geodesic flow on SO(4). 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. . be a canonical homology basis of H1 (C . Moreover. 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 Hénon–Heiles system. . ag . σ bg0 +1 = −bg0 +1 . . . . Notice . On Sg (C ). The concept of algebraic complete integrability (Adler–van Moerbeke) throws a completely new light on these two systems. in relation to Calabi–Yau three-folds and string theory. .17]. OC )/H1 (C . the Kowalewski rigid body motion and Kirchhoff’s equations of motion of a solid in an ideal fluid. via the Serre’s duality. Griffiths. σ ag0 +1 = −ag0 +1 . 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). kg . b1 . . By the Arnold–Liouville theorem [4. two divisors D1 and D2 are linearly equivalent (in short. . Let (a1 . µg some appropriate variables defined on a non empty Zariski open set. . LΩ is the lattice associated to the period matrix Ω and µ1 . . . σ ag0 +n−1 = −ag0 +n−1 . . . in the coordinates of this abelian variety. Z) such that σ (a1 ) = ag0 +n . . . . σ bg0 +n−1 = −bg0 +n−1 . We define the jacobian (Jacobi variety) of C . . .2. bg0 . . 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. Z) Cg /Z2g . . ΩC )∗ /H1 (C .1064 A.14. Namely. Harris [3] and in Mumford’s paper [18]. 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. Mumford around 1970. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 linearity to a map σ : Jac(C ) −→ Jac(C ). . g µk −→ g µ k (t ) ω1 . A dynamical system is algebraic completely integrable (in the sense of Adler–van Moerbeke [1]) if it can be linearized on a complex algebraic torus Cn /lattice (=abelian variety). . from the fundamental exponential sequence. W. k=1 k=1 1 where ω1 . Z) 0 1 H0 (C . . Cornalba. a complex torus. . ωg is a basis of ΩC . the genus of C is g = 2g0 + n − 1. . the affine varieties in C6 obtained by intersecting the four polynomial invariants of the flow are affine parts of Prym varieties of genus 3 curves which are double cover of elliptic curves.11]. This was a total surprise as it was generically believed that only jacobians would appear as invariants manifolds of such systems. . the jacobian Jac(C ) = Pic (C ) of C is the connected component of its Picard group parametrizing degree 0 invertible sheaves. bg ). ∀ω ∈ ΩC . .10. the Kadomtsev–Petviashvili equation (KP equation). By Hurwitz’s formula.16]. . . Up to some points of order two. . E. 0 to be Jac(C ) = Sg (C )/ ≡. σ (b1 ) = bg0 +n .15. In the second part of the paper. One of the remarkable developments in recent mathematics is the interplay between algebraic completely integrable systems and Prym varieties.e. . . . . for the involution σ . are meromorphic functions on an abelian variety. we get an isomorphism Jac(C ) = Pic (C ) = H1 (C . 2. . The basic algebraic tools are known and can be found in the book by Arbarello. 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. Prym varieties Let ϕ : C −→ C0 be a double covering with 2n branch points where C and C0 are nonsingular complete curves. The period of these Prym varieties provide the exact periods of the motion in terms of explicit abelian integrals. . . Let Sg (C ) be the g-th symmetric power of C (the totality of unordered sets of points of C ). ag0 . . in the deformation theory of two-dimensional Schrödinger operators [19]. σ bg0 = bg .. . The theory of Prym varieties was dormant for a long time. . . Indeed. . To be precise. for example integrable systems [1. where g0 is the genus of C0 . It is a compact commutative algebraic group. Consider the following mapping Sg (C ) −→ Cg /LΩ .. . The Kowalewski’s top and the Clebsch’s case of Kirchhoff’s 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 Kötter [12] as a result of complicated and mysterious computations. Jac(C ) splits into an even part and an odd part: the even part is Jac(C0 ) and the odd part is a Prym(C /C0 ).5. D1 ≡ D2 ) D 1 if their difference D1 − D2 is the divisor of a meromorphic function or equivalently if and only if D12 ω = γ ω. . Let OC be the sheaf of holomorphic functions on C .

 ... ag0  . .. . ωg be a basis of holomorphic differentials on C where ωg0 +n ... . . . ag +n−1 0  . are holomorphic differentials on C0 and σ ∗ (ωj ) = −ωj . . . ag +n−1 0 . ag ωg0 +n−1 ω1    ω1 . ..     bg0 +n ... .. . .  . ωj . . ..   ωg0 +n−1     . . ωg . ag ω1 . . L denote the following matrices:     A=    a1 a1 ω1 .. .. . . . . . . g0 + n ≤ j ≤ g.. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1065 that ϕ ag0 +1 . .. . . . where A. ωg0 +n−1 ω1 ag +n−1 0 ω1    .. ωg0 +n−1 . . ωg0 +n−1 ag +1 0 ag0 ω1    . bg +n−1 0 ω1  bg +n−1  0  .   ωg0 +n−1 (1)     B=        C=        D=    b1 b1 ag0 +n ω1 ag +1 0 . .    ωg0 +n . 1 ≤ j ≤ g0 + n − 1. the pullback of ωj . . . ωg ag0 ωg0 +n  .    ag +1 0 . ... . . .   ωg0+n−1 ω1 .. . . ..A. ϕ bg0 +n−1 are homologous to zero on C0 . ... . .     . ϕ bg0 +1 . . ωg0 +n−1 ω1 bg0 . . . .. .   ωg0+n−1     E=        F=        G=        H=    ag +1 0 a1 bg0 +n bg +1 0 bg +1 0 .... ωg0 +n−1 ω1 .    ag0 +n . . Let ω1 . .. ωg ωg0 +n . . . . ..  . . ωg0 +n−1 . . ..  .. ωg0 +n  ωg    . . The period matrix Ω of Jac(C ) is explicitly given by Ω = A G B H C I D J E K F L . bg ωg0 +n−1    .. ωg a1 .. . . . ϕ ag0 +n−1 .. bg . . bg0  . ag0 ag +n−1 0 .

