Professional Documents
Culture Documents
Pub Symmetries of Compact Riemann Surfaces
Pub Symmetries of Compact Riemann Surfaces
Editors:
J.-M. Morel, Cachan
F. Takens, Groningen
B. Teissier, Paris
Emilio Bujalance · Francisco Javier Cirre
José Manuel Gamboa · Grzegorz Gromadzki
Symmetries of Compact
Riemann Surfaces
123
Emilio Bujalance José Manuel Gamboa
Facultad de Ciencias, UNED Universidad Complutense Madrid
Matemáticas Fundamentales Facultad de Matemáticas, UCM
C/ Senda del Rey 9 Departamento de Álgebra
28040 Madrid Plaza de las Ciencias 3
Spain 28040 Madrid
ebujalance@mat.uned.es Spain
jmgamboa@mat.ucm.es
Francisco Javier Cirre
Facultad de Ciencias, UNED Grzegorz Gromadzki
Matemáticas Fundamentales University of Gdansk
C/ Senda del Rey 9 Department of Mathematics
28040 Madrid Wita Stwosza 57
Spain 80-952 Gdansk
jcirre@mat.uned.es Poland
greggrom@mat.ug.edu.pl
The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
springer.com
To Álvaro and my grandchildren
To my family
To Belén, Álvaro and Irene
To the memory of my parents
Preface
vii
viii Preface
number of non-conjugate symmetries. We next consider the same problem for all
the symmetries (conjugate or not) of a Riemann surface. We finally deal with the
total number of ovals of a pair of symmetries in terms of the order of its product and
the genus of the surface.
The monograph is actually devoted to the symmetries of Riemann surfaces of
genus at least two since they are the ones uniformized by the hyperbolic plane. The
theory of symmetries of the remaining surfaces, that is, the Riemann sphere and the
tori, is well-known for a long time but, for the sake of completeness and the reader’s
convenience, we devote the main part of Chap. 4 to this subject. We also outline the
classification of the symmetry types of hyperelliptic Riemann surfaces as being the
double covers of the Riemann sphere.
Finally, Chap. 5 is dedicated to the symmetries of Riemann surfaces with large
groups of automorphisms. Such surfaces are important since on the one hand they
are determined by a 2-generator presentation of their groups of automorphisms, and
on the other hand they can be defined over the algebraic numbers due to the cele-
brated theorem of Belyi from the late seventies. Furthermore, by a recent result of
B. Köck and D. Singerman, these algebraic numbers can be chosen to be reals if
the surface is symmetric. The foundations for the study of symmetries of such sur-
faces were established by Singerman, who found necessary and sufficient algebraic
conditions in terms of the mentioned above generating pair for such a surface to
be symmetric. In the first section, apart from Singerman’s proof, we give formulae
to compute the number of ovals of these symmetries, to which we refer as Singer-
man symmetries. Using these formulae we deal, in the next two sections, with the
significant families of Macbeath-Singerman and Accola-Maclachlan and Kulkarni
surfaces. Finally we describe the symmetries of the last two families by means of
algebraic formulae.
Acknowledgments. The authors are very grateful to the three referees for their very
helpful, deep and accurate comments concerning the first version of the monograph.
We are also very grateful to Dr Ewa Kozłowska-Walania for her careful reading of
the final version and the number of conspicuous and helpful comments.
Emilio Bujalance is partially supported by Spanish MTM2008-00250.
Francisco Javier Cirre is partially supported by Spanish MTM2008-00250.
José Manuel Gamboa is partially supported by Spanish MTM2008-00272, Proyecto
Santander Complutense PR34/07-15813 and GAAR Grupos UCM 910444.
Grzegorz Gromadzki is supported by the Sabbatical Grant SAB2005-0049 of the
Spanish Ministry of Education and by the Research Grant NN201 366436 of the
Polish Ministry of Sciences and Higher Education.
1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1
1.1 NEC Groups and Their Signatures . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1
1.2 Normal Subgroups of NEC Groups . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 6
1.3 Centralizers of Reflections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 8
1.4 Uniformization and Automorphism Groups of Riemann
and Klein Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 11
1.4.1 Maximal NEC Groups .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 13
1.4.2 Teichmüller Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 14
1.5 Symmetric Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 15
1.5.1 Algebraic Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 18
ix
x Contents
6 Appendix . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .145
6.1 Compact Riemann Surfaces Without Symmetries .. . . . .. . . . . . . . . . . . . . . . .145
References .. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .151
Index . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157
Introduction
An easy computation shows that at any point in the complex projective plane P2 (C),
the partial derivatives of Ft are not simultaneously zero. So, for t = 0, 1, each set
ϕ : S0 → S1 ; [x : y : z] → [ξx : ξ 4 y : ξ 3 z],
where ξ = eiπ/4 . However, their sets of R-rational points, that is, the real curves
S0 (R) and S1 (R), are not birationally R-isomorphic. Indeed, both are smooth but
S0 (R) is connected while S1 (R) has two connected components. The paper [32] by
Cirre and Gamboa presents many other examples of non-isomorphic real algebraic
curves with isomorphic complexifications.
xi
xii Introduction
These phenomena lead naturally to the following problems we deal with in this
monograph:
(1) Is a complex smooth algebraic curve C definable over the reals?
(2) Assume that this question has an affirmative answer. How many non-
birationally R- isomorphic real algebraic curves admit C as its complexification?
The projective smooth models of such real curves are usually called the real
forms of C.
(3) What can be said about the topology of the real forms of C?
The expository work by Gromadzki [51] can be understood as the first attempt to
survey the known answers to these questions. Because of the methods to be used, it
seems convenient to translate these questions into a more suitable language. To that
end we use the terminology introduced at the beginning. In particular, the first of the
above problems reads off: is a compact Riemann surface symmetric?
Let σ and τ be symmetries of the compact Riemann surface S. The pairs (S, σ)
and (S, τ ) are real forms of S; they are said to be isomorphic if there exists an
automorphism ϕ of S such that σ = ϕ ◦ τ ◦ ϕ−1 . In this way the second problem
to be treated is the counting of the number of conjugacy classes of symmetries
with respect to the group Aut(S) of analytic and antianalytic automorphisms of
the Riemann surface S.
Finally, the topological type of a symmetry σ of S is determined, together with
the genus of S, by the number of connected components, or ovals (in the nineteenth
century Hilbert’s terminology) of the fixed point set Fix(σ) = {p ∈ S : σ(p) = p}
and the connectedness character of its complement S \ Fix(σ) in S. More precisely,
the triple (g, k, ε) is said to be the topological type of a symmetry σ of a genus
g surface S if the set Fix(σ) has k connected components, and ε = 1 or ε = 0
according to whether S \Fix(σ) is connected or not. We say that σ is non-separating
if ε = 1 and separating otherwise.
A classical result due to Harnack [59] and Weichold [127] states that the nec-
essary and sufficient conditions for a triple to be admissible, that is, to be the
topological type of some symmetry σ, are the following:
it has two connected components. In the first case, i.e., if σ is non-separating, then
the orbit space Xσ = S/σ is non-orientable, while in the separating case Xσ is
orientable.
An expository account of the functorial correspondence between real algebraic
curves and Klein surfaces can be found in [45], see also the condensed versions
[103, 104] by Natanzon.
As we shall see throughout this monograph, a fundamental component to ap-
proach the problems mentioned above is the knowledge of the full automorphism
group Aut(S) of the analytic and antianalytic automorphisms of S and its subgroup
Aut+ (S) consisting of the analytic ones. Moreover, to determine the topology of
a given symmetry σ of S, the centralizer C(Aut(S), σ) of σ in Aut(S) plays a
fundamental role. Although automorphism groups do not constitute the core of this
work, we will need them very frequently. It is worth mentioning that the factor group
C(Aut(S), σ)/σ is isomorphic to the group of automorphisms of the Klein sur-
face S/σ . There is a vast literature concerning groups of automorphisms of such
surfaces. Among them we should mention [22], [58], [79]–[90], [93], the pioneering
papers [114] and [40] and the exceptionally complete work [105].
We now describe briefly the content of this monograph. We also quote the con-
tributions of different authors to the development of the employed techniques and
related topics.
Although in Chap. 4 we study the symmetries of the sphere and the tori, we
will mainly be concerned with compact Riemann surfaces of genus bigger than one.
By the Uniformization Theorem, such a surface S can be presented as the orbit
space of the hyperbolic plane H under the action of a surface Fuchsian group Γ.
Moreover, using covering theory, it can be proved that each automorphism group
of S = H/Γ is a factor group Λ/Γ, where Λ is a non-euclidean crystallographic
(NEC in short) group containing Γ as a normal subgroup. The key point now is
that the algebraic structure of both Fuchsian and NEC groups is well known and
this is why we devote Sect. 1.1 to the presentation of some basic facts about these
groups.
The above shows that in order to move ahead with the combinatorial approach
of the study of symmetries of Riemann surfaces it is essential to understand the
relation between the presentations of two NEC groups Γ and Λ, where the first is a
normal subgroup of the second one. This task is mainly due to E. Bujalance, who
developed in a series of papers [10–12] at the beginning of the eighties, an efficient
method to solve this problem based on surgery of fundamental regions. It is also
worth mentioning the article by J. A. Bujalance [27] concerning this problem. These
results appear, without proofs, in Sect. 1.2.
One of the main elements in the combinatorial approach to the study of symme-
tries of compact Riemann surfaces is the analysis of the centralizers of hyperbolic
reflections in NEC groups. Singerman found in his Ph. D. Thesis [115], see also
[119], the isomorphism type of centralizers of reflections in NEC groups. Going a
bit more into the details of Singerman’s proof, explicit generators of these groups
can be obtained, see the papers [48, 51] by G. Gromadzki. We present them in
Sect. 1.3.
xiv Introduction
valid for those surfaces which have no symmetry with fixed points. The bound,
which depends only on the 2-adic part of g − 1, was obtained originally in [18] and
it was shown to be attained for infinitely many values of g.
Finally, in Sect. 2.4 we obtain an upper bound for the number of conjugacy
classes of symmetries of a genus g surface allowing both fixed point free symmetries
and symmetries with ovals. Once more it turns out that this bound depends only on
the 2-adic part of g − 1.
Chapter 3 deals with several enumerations of ovals of the symmetries of a
Riemann surface. Section 3.1 is crucial for the rest of the monograph; its main result
allows us to find the number of ovals of a symmetry of a Riemann surface S from the
algebraic structure of the full automorphism group Aut(S) and from the topologi-
cal type of the action of Aut(S) on S. It was originally established in [49]. As we
mentioned, a Riemann surface of even genus has at most four non-conjugate sym-
metries and, as an application of the result just quoted, Gromadzki and Izquierdo
found in [54] the maximal total number of ovals of such extremal configuration of
symmetries.
The problem of finding the maximal number of ovals of a fixed number k of non-
conjugate symmetries of a Riemann surface of genus g has been investigated by
many authors throughout the years. However, it has been solved in its full generality
just recently [56]. The first results, concerning low values of k, were obtained by
Natanzon in [96, 100, 105], where he showed that an upper bound for such number
is 2g + 2k−1 for k = 2, 3, 4 and characterized the pairs (g, k) for which this bound
is attained.
Later on, Singerman in [121] showed that for each non-negative integer k there
exist infinitely many values of g for which there exists a Riemann surface of genus
g admitting k non-conjugate symmetries having 2g − 2 + 2k−3(9 − k) ovals in total.
In his work, Singerman also conjectured that this is in fact the best possible upper
bound. This was shown by Gromadzki in [50] to be false for k > 9 by showing
that, for k ≥ 9, the maximal possible number of ovals is 2g − 2 + 2r−3 (9 − k),
where r is the smallest positive integer for which k ≤ 2r−1 . Moreover, this bound
is attained, for arbitrary k ≥ 9, for infinitely many values of g. Later on Natanzon
proved in [107] that Singerman’s conjecture is true under the additional assumption
that the symmetries are separating. The presentation of these results is the main goal
of Section 3.2.
It is worth mentioning that Singerman’s conjecture was found to be true for k = 9
in [50] and it was conjectured to be also true for k in range 5 ≤ k ≤ 8. This has
recently been answered in the affirmative by Gromadzki and Kozłowska-Walania
in [56].
Section 3.3 concerns the total number of ovals of all symmetries of a Riemann
surface. Recall that a simple closed curve on a Riemann surface S is said to be an
oval of S if it is an oval of some symmetry of S. Let S be the number of ovals of
S and let ν(g) be the maximum of S where S runs over all Riemann surfaces of
genus g. Using topological methods, Natanzon proved in [105] that ν(g) ≤ 42(g−1),
and Gromadzki improved this bound in [49] by using combinatorial methods. We
present the complete proofs of these results in this section.
xvi Introduction
The class of p-gonal surfaces has attracted the interest of many authors. In what
concerns symmetries, we quote here the result of Costa and Izquierdo in [35] where
they study the symmetries of cyclic p-gonal Riemann surfaces by means of Fuchsian
and NEC groups. To finish, it is worth mentioning the paper by Bujalance, Costa and
Gromadzki [20], where the behaviour of symmetries with maximal number of ovals
under non-ramified coverings is studied.
Chapter 4 is devoted to the presentation of classical selected examples. To begin
with, we study the Riemann sphere Σ in Sect. 4.1. It is elementary to show that the
maps σ1 : z → z̄ and σ2 : z → −1/z̄ are symmetries of Σ and that they are the only
ones, up to analytic conjugation. Section 4.2 is devoted to classify the symmetries
of the tori, for which we follow closely the approach by Alling [3]. Each torus is
presented as the orbit space C/L for a suitably arranged lattice L. The symmetrical
character of the torus and the topological type of its symmetries are expressed in
terms of the lattice L. As it is classical, the analysis requires the cases of square or
hexagonal lattices to be treated separately.
In Section 4.3 we explain how the complete classification of the symmetries of
hyperelliptic Riemann surfaces was obtained by the authors of this monograph in
their previous work [14]. This work is too extensive even to be completely summa-
rized here, but we explain an example in detail, showing how both the combinatorial
approach and the use of algebraic equations, combined with a topological method,
are fruitful in this case.
In Chap. 5 we deal with symmetries of surfaces S whose group of analytic au-
tomorphisms Aut+ (S) is large enough. Following [52], we call these symmetries
Singerman symmetries. As mentioned above, the symmetrical character of such sur-
faces depends only on Aut+ (S). In Sect. 5.1 we give formulae for the number of
ovals of the symmetries of such surfaces in terms of the orders of the isotropy groups
of some automorphisms acting on Aut+ (S), and the orders of some distinguished
elements in Aut+ (S). These results constitute a fundamental component in the de-
velopment of the next sections of this chapter.
The understanding of the symmetries of the so called Macbeath-Singerman sur-
faces is the goal of Sect. 5.2. These are genus g surfaces admitting the projective
special linear group PSL(2, q), where q is a prime power, as its group of analytic
automorphisms of the maximal order 84(g − 1). Klein [65] was the first to discover
the existence of such surfaces, as he showed that the group of analytic automor-
phisms of the genus 3 surface
known as the Klein quartic, is the projective special group PSL(2, 7) of order 168.
Macbeath [72] proved much later the existence of a unique Riemann surface of
genus 7 on which the group PSL(2, 8) of order 504 acts as its full group of analytic
automorphisms.
Following ideas of Singerman from [118], we show that all Macbeath-Singerman
surfaces are symmetric. We also determine the number of symmetries they admit,
which we call Macbeath-Singerman symmetries, and the topological type of each
xviii Introduction
of them. Remarkably, all of them are non-separating. These results were proved
for the first time by Broughton, Bujalance, Costa, Gamboa and Gromadzki in [8].
The proof we present here is quite different and relies heavily on the results of the
previous section of this chapter.