1 ≤ j ≤ g0 + n − 1. =     bg +1 0 −   bg +1 0 . ωj =     bg +1 0 ωj . . . ωg0 +n  ωg    ..    bg0 +n . . .. g0 + n ≤ j ≤ g. 1 ≤ j ≤ g0 + n − 1.. . ωj ... . =     ag +1 0 −   ag +1 0 . g0 + n ≤ j ≤ g. ωg    . ωj . bg ωg0 +n  .. 1 ≤ j ≤ g0 + n − 1. 1 ≤ j ≤ g0 + n − 1... ωj . bg0 bg +1 0 ωg0 +n .1066 A. σ∗ (ωj ).. . g0 + n ≤ j ≤ g. ωj . . bg Notice that ag +1 0 ωj = − = − σ(ag0 +1 ) ag +1 0 ωj . σ ∗ ωj . . . ag +n−1 0 −    ag +n−1 0 bg +1 0 ωj = − = − σ (bg0 +1 ) bg +1 0 ωj . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079     I=        J=        K=    b1 ag0 +n ωg0 +n . g0 + n ≤ j ≤ g... bg +n−1 0 and     L=    bg0 +n ωg0 +n . . ωg .. .. ωg . ωg bg +n−1 0 . ωg bg0 ωg0 +n  . . . ωj . . ωg    . . ωj =      ag +n−1 0 ωj . .    b1 . . .. . ωg    . ωj ... ag ωg0 +n . bg +1 0 −   bg +1 0 . .    bg +1 0 .    ag0 +n . . ag ωg0 +n  .

. K = O. b1 = −      b1 ωj . ag0 b1 . 1 ≤ j ≤ g0 + n − 1. = A G 2G 2A O C2 D J 2J 2D .. g0 + n ≤ j ≤ g. =    a1 ωj . g0 + n ≤ j ≤ g. ωg . .. . bg0 ωg0 +n  . . Then. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1067 ag0 +n ωj = = σ(a1 ) a 1 −   ωj . σ ∗ ωj . g0 + n ≤ j ≤ g. H = O. 1 ≤ j ≤ g0 + n − 1. ωj =  −      bg0 bg0 ωj . . bg ωj . L = J. . .. where O is the null matrix and therefore Ω = A G B O C2 −A G C3 D J C4 E O C5 −D J C6 . 1 ≤ j ≤ g0 + n − 1.. . ωg ωg0 +n . g0 + n ≤ j ≤ g. 0 ag0 +n ωj . σ ∗ ωj .    (2) a1 . = A G .. a1 ωj . ag0 a1 ωg0 +n . = = where ∆ = G     =    2A O B O O 2D O O 2J .A. we obtain the following matrices: Ω1 = C 1 C2 B O C2 B O C1 + C3 O C4 E O C4 E O O O C5 C4 + C6 . ωg    . C = −A. I = G. C5 C4 − C6 Ω2 = C 1 C1 − C3 D J .. Γ O 2∆ . ωj =    −      ag ag ωj . . ≡ C1 By elementary column operation. 0 and bg0 +n ωj = = σ(b1 ) b 1 ωj . 1 ≤ j ≤ g0 + n − 1. . J . C5 C1 + C3 C4 + C6 Ω3 = C 1 − C 3 C4 − C6 E O O 2G . . . . . ωj . ωg b1 ωg0 +n . bg0 . . F = −D. ...

 mod 2L∆  . .. the tori Cg0 +n−1 /LΓ and Cg0 /2L∆ can be embedded into Cg /LΩ and the map Cg /LΩ3 = Cg0 +n−1 /LΓ ⊕ Cg0 /2L∆ −→ Cg /LΩ . .    ω1   a =  g0 +n−1   2 ω1  b1   .  .   .  tg0 t1  are injectives.    t  → Cg /LΩ :  g0 +n−1  mod LΩ . ω1 . .. it follows that the maps    .   ωg0 +n−1  2  bg0   ωg0 +n−1   bg +1  0  .  .. ωg0 +n−1 . ni ∈ Z . .  . .  Cg0 /2L∆ :  .    . .   . Ω2 .  .  . Since LΩ = LΩ1 = LΩ2 = LΩ3 .  Cg0 +n−1 /LΓ :  . .  ..  .  : mi . Therefore.  .   .   bi   ωg    be the period lattice associated to Ω .   . Let us denote also by LΩ1 .  .  mod LΩ .  mod LΩ →  . .  .  .  ωg0 +n−1  2  b1   .     t1 t1 .   bg +n−1 0 (3) ωg0 +n−1 Let     g LΩ = mi ai   i=1  ω1  . × ..   ωg0 +n−1    ag +n−1 0  .  . LΩ3 and L∆ the period lattices associated respectively to Ω1 .        2  ag0   ωg0 +n−1    ag +1 0   .    2 ω1  ag0    ω1   ag0 +1   ..    0 g → C /LΩ :   mod LΩ .   .   . t1  0  .    2 ω1  bg0    ω1   bg0 +1  . .   .  ...  .. ..   ..  t1     .  tg0 +n−1 t1   .1068 A.  mod LΓ  .  . 2 a1 2 ω1  a1  . Ω3 and ∆.    ωg ω1     .  + ni  .. LΩ2 . ωg0 +n−1  ..  0     . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 and Γ = 2A  B 2D E . tg0 0   . 3 ..   bg +n−1 0 ... tg tg .

  ωg . Moreover. . the dual Jac (C ) of Jac(C ) is isomorphic to Jac(C ). . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1069 shows that the jacobian variety Jac (C0 ) intersects the torus Cg0 +n−1 /LΓ in 22g0 points.      D  ωg0 +n−1   0 .   0   D 2 ωg 0          ⇐⇒          0 D ω1      t1    . u0 : z0 → divisor class(z0 − p0 ).  0        D   . Z)− . Prym(C /C0 ) is thus a subabelian variety of Jac (C ). . ΩC − ∗ /H1 (C .  .    D  . ωg Consequently.  D 0 . we have Prym(C /C0 ) = (Ker Nϕ )0 . .  . D  ωg0 +n    0  .e. The norm map Nϕ is ∗ surjective. 2  ∈ L∆ .  ωg0 +n−1  t   g +n−1   0 = 0 D   0 .    .   0   . (ϕ∗ OC )∗ ) ↓ Nϕ Jac(C ) Jac(C0 ) → ∗ H1 (C0 . OC ) H1 (C0 . i. More precisely. ωg maps Prym(C /C0 ) bi-analtically onto C g0 +n−1 /LΓ . = Im(1Jac(C ) − σ) ⊂ Jac(C ). We have then the following diagram C ↓σ C ↓ϕ C0 −→ −→ u0 u Ker Nϕ ↓ Jac(C ) ←− ←− ⊂ −→ Nϕ u∗ 0 u∗ Jac∗ (C ) ∗ ↑ Nϕ ↓ Nϕ Jac(C0 ) ↓ ϕ∗ Jac(C ) ↓σ Jac∗ (C0 ) ∗ H1 (C . . = Ker(1Jac(C ) + σ)0 . Let D ∈ Prym(C /C0 ).  . . fixed.  .. Nϕ : Jac(C ) → Jac(C0 ). The Prym variety denoted Prym(C /C0 ) is defined by 1 Prym(C /C0 ) = H0 C . where − denote the −1 eigenspace for a vector space on which the involution σ acts.  .A. OC0 ). ωg0 +n  ∈ LΩ ⇐⇒  . mi pi → mi ϕ(pi ) is the norm mapping and Jac∗ (C ) = dual of Jac(C ). . 0  D 0 . p0 ∈ C0 . By Chow’s theorem [9]. fixed with p0 = ϕ(p).  0  .    ωg0 +n   . .    .     ωg0 +n  0    0     D     . p ∈ C . the Abel–Jacobi map          D −→          0 D ω1  Jac (C ) −→ C g /LΩ .  D   . where u : z → divisor class(z − p).