In the 1960’s, Accola [1] and Maclachlan [77] proved, independently, that for
every integer g ≥ 2 there is a compact Riemann surface Xg of genus g whose
automorphism group has order 8g + 8. It is called the Accola-Maclachlan surface
and it is defined by the polynomial equation y 2 = x2g+2 −1. The result is interesting
as 8g + 8 is the largest order of an automorphism group that can be attained for
every genus g. Much later, Kulkarni [70] considered the question of uniqueness of
the surfaces attaining this bound. It turns out that the Accola-Maclachlan surface Xg
is the unique one if g ≡ 0, 1, 2 (mod 4) and g sufficiently large. However, for large
enough g ≡ 3 (mod 4), Kulkarni also proved that, in addition to Xg , there exists
exactly one other surface, called Kulkarni surface, of genus g whose automorphism
group also has order 8g + 8.
In Sect. 5.3 we show that these surfaces are symmetric and, moreover, we deter-
mine the number of conjugacy classes of symmetries they admit and the topological
type of each of them. As in the example of Sect. 5.2, the proof proposed here re-
lies on the results in Sect. 5.1 and it is quite different from the original one which
appeared in [9].
It must be pointed out that the examples selected to this chapter are in some
sense exceptional because it has been possible to decide successfully the separating
character of each symmetry. But, of course, they are not the only ones. In their
paper [2], Akbas and Singerman not only calculated the number of ovals of the
symmetries of the modular surfaces X0 (N ) = H/Γ0 (N ), but also showed that they
are separating for N = 2, 3, 5, 7, 13 and non-separating for all other primes N .
The situation is slightly worse for the symmetries of the modular surfaces X(N ) =
H/Γ(N ). All of them are non-separating in case N ≡ 3 (mod 4) is prime but, as
far as we know, there is no general answer for primes N ≡ 1 (mod 4).
Another interesting example, that we do not explain in the monograph, is due to
Tyszkowska [126], who obtained sharp upper bounds for the number of ovals of the
symmetries of the Belyi surfaces admitting PSL(2, p) as its group of automorphisms.
Section 5.4 is devoted to finding polynomial equations of the sets of points fixed
by the symmetries of families of Riemann surfaces studied in the precedent ones.
The key point is the Galois theory of finite coverings, as explained to the authors by
P. Turbek. In fact Turbek is responsible for the original finding of equations of the
symmetries of the Accola-Maclachlan surfaces occurring in [9], but in this mono-
graph we have chosen a more geometrical approach. However, the presentation of
the part of this section concerning defining equations of the sets of points fixed by
the symmetries of the Kulkarni surfaces follows closely Turbek’s article [124].
It is convenient to explain a little bit the method employed. We begin with a plane
model of our Riemann surface S, possibly with singularities, defined as the zero set
in C2 of a polynomial P ∈ C[X, Y ]. A symmetry σ of S can be seen as an involution
of the quotient field EP of the coordinate ring C[X, Y ]/(P ) of S. We look for a
different polynomial Q ∈ R[X, Y ] which also defines S. Then the quotient fields
Introduction xix
proper. The asymmetric curves are precisely those whose isomorphism classes oc-
cur in the difference Fix(σg∗ ) \ MR g . Seppälä showed in [110] that every asymmetric
curve is in fact a covering of a real algebraic curve.
It is classical that for any integer g > 2 there exists a compact Riemann surface
of genus g whose group of analytic automorphisms is trivial. Indeed, Greenberg
proved in [47] that outside a proper analytic subset of the Teichmüller space, all
compact Riemann surfaces of genus g ≥ 3 have the identity as its only analytic
automorphism. However, it is not easy to construct examples of such surfaces. It
is worth mentioning the paper by Mednykh [91] who constructed, for each pair of
integers (p, r), where p > 3 is prime and r ≥ 2p, a fundamental region of a Fuchsian
group which uniformizes a compact Riemann surface of genus g = (p− 1)(r − 1)/2
with trivial automorphism group.
Later on, Everitt in [42] found new examples for all g > 2, using Schreier
coset graphs for subgroups of triangle groups. Combining covering theory with
Galois theory of algebraic function fields in one variable, Turbek [123, 125] pro-
vided defining equations of compact Riemann surfaces with trivial group of analytic
automorphisms.
In the same vein, Earle in [40] was the first to find examples of pseudo-real
Riemann surfaces, that is, surfaces without symmetries but with orientation revers-
ing automorphisms. Later on, Bujalance and Turbek constructed in [26] algebraic
equations of the elements of an infinite family of pseudo-real Riemann surfaces. The
construction we present in Chap. 6 is a particular case of the one in [26].
More recently, Bujalance, Conder and Costa in [17] have shown that there exist
pseudo-real Riemann surfaces of genus g for each g ≥ 2 and, furthermore, that the
xx Introduction
The Uniformization Theorem for Riemann surfaces says that every compact Rie-
mann surface S of genus bigger than one is the orbit space of the hyperbolic plane
H under the action of a certain subgroup of the group Aut+ (H) of analytic self-
homeomorphisms of H. Since H is simply connected, every analytic automorphism
of S can be lifted to an analytic self-homeomorphism of H. Analogously, every ori-
entation reversing automorphism of S (a symmetry, for instance) can be lifted to an
antianalytic self-homeomorphism of H. We start this section with a description of
the group Aut(H) of analytic and antianalytic self-homeomorphisms of H.
As a trivial consequence of the maximal modulus principle, it follows that
Aut(H) is the following disjoint union:
az + b
Aut(H) = f :z→
with {a, b, c, d} ⊂ R and ad − bc > 0 ∪
cz + d
az̄ + b
∪ f :z→ with {a, b, c, d} ⊂ R and ad − bc < 0 .
cz̄ + d
The elements of Aut(H) are called automorphisms of H. The first set of this union
is the subgroup Aut+ (H) of analytic automorphisms of H. It consists of the orien-
tation preserving hyperbolic isometries of H, while the orientation reversing ones
are those in the second set.
Let GL(2, R) be the group of 2 × 2 non-singular matrices with real entries. It is
clear that the mapping
ab
GL(2, R) → Aut(H) ; A = → fA
cd
where ⎧
⎪ az + b
⎪
⎪ cz + d if det A > 0,
⎨
fA : H → C ; z →
⎪
⎪
⎪
⎩ az̄ + b if det A < 0,
cz̄ + d
is a group epimorphism. Its kernel {λI2 : λ ∈ R∗ }, where I2 is the identity matrix,
is the center C(GL(2, R)) of GL(2, R). Therefore we identify Aut(H) with the
factor group PGL(2, R) = GL(2, R)/C(GL(2, R)). Consequently, Aut(H) is a
topological group and it makes sense to talk about its discrete subgroups.
Definitions 1.1.1. Let Γ be a subgroup of Aut(H).
(1) We say that Γ is a non-euclidean crystallographic group (shortly, NEC group)
if it is a discrete subgroup and the orbit space H/Γ is compact.
(2) An NEC group Γ is said to be a Fuchsian group if it is contained in Aut+ (H).
Otherwise Γ is said to be a proper NEC group.
(3) Given a proper NEC group Γ, its subgroup Γ+ = Γ ∩ Aut+ (H), consisting of
its orientation preserving elements, is called its canonical Fuchsian subgroup.
Obviously, [Γ : Γ+ ] = 2, and Γ+ is the unique subgroup of index 2 in Γ
contained in Aut+ (H).
If Γ is an NEC group then the orbit space H/Γ can be endowed with a dianalytic
structure, see [4, Theorem 1.8.4]. A fundamental region for Γ can be constructed as
a convex bounded hyperbolic polygon with a finite number of sides. A modification
of the region and a suitable labelling of the edges provides the following canonical
surface symbol:
(+) α1 β1 α1 β1 . . . αg βg αg βg ξ1 ξ1 . . . ξr ξr ε1 γ10 . . . γ1s1 ε1 . . . εk γk0 . . . γksk εk
if H/Γ is orientable, or
(−) α1 α∗1 . . . αg α∗g ξ1 ξ1 . . . ξr ξr ε1 γ10 . . . γ1s1 ε1 . . . εk γk0 . . . γksk εk
Taking into account the surface symbol, it is possible to obtain the following
presentation of the group Γ:
generators:
– x1 , . . . , xr (elliptic elements),
– c10 , . . . , c1s1 , . . . , ck0 , . . . , cksk (reflections),
– e1 , . . . , ek (orientation preserving elements, usually hyperbolic and in some
cases elliptic),
– a1 , b1 , . . . , ag , bg (hyperbolic translations) in case (+),
– d1 , . . . , dg (glide reflections) in case (−),
and relations:
– xmi
i
= 1 for i = 1, . . . , r,
– cisi = e−1i ci0 ei for i = 1, . . . , k,
– c2ij−1 = c2ij = (cij−1 cij )nij = 1 for i = 1, . . . , k and j = 1, . . . , si ,
– x1 · · · xr e1 · · · ek a1 b1 a−1 −1 −1 −1
1 b1 · · · ag bg ag bg = 1 in case (+),
2 2
– x1 · · · xr e1 · · · ek d1 · · · dg = 1 in case (−).