. δ δ and Γ ∗ = (U ∗ .     ωg0 +n−1 ωg0 +n−1 . .   ωg0 +n−1 Then. . . pour g0 + 1 ≤ j ≤ g0 + n − 1. . 2 bg0 ω1 2 b1 . the period matrix Γ takes the canonical form (∆δ . dim Prym(C /C0 ) = g − g0 = g0 + n − 1. . . . ω1    ag +n−1 0 ω1          . ωg0 +n−1 . . ω1   a1 ag0 ag +1 ag +n−1   0 0   . . Z)− . . . symmetric and Im > 0.  . 2 ωg0 +n−1 2 a1 ag0 ag +1 0 ag +n−1 0     V =    2 b1 ω1 . in the new basis λ1 . . Write Γ = (U. Let ϕ : C −→ C0 be a double covering where C and C0 are nonsingular algebraic curves with jacobians Jac(C ) and Jac(C0 ).e. .  . Observe that ∆ (2) (resp.1070 A. . .. U=  . the involution σ induces an involution σ : Jac(C ) −→ Jac(C ). 2 ω1 ω1 . δg0 +n−1 U −1 V ∆−1 ). and up to some points of order two.     ωg0 +n−1 2 ag0  . δg0 +n−1 ∆−1 Z ∆−1 .  ..  . This involution extends by linearity to a map (which will again be denoted by σ ) σ : Jac(C ) −→ Jac(C ) and up some points of order two. . ΩC − ∗ /H1 (C . class of D −→ class of σ D.. δn ) and Z = ∆δ U −1 V. eg0 =   . . Jac(C0 ) and an odd part (called Prym variety) denoted Prym(C /C0 ) and defined by 1 Prym(C /C0 ) = H0 C . δ δ δ = (δg0 +n−1 ∆−1 . .. . . The above discussion is summed up in the following statement: Theorem 1. with ∆δ = diag(δ1 .. 2 bg0 . . Z ). ωg0 +n−1 bg +1 0 ω1 . . Jac(C ) splits into an even part Jac (C0 ) and an odd part Prym(C /C0 ): Jac(C ) = Prym(C /C0 ) ⊕ Jac (C0 ) . . with   2 ω1 . δg0 +n−1 ∆−1 (U ∗ )−1 V ∗ . Let σ : C −→ C be the involution exchanging sheets of C over C0 = C /σ . .. .   ωg0 +n−1  eg0 +1    =    ag +1 0 .   ωg0 +n−1 and let us call     e1 =    2 2 a1 ω1       2 ag0 ω1    a1   . . δj = 1 2 pour 1 ≤ j ≤ g0 .  .. .. . with dim Jac(C0 ) = g0 . dim Jac(C ) = g = 2g0 + n − 1. ωg0 +n−1 . . . . λg0 +n−1 where λj = ej δj .. Jac(C ) splits into an even part. . is the period matrix of the dual abelian variety Prym∗ (C /C0 ). bg +n−1 0  . . . δ δ = δg0 +n−1 ∆−1 . . Then Γ ∗ = δg0 +n−1 ∆−1 . . V ∗ ) = (A B D E).. . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 Equivalently. V ). . .i.. bg +1 0 . . Γ (3)) is the period matrix of Jac (C0 ) (resp. eg +n−1 =    0     ag +1 0 ag +n−1 0  .. ωg0 +n−1 bg +n−1 0 ω1    . . . . . Prym(C /C0 )).. .

.D − D KT ). It is known that every positive divisor D on an irreducible abelian variety is ample and thus some multiple of D embeds M into PN .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 ). Moreover. (Kodaira–Serre duality). Γ = (2A B 2D E) and Γ ∗ = (A B D E). if Ω = A G B H C I D J E K F L . . as well as the basic techniques to study two-dimensional algebraic completely integrable systems. any k ≥ 3 will work. 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. Let T = Cn /Λ be a n-dimensional abelian variety where Λ is the lattice generated by the 2n columns λ1 . Im Z > 0. ∗ = dim L(D ) − dim H1 (T. δj ∈ N∗ and δj |δj+1 . we have Jac(C ) = Prym(C /C0 ) ⊕ Jac (C0 ) .. p → [1 : f1 (p) : .D χ(D ) = pa (T ) + 1 + (D . . KT is trivial and pa (T ) = −1. (5) 2 where KT is the canonical divisor on T . . By a theorem of Lefschetz. ∂p1 ∂p n ˙ p1 = − ∂H ∂H ˙ . . ∂q 1 ∂q n (9) . (4) In the case of a 2-dimensional abelian varieties (surfaces). . OT (D )) − dim H1 (T. therefore combining relations (4)–(7). . To be precise. However for abelian surfaces.1.e.A. The pair (M.e. .D ∂H ∂H ˙ . (Kodaira vanishing theorem). Algebraic complete integrability 3. there exists a complex basis of Cn such that the lattice expressed in that basis is generated by the columns of the n × 2n period matrix   δ1 0 |   . . (8) 2 Let M be a 2n-dimensional differentiable manifold and ω a closed non-degenerate differential 2-form. are respectively ∆ = (G H). . . ω) with hamiltonian H can be written in the form ˙ q1 = χ (D ) = dim L(D ) = D . Define L(D ) = {f meromorphic on T : (f ) ≥ −D}. with 2n branch points. dim Jac (C ) = genus g of C = 2g0 + n − 1 and dim Prym(C /C0 ) = g − g0 = g0 + n + 1. D . OT (D)). then the period matrices of Jac (C0 ) . Let H : M → R be a smooth function. we obtain = g (D ) − 1 = δ1 δ2 .  | Z . = dim L(D ). The divisor D is called ample when a basis (f0 . where as the Riemann–Roch theorem for line bundles on a surface tells you that (6) 2 where pa (T ) is the arithmetic genus of T and χ(D ) the Euler characteristic of D . . . . with dim Jac (C0 ) = genus g0 of C0 . . the zero-locus of a holomorphic 2-form. Prym (C /C0 ) and the dual Prym∗ (C /C0 ) of Prym (C /C0 ).. Moreover. : fN (p)]. A hamiltonian system on (M. . The integers δj which provide the so-called polarization of the abelian variety T are then related to the divisor as follows: dim L(D ) = δ1 . .D + D . . To study abelian surfaces using Riemann surfaces on these surfaces. pn = − . 1 χ(D ) = dim H0 (T. λ2n of the n × 2n period matrix Ω and let D be a divisor on T . fN ) of L(kD ) embeds T smoothly into PN for some k. . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1071 1 where ΩC is the sheaf of holomorphic 1-forms on C and − denote the −1 eigenspace for a vector space on which j acts. for D = kj Dj a function f ∈ L(D ) has at worst a kj -fold pole along Dj . is the period matrix of Jac(C ) where A. then kD is called very ample. (7) ∗ whenever D ⊗ KT defines a positive line bundle. δn . i. 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 + 1. qn = . L denote the matrices (1). .. ω) is called a symplectic manifold. we recall that g(D ) = KT .   .. via the map T → PN . . 3. i. 0 δn | with Z = Z . . Ω 2 (D ⊗ KT )). .