Throughout the monograph, a set of generators as the above one will be called a
set of canonical generators of Γ.
The first presentations for NEC groups appeared in [128] and their structure was
clarified by the introduction of signatures in [73].
Definitions 1.1.2. Let g, k be non-negative integers and let mi for i = 1, . . . , r,
and nij for i = 1, . . . , k and j = 1, . . . , si , be integers ≥ 2.
(1) An abstract signature is a collection of symbols and non-negative integers of
the form
The non-negative integer g is called the orbit genus of s. If the sign “+”
appears then we write sign(s) = “+”; otherwise sign(s) = “−”. The inte-
gers m1 , . . . , mr are called the proper periods of s and the nij are called the
link periods of the period cycle (ni1 , . . . , nisi ). An empty set of proper periods,
(i.e., r = 0), will be denoted by [−], an empty period cycle (i.e., si = 0) by
(−), and the fact that s has no period cycles (i.e., k = 0) by {−}.
(2) Given an NEC group Γ with the above presentation, the signature (1.1) is de-
fined as its signature s(Γ), the sign of s(Γ) being “+” in the case (+) and “−”
otherwise. The orbit genus of s(Γ) is usually called the orbit genus of Γ.
(3) Since a Fuchsian group contains no orientation reversing elements, its signa-
ture has no period cycles and its sign is always “+”. Hence we may drop such
data and in the sequel the signature of a Fuchsian group will be represented
simply by
(g; m1 , . . . , mr ).
Signatures of the form (0; k, , m), that we abbreviate as [k, , m], are called
triangle Fuchsian signatures. Fuchsian groups with such signatures are called
triangle Fuchsian groups.
4 1 Preliminaries
(4) A signature of the form s = (0; +; [−]; {(n1 , . . . , ns )}) will be abbreviated
as (n1 , . . . , ns ), when no confusion can arise. If s = 3 then s is called trian-
gle NEC signature. NEC groups with these signatures are called triangle NEC
groups.
In the obvious manner, a presentation of an NEC group Γ can be read from its
signature. In fact, signatures give a procedure to classify NEC groups up to isomor-
phism, as it was proved by Macbeath [73] and Wilkie [128].
Proposition 1.1.3. Let Γ be an NEC group with signature s = s(Γ) as in (1.1). Let
Γ be another NEC group with signature
Let us write Ci = (ni1 , . . . , nisi ) and Ci = (ni1 , . . . , nis ). Then Γ and Γ are
i
isomorphic as abstract groups if and only if
(1) sign(s) = sign(s ),
(2) g = g , r = r , k = k and si = si for i = 1, . . . , k,
(3) mi = mϕ(i) for a permutation ϕ of {1, . . . , r},
(4) if sign(s) = “+” then there exists a permutation φ of {1, . . . , k} such that one
of the following conditions holds true:
(4a) Ci is a cyclic permutation of Cφ(i) for each i ∈ {1, . . . , k} or
(4b) Ci is a cyclic permutation of the inverse of Cφ(i) for each i ∈ {1, . . . , k},
(5) if sign(s) = “−” then there exists a permutation φ of {1, . . . , k} such that
Ci is a cyclic permutation of either Cφ(i) or of the inverse of Cφ(i) , for each
i ∈ {1, . . . , k}.
If the sign is “+” then the corresponding period cycles are all paired in the same
way (either all directly or all inversely). If the sign is “−” then some period cycles
may be paired directly and some inversely.
In addition to this algebraic information, the signature of an NEC group Γ also
provides topological information about the canonical projection H → H/Γ, see
Proposition 1.1.4 below.
Let Γ be an NEC group and consider the canonical projection f : H → H/Γ. At
each point p ∈ H, the map f behaves locally as z → z m (see [4, Sect. 5] for a more
rigorous statement). The integer m is the ramification index of f at p; we say that
f is ramified at p if m > 1. The ramified points of f are precisely those fixed by
orientation preserving elements of Γ. The elliptic elements of Γ fixing p constitute a
cyclic group, whose order is the ramification index of f at such a point. Moreover, it
turns out that all the points in the same fiber as p have the same ramification index.
We say that f (p) is a branch point of f with branching order m. The next result
collects the topological interpretation of s(Γ), as proved by Wilkie in [128].
Proposition 1.1.4. Let Γ be an NEC group with signature (1.1) and let S = H/Γ.
Then:
(1) g is the topological genus of S.
1.1 NEC Groups and Their Signatures 5
Remark 1.1.6. It follows from the presentation of an NEC group that the number
of conjugacy classes of reflections associated to a period cycle with v > 0 even link
periods is v. If the period cycle is empty or all its link periods are odd then all
reflections associated to such period cycle are pairwise conjugate.
Definition 1.1.7. Let s be an abstract signature as given in (1.1) and define η = 2
if sign(s) = “+” and η = 1 otherwise. The area of s is defined to be
⎛ ⎞
r k si
1 1 1 ⎠
Area(s) = 2π ⎝ηg + k − 2 + 1− + 1− .
i=1
m i 2 i=1 j=1
n ij
Part (1) in the following result justifies this definition, see [22].
Theorem 1.1.8. Let Γ be an NEC group with signature s(Γ).
(1) The hyperbolic area of any fundamental region for Γ is Area(s(Γ)). It makes
sense to call area of the NEC group Γ to such common value. We denote it by
Area(Γ).
6 1 Preliminaries
(2) The signature s is the signature of some NEC group Γ if and only if Area(s) > 0
and sign(s) = “+” if g = 0.
(3) If Γ is a subgroup of finite index of an NEC group Γ then Γ is also an NEC
group and the so called Hurwitz–Riemann formula holds:
Area(Γ ) = Area(Γ)[Γ : Γ ].
Theorem 1.2.2. Let {(cij−1 , cij ) : i ∈ I, j ∈ Ji } be the set of all pairs of consecu-
tive canonical reflections in Λ \ Γ. Let let qij be the order of the image in Λ/Γ of the
product cij−1 cij , for i ∈ I and j ∈ Ji , and let p denote the order of the image in
Λ/Γ of the canonical elliptic generator x of Λ corresponding to the proper period
m . Then the proper periods of s(Γ) are the following:
Finally we shall present results concerning the sets of period cycles. If N is odd
then all reflections of Λ belong to Γ, which makes things much easier.
Theorem 1.2.3. Let N be odd. Then each period cycle C = (n1 , . . . , ns ) of s(Λ)
induces N/ period cycles of s(Γ), all of them having the form
(n1 , . . . , ns , . . ., n1 , . . . , ns ),
where is the order in Λ/Γ of the image of the canonical hyperbolic generator of Λ
corresponding to the period cycle C.
The case of even N is more involved and we divide it into two subcases.
Theorem 1.2.4. Let N be even, let C = (n1 , . . . , ns ) be one of the period cycles
of s(Λ) and let be defined as in the previous theorem. Assume that all canonical
reflections corresponding to C belong to Γ. Then C produces N/ period cycles in
s(Γ) of the form
(n1 , . . . , ns , . . ., n1 , . . . , ns )
if the sign of s(Λ) is “ − ”. If the sign is “+” then there exists a non-negative integer
N1 ≤ N/ such that C produces N1 period cycles in s(Γ) of the above form and
N/ − N1 period cycles in s(Γ) of the form
(ns , . . . , n1 , . . ., ns , . . . , n1 ).
The last result deals with the remaining case, that is, not all reflections of a given
period cycle of s(Λ) belong to Γ.