. . . if {F. that is. H} of F and H is defined by k=1 ∂p ∂q − ∂q ∂p k k k k = n XH F = k=1 ∂H ∂F ∂H ∂F − = {F. The system (11) with polynomial right x hand side will be called algebraically completely integrable (a. there exists a diffeomorphism n+k i=1 x ∈ Rm : Hi = ci → Rn /Lattice. and since XHk Hl = {Hk . . and every ˙ vector field XHk can be written in the form ϕk = hk (s1 . . H2 . abelian variety) and D a divisor. . . where Cn /Lattice is a complex algebraic torus (i.. 1 ≤ i ≤ k. ∂x ∂x ∂xk ∂xl k . . . .1072 A. . . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 where (q1 . it is diffeomorphic to an n-dimensional real torus and each vector field will define a linear flow there. x ∈ Cm }. . the integral curves of each XHk will lie in Mc and the vector fields XHk span the tangent space of Mc . ϕk = hk (s1 . . thought of as affine varieties n+k A= i=1 {Hi = ci .. in Cm can be completed into complex algebraic tori. A hamiltonian system is completely integrable in the sense of Liouville if there exist n invariants H1 = H. Hl } = 0. The functions ϕk (called angle-variables) give standard angular coordinates on the torus. in particular. . . satisfies the Jacobi identities. ∂x x = (q1 . (9) and (10) can be written in more compact form ˙ x=J ∂H . and can be solved by quadratures. .. where ω1 . XN ) = 0} involving a large i= number of homogeneous polynomials Pi . Remark 3. . .e..J = Jkl . ϕn in which ω takes the canonical form ω = k=1 dsk ∧ dϕk . for which the ∂H corresponding Poisson bracket {Hi . . Hn+k of which k lead to zero vector fields ∂H J ∂n+i (x) = 0. . . . there is an algebraic map (x1 (t). .15]. xm (t)) → (µ1 (t). {F. x +k the invariant varieties n=1 {x ∈ Rm : Hi = ci } are assumed compact and connected. . . . qn . ωn denote holomorphic differentials on some algebraic curves. . . J ∂xj .e. in a neighbourhood of the torus the hamiltonian vector field XHk take the following form ˙ s˙ = 0. ∂x x ∈ Rm . sn . Hi . Consequently. . cn ) the level set Mc = {H1 = c1 .e. . a skew-symmetric matrix where I is the n × n unit matrix and O the n × n zero matrix. k Consider now hamiltonian problems of the form ˙ XH : x = J ∂H ≡ f (x). ϕ1 . The functions F and H are said to be in involution or to commute. . . in some open neighbourhood of the torus one can introduce regular symplectic n coordinates s1 . its integral trajectories define a conditionally-periodic motion on the torus. i. Hj } = ∂∂Hi .1. Then. . Here the functions sk (called action-variables) give coordinates in the direction transverse to the torus and can be expressed functionally in terms of the first integrals Hk . pn ) . . .e. . . Note that Eqs. and the solutions of the system (11) are straight lines motions on these tori. according to the Arnold–Liouville i theorem. . . . Hj = 0) and m = 2n + k . . 1 ≤ j ≤ n. Algebraic means that the torus can be defined as an intersection M 1 {Pi (X0 . . . dH1 ∧ · · · ∧ dHn = 0). . . H} . If F is a smooth function on the manifold M. H} = O I ∂F ∂H ∂F ∂H . . . . Hl } all vanish) with linearly independent gradients (i. sn ). tn ) are meromorphic and (11) defines straight line motion on Cn /Lattice.l=1 n with J = −I O . . . (ii) The invariant varieties. To be precise. the functions xi = xi (t1 . . . . H} = 0. . . p1 . A ∪ D = Cn /Lattice. . . tn ) of Cn /Lattice coming from Cn . By a theorem of Arnold [4. qn . ∂pk ∂qk ∂qk ∂pk (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 XH is identically zero. the n remaining ones are in involution (i. Hn = cn } will be an n-manifold. . . pn ) are coordinates in M. . . Condition (i) means. . . . dj ∈ C. In the natural coordinates (t1 . . Thus the hamiltonian vector field XH is defined by XH n ∂H ∂ ∂H ∂ . such that the associated Poisson bracket {Hk . p1 . . (11) where H is the hamiltonian and J = J(x) is a skew-symmetric matrix with polynomial entries in x. sn ). For most values of ci ∈ R.e.c. Hn in involution (i. if Mc is compact and connected. .) in the sense of Adler–van Moerbeke [1] when: (i) The system possesses n + k independent polynomial invariants H1 . . For generic c = (c1 . . the Poisson bracket {F. . µn (t)) making the following sums linear in t: n i=1 µi (t) µi (0) ωj = d j t .i. . ..

2. xm ) . ∂q2 (15) H ≡ H1 = p2 + p2 + Aq2 + Bq2 + q2 q2 + 6q3 . x(k−1) for k > 1. . 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 Cn /Lattice and the complex flows generated by the vector fields are straight lines on this complex torus. this can never work for a general reason: An abelian variety A of dimension bigger or equal than two is never a complete intersection. An exposition of such methods and their applications can be found in [1.. definition. that is it can never be descriped in some projective space Pm (C) by m-dim A global polynomial homogeneous equations.e. going with each variable xi . Moreover through every point of D . ∂q 1 p2 = − ∂H . they must either come from the non-linear equations (13) or from the eigenvalue problem (14). Hence the differential equations must admit Laurent expansions which depend on the n − 1 parameters defining D and the n + k constants ci defining the torus Cn /Lattice . x(0) + gi xi(0) = 0.. So. L must have at least m − 1 integer eigenvalues. A must have a singularity somewhere along the locus at infinity A ∩ {X0 = 0}. the question is how does one prove directly that the system is effectively a. the affine variety A is 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 XH1 .. . . without ever going through the delicate procedure of blowing up and down. fi x 1 (0) . an abelian variety and the coordinates xi restricted to A are abelian functions.i. the total count is therefore m − 1 = dim (phase space) − 1 parameters. 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) Arnold–Liouville theorem discuted above. k ∈ Z.. because of the fact that the homogeneity k of the constant H must be an eigenvalue of L. . . ανm xm ) = ανi +1 fi (x1 . . (13) and at the kth step. Moreover the formal series solutions are convergent as a consequence of the majorant method [1]. . A naive guess would be to take the natural compactification A of A in Pm (C).c. These are much less conditions than expected. (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. i. Observe that then the constants of the motion H can be chosen to be (weight)-homogeneous: If the flow is algebraically completely integrable. . . . it means that the variables xi are meromorphic on the torus Cn /Lattice and by compactness they must blow up along a codimension one subvariety (a divisor) D ⊂ Cn /Lattice. Assume now hamiltonian flows to be (weight)-homogeneous with a weight νi ∈ N. the differential equations (11) must admits Laurent series solutions (12) depending on m−1 free parameters. A is an algebraic complex torus.A. . xm ) .2]. . So.i. some ki > 0. . ∂z If m − 1 free parameters are to appear in the Laurent series. . Now. ∂p1 1 2 q2 = ∂H . The Hénon–Heiles system The Hénon–Heiles system ˙ q1 = with ∂H .c. ανm xm ) = H (αν1 x1 . if A is to be the affine part of an abelian variety. m 1 ≤ i ≤ m. fi (αν1 x1 . 3. if the hamiltonian flow (11) is a. ∂p 2 ˙ p1 = − ∂H .i. Indeed. Namely. . In fact. . i. . . .c. 1 2 1 2 1 2 . 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. By the a. the flow (11) is a straight line motion in Cn /Lattice and thus it must hit the divisor D in at least one place. We must have ki = νi and coefficients in the series must satisfy at the 0th step non-linear equations. (j) ki ∈ Z.e.c. . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1073 Adler and van Moerbeke [1] have shown that the existence of a coherent set of Laurent solutions: ∞ xi = j=0 xi tj−ki .e. Information about the tori can then be gathered from the divisor. If this is possible. (14) L = Jacobian map of (13) = ∂f + gI |z=z(0) . . i. . . linear systems of equations: (L − kI) z(k) = where 0 for k = 1 some polynomial in x(1) . . ∀α ∈ C. αk H (x1 .i. In other words. . . . XHm extend holomorphically along the divisor D and remain independent there. . there is a straight line motion and therefore a Laurent expansion around that point of intersection.

where P8 (α) = − (16) 1 13 1 671 2 7 α8 − 5A − B α6 − B + 15552 432 18 36 15120 1 13 2 2 10 1 4A3 − B3 − A B + AB2 − c1 α2 + − 36 2520 6 9 7 17 7 1 36 A2 − c2 . the map σ : D −→ D . p2 = q2 and q1 q1 q1 (4) (5) (6) = αAB 24 − α5 72 + 11α3 B 720 Aβ − 11α3 A 120 − αB2 720 − αA2 8 . where the lattice is generated by the period matrix Im a c c b 2 0 0 4 a c c b . by adjoining the divisor D . (β. ρ : D −→ E . according to (8). The divisor 2D is very ample and the functions 1. (β. Hi (x) = ci . p1 . (19) ramified at the four points covering ζ = 0 and ∞. the affine surface Mc completes into an abelian surface Mc . Moreover.1074 A. D = t0 − coefficient of Mc .2]. Following the methods in Adler–van Moerbeke [1. 4). α) −→ (β. α5 A 2160 = − αγ 9 α7 15552 − + α5 B 12960 + α3 B2 25920 + α3 A2 1440 − α3 AB 4320 + αAB2 1440 − αB3 19440 − αA2 B 288 + αA3 144 . p2 are canonical coordinates and momenta. y1 . The latter defines on Mc a polarization (1. 943 1260 BA α4 The curve D determined by an eight-order equation is smooth. γ . It provides a model for the oscillations of atoms in a three-atomic molecule. αA 2 B − + αB 12 t + βt2 + q1 t3 + q1 t4 + q1 t5 + · · · . The system (15) possesses Laurent series solutions depending on 3 free parameters α. y2 . embed Mc smoothly into 1 1 1 2 CP7 with polarization (2. where p1 = q1 . i. to be precise D is the closure of the continuous components of the set of Laurent series solutions x (t) such that Hi (x (t)) = ci . Let D be the pole solutions restricted to the surface 2 Mc = i=1 x ≡ (q1 . + α2 12 α4 48 12 + α2 A 10 − α2 B 60 − B2 240 t2 + αβ 3 t3 + γ t4 + · · · . > 0.e. 2). 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. The hyperelliptic curve D is thus a 2-sheeted ramified covering of the elliptic curve E (18). p2 ) ∈ C4 . The abelian surface Mc which completes the affine surface Mc is the dual Prym variety Prym∗ (D /E ) of the genus 3 hyperelliptic curve D (16) for the involution σ interchanging the sheets of the double covering ρ (19) and the problem linearizes on this variety. (4) (5) (6) q2 = − . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 has another constant of motion H2 = q4 + 4q2 q2 − 4p1 (p1 q2 − p2 q1 ) + 4Aq2 q2 + (4A − B) p2 + Aq2 1 1 2 1 1 1 . −α). y2 . x2 + y2 y2 . x1 x2 + 2Ay1 y2 + 2y1 y2 . Thus we find an algebraic curve defined by D : β2 = P8 (α). (17) is an involution on D and the quotient E = D /σ is an elliptic curve defined by (18) where P4 (ζ) is the degree 4 polynomial in ζ = α2 obtained from (16). namely q1 = α t + 1 t2 α3 12 + − . It well-known from applications in stellar dynamics. =− βα2 12 + − βB 60 − 10 . B are constant parameters and q1 . 1 ≤ i ≤ 2. where A. Theorem 2. hyperelliptic and its genus is 3. E : β2 = P4 (ζ). q2 . respectively. β. x1 . x2 y1 − 2x1 y2 . Then the system (15) is algebraic complete integrable and the corresponding flow evolves on an abelian surface Mc = C2 /lattice. statistical mechanics and quantum mechanics. p1 . q2 . α) −→ (β.. . ζ).