Theorem 1.2.5. Let N be even, let C = (n1 , . . . , ns ) be a non-empty period cycle
of s(Λ) and let {e, c0 , . . . , cs } be the set of canonical generators of Λ corresponding
to C. Assume that the set
is not empty. Denote by n(i, j) the order of the image of ci−1 cj+1 in Λ/Γ. Then,
for each pair (i, j) ∈ J, the numbers ni and nj are even and s(Γ) has N/2n(i, j)
period cycles, each of them consisting of n(i, j) copies of the periods
where the quotients nj+1 /2 and ni /2 are omitted if they are equal to 1.
8 1 Preliminaries
Remark 1.2.6. The proper periods and period cycles of s(Γ) are exactly those
induced by the ones of s(Λ) and described in the above theorems. There are no
more proper periods or period cycles in s(Γ).
An immediate consequence of the above results is the following corollary, originally
proved by Singerman [119]. We provide a different proof which, however, is not a
consequence of the above results. The corollary is essential in the study of symmet-
ric Riemann surfaces since, up to certain extent, it describes how the symmetry type
of a Riemann surface S (see Definition 1.5.5) depends just on the topological data
of the action on S of the group Aut+ (S) of all analytic automorphisms of S, see
Sect. 1.5.1.
Corollary 1.2.7. If
is the signature of the NEC group Γ then the signature of its canonical Fuchsian
subgroup Γ+ is
Remark 1.2.8. Corollary 1.2.7 also follows directly from a more general result of
Singerman [116] about the relation between the signatures of two Fuchsian groups
Λ1 and Λ2 , where Λ1 is a subgroup of Λ2 but not necessarily a normal one.
Proof. Observe that in each case the centralizer C(Λ, ci ) contains the group in the
statement. We shall prove the converse inclusions. For i = 0, let γ i−1 , γ i , γ i+1 be
the edges of a fundamental region F corresponding to ci−1 , ci , ci+1 . Let be the
hyperbolic line containing γ i . Then for λ ∈ Λ, the product λci λ−1 is a reflection
with axis λ(). So Λ centralizes ci if and only if λ() = , or equivalently, if and only
if λ(F) is adjacent to . The element (ci−1 ci )ni /2 (ci ci+1 )ni+1 /2 is the composition
of two half-turns with respect to the ends of γ i and so it is a hyperbolic isometry
with axis and whose translation length equals twice the hyperbolic length of γ i .
Now, composing some power of (ci−1 ci )ni /2 (ci ci+1 )ni+1 /2 with ci (ci ci+1 )ni+1 /2
we produce an element λ ∈ Λ such that λ(F) is an arbitrary face adjacent to and
lying on the same side as F with respect to . On the other hand, composing this last
with ci we produce an element μ ∈ Λ such that μ(F) is an arbitrary face adjacent to
and lying on the opposite side as F with respect to .
The case i = 0 is similar; actually one must repeat the above arguments for the
triple of reflections ecs−1 e−1 , c0 , c1 .
For s = 0, the orientation preserving element e fixes . So F and e(F) have a
common edge and both of them are adjacent to . Hence, for a suitable choice of k,
ek (F) is an arbitrary face adjacent to and lying on the same side as F with respect
to , while considering c0 ek (F) we obtain an arbitrary face lying on the other side
as F with respect to . Therefore, when k runs over all integers and ε = 0 or 1, the
product cε0 ek runs over all elements of the centralizer of c0 .
Using similar ideas we prove the following.
Lemma 1.3.2. Let e, c0 , . . . , cs be a set of canonical generators corresponding to
a period cycle (n1 , . . . , ns ) of an NEC group Λ and let C(Λ, ci ) be the centralizer
in Λ of ci .
(1) If s = 0 and all ni are odd then
s−1
(ni+1 −1)/2 −1
C(Λ, c0 ) =
c0 ⊕ (ci+1 ci ) e .
i=0
(2) If ni , nj are even and ni+1 , . . . , nj−1 are odd, with i < j ≤ s, then
C(Λ, ci ) =
ci ⊕
(ci−1 ci )ni /2 ∗
x−1 (cj−1 cj )nj /2 x ,
(3) If ni , nj are even and ni+1 , . . . , ns , n1 , . . . , nj−1 are odd, with 1 ≤ j ≤ i, then
(ci ci+1 )(ni+1 −1)/2 ci (ci ci+1 )−(ni+1 −1)/2 = ci+1 and ecs e−1 = c0 .
So the segment labelled by γ i is followed by γ i+1 for i = 0, . . . , s − 1 while
γ s is followed by γ 0 . Therefore γ 0 , γ 1 , . . . , γ s , γ 0 are the labels of consecutive
segments on and this labelling repeats on periodically. Hence c0 and β generate
C(Λ, c0 ) indeed. Proofs of claims (2) and (3) are similar and we omit them.
Of particular interest in this monograph is the description of the centralizers of
the canonical reflections of a triangle NEC group. For further reference, we display
the corresponding result in Lemma 1.3.3. The fact that two reflections ci and cj are
conjugate will be denoted by ci ∼ cj .
Lemma 1.3.3. Let Λ be an NEC group with signature (k , , m ) and let c0 , c1 , c2
be a set of its canonical generators. Then, according to the parity of k , , m , the
centralizers in Λ of the reflections c0 , c1 and c2 are the following:
(1) For k = 2k, = 2 and m = 2m we have
C(Λ, c0 ) =
c0 ⊕
(c0 c1 )k ∗
(c0 c2 )m ,
C(Λ, c1 ) =
c1 ⊕
(c0 c1 )k ∗
(c1 c2 ) ,
C(Λ, c2 ) =
c2 ⊕
(c0 c2 )m ∗
(c1 c2 ) .
Theorem 1.4.2. Let X be a compact Klein surface of genus ≥ 2. Then there exists
a surface NEC group Γ such that X = H/Γ as Klein surfaces.
Remark 1.4.3. (1) If X is the orbit space of the compact Riemann surface S under
the action of the antianalytic involution σ and Γ is as above, then S = H/Γ+
as compact Riemann surfaces and
σ = Γ/Γ+ , where Γ+ is the canonical
Fuchsian subgroup of Γ.
(2) If we write the Klein surfaces X and X as H/Γ and H/Γ respectively, then
X and X are isomorphic if and only if Γ and Γ are conjugate in the group
Aut(H).
Theorem 1.4.2, together with the classification of NEC groups by means of signa-
tures, opens the door to the combinatorial approach to the theory of Klein surfaces
and their automorphism groups. We now summarize some general results concern-
ing automorphisms of Klein surfaces due to May [82]. The full group of dianalytic
automorphisms of the Klein surface X will be denoted by Aut(X).
Theorem 1.4.4. Let Aut(X) be the full group of automorphisms of the Klein sur-
face X = H/Γ and let N(Γ) be the normalizer of the surface NEC group Γ in
Aut(H). Then,
(1) N(Γ) is an NEC group.
(2) Aut(X) N(Γ)/Γ.
(3) A group G is a subgroup of Aut(X) if and only if it is isomorphic to a factor
group Λ/Γ for some NEC group Λ containing Γ as a normal subgroup.
Remark 1.4.5. If X is not a Klein surface but a Riemann surface S (in which case
Γ is not a surface NEC group but a surface Fuchsian group) then the above holds
true provided that Aut(S) stands for the full group of analytic and antianalytic au-
tomorphisms of S.
Given an NEC group Λ, a factor group Λ/Γ, where Γ is a surface Fuchsian group,
will be called a smooth factor. An epimorphism θ : Λ → G onto a finite group G,
whose kernel is a surface Fuchsian group, will be called a smooth epimorphism.
Observe that a necessary and sufficient condition for an epimorphism θ : Λ → G
to be smooth is that θ preserves the orders of the elements of Λ of finite order. So
an action of a finite group G on the Riemann surface H/Γ is defined by a smooth
epimorphism θ : Λ → G where Λ is an NEC group and ker θ = Γ.
Moreover, two such actions given by θ : Λ → G and θ : Λ → G are said to be
topologically equivalent if there exists a commutative diagram
Λ
ϕ
/ Λ
θ θ
G
ψ
/ G
Definition 1.4.6. An NEC signature s is said to be maximal if for every NEC group
Λ containing an NEC group Λ with signature s the equality dim(Λ) = dim(Λ )
implies Λ = Λ . The definition of maximal Fuchsian signature is analogous.