Let us denote also by L∆ . Let (a1 . ω3 = αdα . σ (b1 ) = b3 .  Let ω1 ω2  + ni mi LΩ =  ai i=1 ω3   3   ω1  ω2  : mi . We consider the period matrix Ω of Jac(D)      Ω =   a1 a1 a1 a2 a2 a2 ω1 ω2 ω3 a3 a3 a3 ω1 ω2 ω3 b1 b1 b1 ω1 ω2 ω3 b2 b2 b2 b3 b3 b3 ω2  .2). E ). H2 such that: dt1 |D = ω1 and dt2 |D = ω2 . has kernel (ϕ∗ E ) Prym(D /E ) is of type (1. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1075 Proof. ω2 on the curve D (16) we take the differentials ω1 = α2 dα . the period lattice associated ∆.e. a2 . β ω1 ω2 ω3 ω2 = dα β . σ (b2 ) = −b2 . b2 . As a basis of holomorphic differentials ω0 . ni ∈ Z . for the involution σ (17)..  0 ω3 0 ω3 and therefore the period matrices of Jac(E ) (i. nk  k LΛ =     k=1  dt2   νk . a3 .  ω3     By Theorem 1. Let Mc → C2 /LΛ : p p p0 Z2 × Z2 and the induced polarization on dt1 dt2 . β ω1 ω2 ω3 ω1 and obviously σ ∗ (ω1 ) = −ω1 .  bi ω3    be the period lattice associated to Ω . σ (a2 ) = −a2 . where dt1 .σ ∗ (ω3 ) = ω3 . 1 Γ =   ω2 ω2 2 ω2 ω2 2 a1 a2 b1 b2 a1 ω3 b1 ω3 ). . σ ∗ (ω2 ) = −ω2 . and   Γ∗ =   a1 a1 ω1 ω2 a2 a2 ω1 ω2 b1 b1 ω1 ω2 b2 b2 ω1  ω2  . dt2 are two differentials on Mc corresponding to the flows generated respectively by H1 . we get the following diagram: 0 ↓ E ↓ ϕ∗ 0 D ↓ϕ −→ Nϕ −→ ker Nϕ −→ τ Prym(D /E ) ⊕ E = Jac(D ) E −→ 0 ↓ Mc = Mc ∪ 2D ↓ 0 C2 /lattice The polarization map τ : Prym(D /E ) −→ Mc = Prym∗ (D /E ). Prym(D /E ) and Prym∗ (D /E ) are respectively ∆ = (   ω1  ω1 2 ω1 2 a ω1 b2 b1 a2 . b1 .        4     ν dt1   : nk ∈ Z . ω1 . According to Theorem 1. b3 ) be a canonical homology basis of D such that σ (a1 ) = a3 .A. be the uniformizing map.     Ω =   a1 a1 a1 ω1 ω2 ω3 a2 a2 ω1 ω2 − a1 ω1 ω2 b1 b1 b1 ω1 ω2 ω3 b2 b2 ω1 ω2 − b1 ω1      − a1 a1 − b1 b1 ω2  .

 is a Riemann matrix. The Kowalewski rigid body motion The motion for the Kowalewski’s top is governed by the equations m = m ∧ λm + γ ∧ l. H3 = H4 = 2 2 2 γ1 + γ2 + γ3 = c3 = 1. γ and l denote respectively the angular momentum. 1 t 2t2 . 1 t 2t2    iα1  i   − α2 α2 + ◦ (t) .i. The system (20) can be written m1 = m2 m3 . 0) and λm = (m1 /2. ν4 in D which we look for are a1 . ν4 ) is a basis of H1 (Mc . the period matrix of Prym∗ (D /E ) dual of Prym(D /E ). + ◦ (t) . Z) such that    Λ=  a1 a1 ω1 ω2 b1 b1 ω1 ω2 a2 a2 ω1 ω2 b2 b2 ω1  ω2 . (20) where m. γ = γ ∧ λm. Consequently Mc and Prym∗ (D /E ) are two abelian varieties analytically isomorphic to the same complex torus C2 /LΛ . . m2 = −m1 m3 + 2γ3 . Z).1076 A. We show that Λ = Γ ∗ . ν2 . (22) m1 − im2 2 m1 + im2 2 2 − (γ1 + iγ2 ) 2 − (γ1 − iγ2 ) = c4 . a2 . 3 H2 = m1 γ1 + m2 γ2 + m3 γ3 = c2 .e. . Z) −→ H1 (Mc . . b1 . 0.  dt2 and (ν1 .  and       : nk ∈ Z . m2 /2. Ac and Prym∗ (D /E ) are then algebraically isomorphic.. ν3 . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 is the lattice associated to the period matrix   dt1  Λ =  ν1  dt2 ν1 ν2 ν2 dt1 ν4 ν3 dt1 ν4 ν4 dt2 dt2 dt1  . 1 t γ1 (t) = 2t m1 (t) = α  1  1    − i α2 − 2 α2 + ◦ (t) . ν2 . . . m3 /2).3.   2 + ◦ (t) . . Z) induced by the inclusion D → Mc is surjective and consequently we can find 4 cycles ν1 . the map H1 (D . the directional cosine of the z-axis (fixed in space). By the Lefschetz theorem on hyperplane section [9]. nk  νk LΛ =     k=1  ω2    4     ω1   νk The cycles ν1 . The system (21) admits two distinct families of Laurent series solutions:   α1  1   + i α2 − 2 α2 + ◦ (t) . ν3 .   2 + ◦ (t) . By Chow’s theorem. ν4 on the curve D such that   ω1  Λ =  ν1  ν1 ν2 ν2 ω1 ν4 ν3 ω1 ω2 ν4 ν4 ω1  ω2 ω2 ω2 . γ 2 = m1 γ3 − 2m3 γ1 . . . ν2 . 1 t m2 (t) = γ2 (t) = 2t  −iα1  −i     − α2 α2 + ◦ (t) . γ 1 = 2 m3 γ2 − m2 γ3 . the center of gravity which after some rescaling and normalization may be taken as l = (1. ν3 . 3. + ◦ (t) . m3 = −2γ2 . b2 and they generate H1 (Mc . (21) with constants of motion 1 2 2 H1 = 2 m1 + m2 + m2 + 2γ1 = c1 . γ 3 = m2 γ1 − m1 γ2 .

accordind to (8). H4 = 1 2 b1 p2 + b2 p2 + b3 p2 + 1 2 3 l2 + l2 + l2 1 2 3 . These equations can be regarded as the equations of the geodesics of the right-invariant metric on the group E (3) = SO (3) × R3 of motions of 3-dimensional euclidean space R3 . . 4). l2 .1). the Hamiltonian flows generated by the vector fields XH1 and XH4 are straight lines on Mc . . The affine surface Mc = 4=1 {Hi = ci } ⊂ C6 . . The tori Mc can be identified as Mc = Prym∗ (Dε /D0 ). ∂x ∂y where x = (x1 . of geometric genus 9. f1 (t1 . we obtain the following theorem: Theorem 3. l ˙2 l ˙3 l ∂p 3 ∂p 2 ∂l3 ∂l2 ∂H ∂H ∂H ∂H = p3 − p1 + l3 − l1 . (25) have the four invariants: H1 = H = 1 2 a1 p2 + a2 p2 + a3 p2 + b1 l2 + b2 l2 + b3 l2 1 2 3 1 2 3 . . (l1 . ∂p 2 ∂p 1 ∂l2 ∂l1 (25) where (p1 . Kirchhoff’s equations of motion of a solid in an ideal fluid The Kirchhoff’s equations of motion of a solid in an ideal fluid have the form ˙ p1 = p2 ∂H ∂H − p3 . f6 = 4f1 f4 − f3 f5 . generated by rotations and translations. Each of the curve Dε is a 2 − 1 ramified cover (α1 . The 8 functions 1 2 2 . defined by putting the four invariants (22) of the i Kowalewski flow (21) equal to generic constants. f2 = m2 . In Clebsch’s case. is very ample and defines a polarization (2. p and p. then D is a set of two isomorphic curves of genus 3. It was shown by the author [14] that each divisor Dε is ample and defines a polarization (1. f1 = m1 . (26) H2 = p 2 + p 2 + p 2 . ∂l1 ∂l3 ∂H ∂H ˙ − p2 . D = D1 + D−1 : Dε : P (α1 . Eqs.. l . i. 7 (t1 . t2 ) → [(1. γi blow up. f4 = γ3 . α2 ) = α2 − 1 1 α2 − 1 α2 − P (α2 ) + c4 = 0. p/ε) and letting ε → 0. The first set can be obtained from the second by putting (x. ∂l3 ∂l2 ∂H ∂H ˙ p2 = p3 − p1 . t2 ))] . whereas the divisor D. . Hence the motion has the trivial coadjoint orbit invariants p. α5.4.  m3 (t) = t  −i   + α1 α2 + ◦ (t) . 1 2 (23) where P (α2 ) = c1 α2 − 2εc2 α2 − 1 and ε = ±1. 2 (24) ramified at the 4 points α1 = 0 covering the 4 roots of P (α2 ) = 0. y2 . t2 ) . x3 ) ∈ R3 et y = (y1 . ∂p 1 ∂p 3 ∂l1 ∂l3 ∂H ∂H ∂H ∂H = p1 − p2 + l1 − l2 . . 3. 1 2 3 H3 = p1 l1 + p2 l2 + p3 l3 . x2 .e. Following the method (Section 3. l3 ) 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. β) of elliptic curves D0 : ε 2 D0 : ε β2 = P2 (α2 ) − 4c4 α4 . which is nothing but the equation of the divisor D along which the mi . As it turns out. one eliminates three parameters linearly. is an embedding of Mc into CP . Since the system (21) admits two families of Laurent solutions. f7 = (f2 γ1 − f1 γ2 ) f3 + 2f4 γ2 . The latter set of equations is the geodesic flow on SO(4) for a left invariant metric defined by the quadratic form H. t  α2  + ◦ (t) . dual of Prym(Dε /D0 ) and the problem linearizes on this ε ε Prym variety. 2). p3 = p1 ∂l2 ∂l1 ˙1 = p2 ∂H − p3 ∂H + l2 ∂H − l3 ∂H . ∂x ∂y ˙ y=y∧ ∂H ∂H +x∧ . By substituting these series in the constants of the motion Hi (22). f3 = m3 .A. y) = (l.  t γ3 (t) = α  2  + ◦ (t) . form a basis of the vector space of meromorphic functions on Mc with at worst a simple pole along D Moreover. α2 . . Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1077  i  + α1 α2 + ◦ (t) . y3 ) ∈ R3 . p2 . . Each Dε is a double cover of an elliptic curve D0 (24) ramified at 4 points. this is a special case of a more general system of equations written as ˙ x=x∧ ∂H ∂H +y∧ . p3 ) is the velocity of a point fixed relatively to the solid. leading to algebraic relation between the two remaining parameters. t which depend on 5 free parameters α1 . the map Mc C2 /Lattice → CP7 . f5 = f1 + f2 . ε Moreover. is the affine part of an abelian surface Mc with Mc \ Mc = D = one genus 9 curve consisting of two genus 3 curves Dε (23) intersecting in 4 points. . f7 (t1 .