Remark 1.4.7. (1) Let s be the signature of a proper NEC group. If the signature
s+ of its canonical Fuchsian subgroup is maximal then so is s.
(2) Almost all Fuchsian signatures turn out to be maximal. A list of those which
fail to be so was obtained by Greenberg in [47] and completed by Singerman
in [117]. The list consists of nineteen pairs of signatures (s, s ) where s is non-
maximal and s is the signature of a Fuchsian group Δ properly containing a
group Δ with signature s and such that dim(Δ) = dim(Δ ). If Δ is normal in
Δ then (s, s ) is said to be a normal pair ; otherwise it is non-normal. The cor-
responding list of normal pairs of NEC signatures was obtained in [12], while
the non-normal pairs were obtained by Estévez and Izquierdo in [41].
It must be pointed out that the maximality of the NEC signature s(Λ) does not imply
the maximality of the NEC group Λ. However, the following two results, proved in
[22, Chap. 5], should be mentioned.
Theorem 1.4.8. If s is a maximal NEC signature then there exists a maximal NEC
group Λ with s(Λ) = s.
Remark 1.4.10. This corollary is also true if H/Γ is a Riemann surface, in which
case Γ is not a surface NEC group but a surface Fuchsian group.
14 1 Preliminaries
The last results in this section rely on rather deep mathematics: Teichmüller theory.
Although we do not enter here into their proofs, we consider it convenient to present
the basics about Teichmüller spaces in the context of NEC groups. The reader is
referred to Sect. 4.7 in the book [113] by Seppälä and Sorvali and the paper [76] by
Macbeath and Singerman.
Let PGL(2, R) be the group of automorphisms of the hyperbolic plane H. Given
an NEC group Λ, let R(Λ) be the set of group monomorphisms t : Λ → PGL(2, R)
such that t(Λ) is also an NEC group. The automorphism group Aut(PGL(2, R)) of
PGL(2, R) acts on R(Λ) by left multiplication, and the orbit space
We start this section with a short proof of the classical Harnack–Weichold Theorem,
[59] and [127].
Theorem 1.5.3. A triple (g, k, ε) is the topological type of some symmetry of a
compact Riemann surface of genus g if and only if
satisfies that ker θ is a surface Fuchsian group having orbit genus g. Thus S =
H/ ker θ is a Riemann surface of genus g having a symmetry of topological type
(g, k, 0).
For ε = 1, let g = g + 1 − k and let Λ be an NEC group with signature
(g ; −; [−]; {(−), . k. ., (−)}). Let θ : Λ → Z2 =
a be the epimorphism induced by
the assignment
Again its kernel ker θ is a surface Fuchsian group of orbit genus g and therefore
S = H/ ker θ is a Riemann surface of genus g having a symmetry of type (g, k, 1).
This completes the proof.
For example, the possible topological types of the symmetries of a genus two
Riemann surface are (2, 1, 0), (2, 3, 0), (2, 0, 1), (2, 1, 1) and (2, 2, 1).
The topological classification of a symmetry σ depends just on the number k of
ovals of Fix(σ) and on the separating character of Fix(σ). These two data constitute
the so called species of σ.
Definition 1.5.4. The species of a symmetry σ, denoted by sp(σ), is +k if its topo-
logical type is (g, k, 0) and −k if its topological type is (g, k, 1).
Let Aut(S) denote the group of analytic and antianalytic self-homeomorphisms
of the Riemann surface S. It is called the automorphism group of S, and its ele-
ments automorphisms of S. Let Aut+ (S) be its subgroup consisting of the analytic
automorphisms. Observe that with this notation, a symmetry of S is an involu-
tion σ ∈ Aut(S) \ Aut+ (S). Two symmetries σ and τ are said to be conjugate
if σ = f ◦ τ ◦ f −1 for some automorphism f (either analytic or antianalytic, see
Remark 1.5.6) of S. Clearly, two conjugate symmetries have the same species. This
motivates the following definition.
Definition 1.5.5. The symmetry type of S is the unordered list of species of all con-
jugacy classes of symmetries of S.
1.5 Symmetric Riemann Surfaces 17
Remark 1.5.6. Notice that two symmetries σ and τ of S are conjugate via a diana-
lytic automorphism if and only if they are conjugate via an analytic automorphism.
Indeed, if τ = ϕσϕ−1 where ϕ is antianalytic then τ = ψσψ −1 where ψ = ϕσ
is analytic. This allows us, if necessary, to focus just on analytic conjugation of
symmetries.
Corollary 1.5.7. The Riemann surface S = H/Γ is symmetric if and only if there
exists a proper NEC group Λ containing Γ as a subgroup of index 2.
The following Lemma 1.5.9 will be useful in the sequel in order to determine
whether a given symmetry fixes points or not.
Lemma 1.5.9. Let σ be a symmetry of a compact Riemann surface S = H/Γ with
Aut(S) = Λ/Γ. Let θ : Λ → Aut(S) be the corresponding smooth epimorphism
with ker θ = Γ. Let us write σ = θ(λ) for some λ ∈ Λ. Then
18 1 Preliminaries
What makes a compact Riemann surface symmetric? Let us look for necessary al-
gebraic conditions on the group Aut+ (S) of analytic automorphisms of a compact
Riemann surface S for it to be symmetric. Let S and Aut+ (S) be represented re-
spectively as H/Γ and Δ/Γ for some Fuchsian groups Γ and Δ, where this last
contains the first as a normal subgroup. Let us view Δ as an abstract group K and
let H be the corresponding normal subgroup of K isomorphic to Γ. By Corollary
1.2.7, there is only a finite number of groups L1 , . . . , Ln which can be realized as
NEC groups and which contain K as a subgroup of index 2. From the proof of
this result we also know the way in which such embeddings K ⊆ Li look like.
Now it is an entirely algebraic matter to decide which Li contains H as a normal
subgroup. Let us call such Li algebraically admissible for S, and let us say that
Li is conformally admissible if in addition it can be realized as an NEC group Λi
containing Δ. It is clear that the existence of algebraically admissible groups is a
necessary condition for S to be symmetric.
The converse is not true: the existence of such a group L may not be sufficient
for S to be symmetric, for reasons of conformal nature. This is so because, as we
already pointed out, dim(Λ) = dim(Λ+ )/2 for the Teichmüller dimensions of a
proper NEC group Λ and its canonical Fuchsian subgroup Λ+ . Simply saying this
means that there are “more” Fuchsian groups isomorphic to K than NEC groups
isomorphic to L. However it is worth mentioning that in such a situation a confor-
mal structure on the underlying topological surface can be defined so that the new
Riemann surface S is symmetric, its group Aut+ (S ) is isomorphic to Aut+ (S)
and the actions of both groups are topologically equivalent. In fact, any NEC group
Λ realizing L contains normal subgroups Δ and Γ isomorphic to Δ and Γ respec-
tively and we have a commutative diagram:
∼ / Δ
Δ
∼ / Δ /Γ
Δ/Γ
1.5 Symmetric Riemann Surfaces 19
where the horizontal arrows are isomorphisms and the vertical ones are the corre-
sponding canonical projections.
In this monograph we shall see that the conformal structure of S usually plays a
modest role in both the qualitative and the quantitative study of symmetries, just up
to determine Aut+ (S) and to decide which of the algebraically admissible groups
are actually conformally admissible. So in certain qualitative and quantitative stud-
ies of symmetries of Riemann surfaces, the matter depends on Aut(S) and on the
topological characteristics of this action. The signature of Λ and the epimorphism
θ : Λ → Aut(S) take care of both. In one word, and this is a general philosophy of
our approach in this monograph, “most of it is an algebraic question”.
Fortunately, the above quoted necessary condition for the surface H/Γ to be
symmetric is also sufficient when Δ is a triangle group, since dim(Δ) = 0. In such
a case any algebraically admissible group is also conformally admissible. Assume
that Δ has signature [k, , m] and let {x1 , x2 , x3 } be a canonical set of generators.
Now a Riemann surface S admitting a group G isomorphic to K/H as the group of
analytic automorphisms corresponds to a pair (a, b) of generators of G of orders k
and respectively and whose product has order m; the surface S can be written as
S = H/Γ, where Γ = ker θ for a group homomorphism θ : Δ → G induced by the
assignment θ(x1 ) = a, θ(x2 ) = b.
To continue, we shall need the following result due to Singerman [118]. As the
original proof contains a harmless gap, we provide a somewhat different one here.
Theorem 1.5.10. Let S be a Riemann surface corresponding to a generating pair
(a, b) where a, b and ab have orders k, and m respectively. Then, S is symmetric if
and only if the mapping ϕ : a → a−1 , b → b−1 or ϕ : a → b−1 , b → a−1 induces
an automorphism of G =
a, b = Aut+ (S).
Proof. Let us write S = H/Γ with Γ a surface Fuchsian group, and G = Δ/Γ =
Aut+ (S) where Δ is a Fuchsian group with signature [k, , m] containing Γ as a
normal subgroup.
If S is symmetric then there exists a proper NEC group Λ containing Δ as a
subgroup of index 2 and Γ as a normal subgroup. Then Λ/Γ = Aut(S). According
to Corollary 1.2.7, there are two possibilities for the signature of Λ. If the periods
of Δ are pairwise different then s(Λ) = (k, , m), but if two periods of Δ coincide,
say k = , then either s(Λ) = (k, , m) or s(Λ) = (0; +; [k]; {(m)}). Assume first
that Λ has signature (k, , m) and let {c0 , c1 , c2 } be a set of canonical generators of
Λ. Then
x1 = c0 c1 , x2 = c1 c2 , x3 = c2 c0 (1.2)
constitute a set of canonical generators for Δ. Note that
c1 x1 c1 = x−1
1 and c1 x2 c1 = x−1
2 . (1.3)
Then, for a = Γx1 , b = Γx2 and v = Γc1 ∈ Λ/Γ we have av = a−1 and bv = b−1 .
Therefore, the assignment a → a−1 , b → b−1 induces in fact an automorphism
of G. Observe that Aut(S) = Λ/Γ is a semidirect product G Z2 .
20 1 Preliminaries
Suppose now that s(Λ) = (0; +; [k]; {(m)}) and let {x, e, c0 , c1 } be a set of
canonical generators of Λ. Then
x1 = x, x2 = c0 x−1 c0 , x3 = c0 c1 (1.4)
Now for a = Γx1 , b = Γx2 and v = Γc0 ∈ Λ/Γ we have av = b−1 and bv = a−1 ,
and therefore the assignment a → b−1 , b → a−1 induces an automorphism of G.
Suppose conversely that the assignment a → a−1 , b → b−1 induces an au-
tomorphism of G. As dim(Δ) = 0, there exists an NEC group Λ with signature
(k, , m) containing Δ. Now a word w = w(a, b) is the identity in Λ/Γ if and only
if w(x1 , x2 ) belongs to Γ, and analogously, w(a−1 , b−1 ) = 1 if and only if w =
w(x−1 −1
1 , x2 ) ∈ Γ. So, as a → a
−1
, b → b−1 induces an automorphism of G, we
see from (1.3) that for w = w(x1 , x2 ) ∈ Γ, also c1 w(x1 , x2 )c1 = w(x−1 −1
1 , x2 ) ∈ Γ
and therefore Γ is a normal subgroup of Λ. In this case, the image of c1 in the factor
group Λ/Γ can be chosen as a symmetry of S.
Finally, assume that the assignment a → b−1 , b → a−1 induces an auto-
morphism of G. Then k = and there exists an NEC group Λ with s(Λ) =
(0; +; [k]; {(m)}) containing Δ as a subgroup of index 2. Also now w = w(x1 , x2 )
belongs to Γ if and only if the same happens to w(x−1 −1
2 , x1 ). Thus, using the above
−1 −1
equalities (1.5) we see that c0 w(x1 , x2 )c0 = w(x2 , x1 ) ∈ Γ for every w ∈ Γ and
therefore Γ is a normal subgroup of Λ. So the surface H/Γ is symmetric because
the image of c0 in Λ/Γ is a symmetry of S.
Remark 1.5.11. From equalities (1.2) and (1.4) above it follows easily that for a
symmetric Riemann surface S with Aut+ (S) =
a, b, its automorphism group
Aut(S) is the semidirect product Aut(S) = Aut+ (S) Z2 =
a, b
t, where
t acts as the automorphism given in the proof of Theorem 1.5.10, and the epimor-
phism from Λ onto Aut+ (S) Z2 is defined as follows: if s(Λ) = (k, , m) then
Proof. Let
2 2 2 2 2
ZN = H0 ≤ H1 ≤ H2 ≤ · · · ≤ Hr−1 ≤ Hr = G
So wx−m = xm w and therefore x+sm w has order 2 for each integer s. Moreover,
Lemma 2.1.4. Let S be a Riemann surface of genus g ≥ 2, and let 2r−1 be the
largest power of 2 dividing g −1. Let G be a 2-group of automorphisms of S of order
2t and assume that t ≥ r + 1. Then G contains a cyclic or a dihedral subgroup of
index 2r .
Proof. Let us write S = H/Γ for some surface Fuchsian group Γ and G = Λ/Γ for
some NEC group Λ containing Γ as a normal subgroup. Assume that Λ has signature
We claim that s(Λ) has either a proper period or a link period. In fact, by the
Hurwitz–Riemann formula we have
⎛ ⎞
g−1 ν
mi − 1 k si
nij − 1 ⎠
= 2t−r ⎝ηh − 2 + k + +
2r−1 i=1
m i i=1 j=1
2nij
Remark 2.1.5. The proof shows that in fact G contains a cyclic subgroup gener-
ated by an orientation preserving element or a dihedral subgroup generated by two
orientation reversing elements, of index 2r in both cases.
Remark 2.1.6. Let G+ denote the subgroup of G consisting of its orientation pre-
serving elements. With the notations in the above proof of Lemma 2.1.4, the
existence of a proper period or a link period in the signature of Λ shows that G+
acts on S with fixed points.
with real points. He also showed that this bound is attained for infinitely many values
of g, those being of the form (2n − 1)2 . Here we go further, namely, we determine
the maximal number of conjugacy classes of symmetries with fixed points that a
compact Riemann surface S of genus g ≥ 2 can admit.
Assume that σ1 , . . . , σk are representatives of the conjugacy classes of symme-
tries of S. Since each σi belongs to a Sylow 2-subgroup of Aut(S) and all Sylow
2-subgroups are conjugate, we may assume that all these symmetries generate a
2-group G. We now establish a fundamental result on this topic, whose first proof
appeared in [23].
Theorem 2.2.1. Let S be a Riemann surface of genus g ≥ 2 and let us write g =
2r−1 u+1 with u odd. Then the maximum number of non-conjugate symmetries with
fixed points that S admits is 2r+1 . Furthermore, this bound is attained if and only if
u ≥ 2r+1 − 3.
Proof. Let k be the number of conjugacy classes of symmetries with fixed points
of S. As we observed above, we can choose representatives of these classes such
that they generate a 2-group, say of order 2t . If t ≤ r then k < 2t ≤ 2r and so the
first part of the statement is proved in this case. If t ≥ r + 1 then the first part is a
direct consequence of Lemma 2.1.4, Corollary 2.1.3 and Lemma 2.1.2.
Let now S = H/Γ be a Riemann surface with the maximum number 2r+1 of
conjugacy classes of symmetries with fixed points and let G be a 2-group generated
by 2r+1 representatives of these classes. Let us write G = Λ/Γ for some NEC group
Λ with signature (2.3). Let C1 , . . . , Cn be the different period cycles of Λ involving
these symmetries, and assume that C1 , . . . , Cm are non-empty and Cm+1 , . . . , Cn
are empty. Observe that n > 0 because they are symmetries with fixed points. As
each empty period cycle involves at most one symmetry we see that C1 , . . . , Cm
involve at least 2r+1 − (n − m) symmetries. As each non-empty period cycle Ci of
length si involves at most si non-conjugate symmetries, see Remark 1.1.6, we get
s1 + · · · + sm ≥ 2r+1 − n + m.