(28) which depend on dim(phase space) − 1 = at the 0th step a non-linear equation. solving inductively for the x(k) . d2 such that: d1 + d2 + 1 = 0. p3 . 4.1078 A. b2 . ω1 = ζ dζ . The curve D is a 2-sheeted ramified covering of the elliptic curve E .e. i. x = (p1 . a1 −a2 The system ˙ x = f (x) ≡ J where ∂H . p2 .  (27)  J= O P P L 0 −p 3 0 p1 P =  p3 −p2  p2 −p1  . c2 . Putting (28) into (27). (29) = an algebraic curve defined by θ2 + c1 β2 γ 2 + c2 α2 γ 2 + c3 α2 β2 + c4 αβγ = 0. a3 . Upon putting ζ ≡ α2 . 2 0 0 4 a c c b form a basis of holomorphic differentials on C and τ ∗ (ω0 ) = ω0 . Haine [10] shows that the flow evolves on an abelian surface Mc ⊆ CP7 of period matrix Prym(C /C0 ). Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 − − − with a2b a3 + a3b a1 + a1b a2 = 0. l1 . The abelian surface Mc can be identified as Prym(C /C0 ). (θ. Taking into account only solutions trajectories lying on the invariant surface Mc = 4 1 {Hi (x) = ci } ⊂ C6 .. 0 −l3 0 p1 0 L =  l3 −l2  l2 −l1  . i. x(0) + f (x(0) ) = 0. More precisely 4 i=1 x ∈ C6 . and at the kth step. ζ). the curve D can also be seen as a 4-sheeted unramified covering of the following curve of genus 3: C: E : β2 = d2 α2 − 1. c3 . γ = x6 parameterizes the elliptic curve (0) (0) (0) γ 2 = d2 α2 + 1. a linear system of equations. . l2 . . the genus of D is 9. . is an involution on C and the quotient C0 = C /τ is an elliptic curve defined by C0 : 2 η2 = c4 ζ d2 d2 ζ2 + d2 − d2 ζ − 1 . related to specific values of c1 . ζ) → (η. Using the Poincaré residue map [9]. which is a ramified cover of an elliptic curve E (30) with 16 branch points. 1 θ2 + c1 β2 γ 2 + c2 γ 2 + c3 β2 ζ 2 2 − c4 ζβ2 γ 2 = 0. c4 . To be precise we search for the set of Laurent solutions which remain confined to a fixed affine invariant surface. . b1 . ζ) → (−θ. where θ is an arbitrary parameter and where α = x4 . 4 D = i=1 t0 − coefficient of Hi (x (t)) = ci . . one finds (L − kI)x(k) = 0 for k = 1 quadratic polynomial in x(1) . (θ. 2). where D is a genus 9 curve (29). . τ ∗ (ωk ) = −ωk (k = 1. 4 1 2 1 2 Let (a1 . a2 . 1 2 1 2 The curve C is a double ramified covering of C0 . 0 Consider points at infinity which are limit points of trajectories of the flow.  θ2 = −c1 β2 γ 2 − c2 γ 2 + c3 β2 ζ + η C: η2 = c2 ζ d2 d2 ζ2 + d2 − d2 ζ − 1 . One parameter appear at the 0th step. C → C0 . . b3 ) be a canonical homology basis of C such that τ (a1 ) = a3 . The curve D is unramified at infinity and by Hurwitz’s formula. 5 free parameters. Im a c c b > 0 and also identified Mc as Prym variety Theorem 4. Moreover. in the resolution of the non-linear equation and the 4 remaining ones at the kth step. to be precise = b1 (b2 −b3 ) a2 −a3 = b2 (b3 −b1 ) a3 −a1 = b3 (b1 −b2 ) . l3 ) . x (t) = t−1 x(0) + x(1) t + x(2) t2 + · · · . . we obtain one-parameter families which i= are parameterized by a curve. k = 1. (30) 2 2 2 with d1 . In fact. τ (a2 ) = −a2 and τ (b2 ) = −b2 for the involution τ . x(k) for k ≥ 1.e. ζ). there is a Laurent decomposition of such asymptotic solutions. the map τ : C → C . ∂x . . η. and the constant satisfies the conditions 1 2 3 (25) can be written in the form (11) with m = 6. we show that ω0 = dζ η . where L denotes the jacobian map of the non-linear equation above. θη ω2 = dζ θη . β = x5 . .. τ (b1 ) = b3 . The branch points are defined by the 16 zeroes of c1 β2 γ 2 + c2 α2 γ 2 + c3 α2 β2 + c4 αβγ on E . Hi (x) = ci = Prym(C /C0 ) \ D.

Enol’skii. 2)-polarization. 31 (1999) 265–286. 263 (1983) 435–472. [7] A. Geom. [13] S. Le théorème d’Arnold–Liouville et ses conséquences. 398 (1989) 169–179. Invent. 4 21 (1988) 193–223. Adler. Arbarello. 89–111. [15] A. 41 (1977) 149–196. Sendai. [6] A. Adv. 1978. . Painlevé geometry and Lie algebras. Griffiths. Its. M.V. The complex geometry of the Kowalewski–Painlevé analysis. Algebraic integrability. [12] F. Lesfari. [18] D. Ann.Z. Math. Math. Springer-Verlag. Lesfari.D. Abelian surfaces with (1. Harris. Math. Barth. Pure Math. [8] E. Springer-Verlag. 12 (1989) 177–232. Vanhaecke. Spectral curves and the generalized theta divisor. A. Sup. Novikov. Sur le problème de la rotation d’un corps solide autour d’un point fixe. P. Geodesic flow on SO(4) and Abelian surfaces. J. Wiley-Interscience. Maskit. Belokolos. L. Academic Press. Prym varieties I. V. vol. V. Stud.S. Niremberg (Eds. Ahlfors. Phys. B. 2004.. 325–350. Geom. Completely integrable systems: Jacobi’s heritage. A. 54 (1987) 91–114. [19] S. [17] A. New-York. Griffiths. J. Arnold. École Norm. Uber die Bewegung eines festen Körpers in einer Flüssigkeit I. Pacific J.I. [2] M. Math. 97 (1989) 3–51. 58 (1) (2003) 6–20. World Scientific. Scient. [3] E. Elem. Duke Math. J. Contributions to Analysis. in: L. Beauville. Lesfari.I. 47.A. Springer-Verlag. Paris sér. Harris. 1978. pp. Prym varieties and the Schottky problem. Two-dimensional Schrödinger operator and solitons: Three-dimensional integrable systems. Haine. 1974. [11] N. pp. Matveev. Hitchin. Stable bundles and integrable sytems. Invent. 1985. J. I. Kötter. in: Proceedings of the VIIIth International Congress on Mathematical Physics. Springer-Verlag. Acta Math. Abelian surfaces and Kowalewski’s top. 109 (1892) 51–81. Beauville. van Moerbeke. Reine Angew. Ramanan. J. Ann. II. Kra. Narasimhan. Math. Le système différentiel de Hénon–Heiles et les variétés Prym. P. [5] W. [9] P. P. Mumford.A. Algebro-Geometric Approach to Nonlinear Integrable Equations.P.A. Cornalba. Adler. J. Geometry of Algebraic Curves I. on Alg.).. Lesfari / Journal of Geometry and Physics 58 (2008) 1063–1079 1079 References [1] M. Kowalewski. Principles of Algebraic Geometry. [10] L. [16] A. P. Mathematical Methods in Classical Mechanics. 10 (1987) 41–84. 212 (1) (2003) 125–132. in: A Series of Modern Surveys in Mathematics. 1987. Lesfari. S. [4] V. 1994. [14] A. in: Conf. Reine Angew. 226–241. 1994. Math. Bobenko. M.B. Math. van Moerbeke.R.

Sign up to vote on this title
UsefulNot useful