Observe that each term (1/2)(1 − 1/nij ) occurring in the formula of Area(Λ) is
not smaller than 1/4 because nij ≥ 2.
Since G is generated by 2r+1 orientation reversing involutions, we see that |G| ≥
r+2
2 . In particular, we may repeat the proof of Lemma 2.1.4 to show that Λ has a
proper period ≥ |G|/2r or a link period ≥ |G|/2r+1 . In the first case
2r s 1 + · · · + sm
Area(Λ) ≥ 2π n − 2 + 1 − +
|G| 4
r+1 r+1
2 + 3n + m − 4 2r 2 −3 2r
≥ 2π − > 2π − .
4 |G| 4 |G|
2.2 Symmetries with Non-Empty Set of Fixed Points 25
|G|
g − 1 ≥ (2r+1 − 3) − 2r−1 ≥ (2r+1 − 3)2r−1 − 2r−1 = 2r−1 (2r+1 − 4).
8
Every even value of g can be written as 2r−1 u + 1 with r = 1 and u odd. In this
way we obtain the main result in [53] as a corollary of Theorem 2.2.1.
√
Henceforth 2( g + 1) > 2r+1 and thus, for the values of g corresponding to
√
u ≥ 2r+1 − 3, the bound 2( g + 1) obtained by Natanzon in [95] for the number
of non-conjugate symmetries with fixed points is not sharp. On the other hand, if
u ≤ 2r+1 − 5 then
√
Hence 2( g + 1) < 2r+1 in this case and so, for the values of g corresponding to
u≤2 r+1
− 5, Natanzon’s bound is better than the one in Theorem 2.2.1. We now
calculate sharp bounds for the remaining values of g.
Notation 2.2.4. For each integer g ≥ 2 we denote by μf (g) the maximal number of
conjugacy classes of symmetries with fixed points that a genus g Riemann surface
can admit.
Next we shall calculate the remaining values of this function, as stated in [23]. To
that end we fix r ≥ 2 (since the case r = 1 is solved in Corollary 2.2.2) and consider
the function
2s−4 − 1
f (s) = ,
2r−s
which is strictly increasing for s > 3. Since f (4) = 0 and f (r + 3) = 2r+2 − 8
we see that for each odd positive integer u < 2r+2 − 7 there exists a unique integer
s ∈ {4, . . . , r + 2} such that f (s) < u ≤ f (s + 1), that is,
2s−4 − 1 2s−3 − 1
< u ≤ .
2r−s 2r−s−1
The next theorem shows that μf (g) depends on this value of s; in fact, it shows
that μf (g) = min{2r−s+2 u + 4, 2s−1 }.
Theorem 2.2.5. Let g = 2r−1 u + 1, where r ≥ 2 and u < 2r+2 − 7 is odd. Let s
be defined as above. Then
2.2 Symmetries with Non-Empty Set of Fixed Points 27
⎧
⎪ 2s−4 − 1 2s−3 − 1
⎪
⎨2
r−s+2
u+4 if < u ≤ ;
2r−s 2r−s
μf (g) =
⎪
⎪ 2s−3 − 1 2s−2 − 2
⎩ 2s−1 if < u ≤ .
2r−s 2r−s
If t ≥ s + 1 then k ≤ 2r−s+1 u + 4 ≤ 2s−1 , where we have used (2.5) for the first
inequality and (2.7) for the second. If t ≤ s then k ≤ 2t−1 ≤ 2s−1 by (2.4).
To finish the proof we consider an arbitrary integer s ∈ {4, . . . , r + 2} and an
arbitrary odd integer u in the range
Let G = Zs2 = x1 ⊕ · · · ⊕ xs . Let A be the set consisting of the 2s−1 involu-
tions of G which can be written as words of odd length in x1 , . . . , xs . Let us write
k = 2r−s+2 u + 4 (k ≥ 5) and let Λ be a maximal NEC group with signature
(0; +; [−]; {(2, . k. ., 2)}). Observe that k > s because u > (2s−4 − 1)/2r−s . Hence
there exists an epimorphism θ : Λ → G such that the image θ(ci ) of each canonical
reflection belongs to A and θ(ci ) = θ(ci+1 ). In addition, if k ≤ 2s−1 then θ can
be defined so that the k canonical reflections c1 , . . . , ck are mapped onto distinct
elements of A.
28 2 Number of Conjugacy Classes of Symmetries
Then ker θ is a surface Fuchsian group and the orbit space S = H/ ker θ is a
Riemann surface of genus 2r−1 u + 1 having G as its full group of automorphisms.
Now, if k ≤ 2s−1 , which happens if and only if u ≤ (2s−3 − 1)/2r−s, then {θ(ci ) :
i = 1, . . . , k} are representatives of the conjugacy classes of symmetries with fixed
points in S. On the other hand, if k > 2s−1 , which happens if and only if u >
(2s−3 − 1)/2r−s, then the 2s−1 elements in A are representatives of the conjugacy
classes of symmetries with fixed points in S.
Remark 2.2.6. The values of u in the range 2r+1 − 3 ≤ u < 2r+2 − 7 are covered
by both Theorems 2.2.1 and 2.2.5. Let us check that the formulae of μf (g) given
by these theorems coincide for these values of u. First, Theorem 2.2.1 gives directly
μf (g) = 2r+1 . To apply the formula of Theorem 2.2.5 we observe that the value of
the parameter s corresponding to those u in the range 2r+1 − 3 ≤ u < 2r+2 − 7 is
s = r + 2. So,
because u + 4 ≥ 2r+1 + 1.
Example 2.2.7. The function g → μf (g) is not increasing because μf (g) = 4 for
all even values of g (see Corollary 2.2.2). However, if we write g = 2r−1 u + 1 and
fix a value of r then the function u → μf (2r−1 u + 1) is increasing (but not strictly)
as a function of u. It attains the maximal value 2r+1 for u = 2r+1 − 3 and remains
constant from that moment onwards. We illustrate this in Table 2.1, where the pairs
(g, μf (g)) are computed for small values of r.
In this section we shall deal with symmetries without fixed points. These symme-
tries correspond to the so called purely imaginary curves, that is, complex algebraic
curves which can be defined over the reals but have no R-rational points.
For an arbitrary value of g ≥ 2, let μi (g) denote the maximal number of con-
jugacy classes of fixed point free symmetries that can be admitted by a Riemann
surface S of genus g which has no symmetry with fixed points.
Theorem 2.3.1. Let us write g = 2r−1 u + 1 with u odd. Then μi (g) ≤ 2r . Further-
more, this bound is attained whenever u ≥ 2r + 1.
|G|
m |G|
u= r
h − 2 + s
= r+s (2s (h − 2) + m) .
2 2 2
This yields that the order of G divides 2r+s because u is odd. The image in G of the
elliptic element of order 2s is an orientation preserving element which generates a
cyclic subgroup of index 2r . Hence μi (g) ≤ 2r by Lemma 2.1.2.
To prove the second part, let u ≥ 2r + 1 and let Λ be a maximal NEC group
with signature (h; −; [2, 2, 2]; {−}), where h = (u + 1)/2 ≥ r + 1. Take G =
Zr+1
2 with generating basis {z1 , . . . , zr+1 }, and let θ : Λ → G be the epimorphism
given by θ(di ) = zi for 1 ≤ i ≤ r + 1, θ(di ) = z1 for r + 2 ≤ i ≤ h and
θ(x1 ) = z1 z2 , θ(x2 ) = z2 z3 , θ(x3 ) = z1 z3 . Then Γ = ker θ is a surface Fuchsian
group and X = H/Γ is a Riemann surface of genus g = 2r−1 u + 1, without
symmetries with fixed points and admitting 2r conjugacy classes of fixed point free
symmetries.
The results are more precise if we restrict our considerations to Riemann surfaces
whose full group Aut(S) acts without fixed points, that is, no automorphism of S
(either analytic or antianalytic) fixes points in S. These are the surfaces S for which
the normal covering S → S/ Aut(S) is unramified.
For each g ≥ 3, let μwi (g) denote the maximal number of conjugacy classes of
symmetries that a genus g Riemann surface whose full group Aut(S) acts without
fixed points may admit. Observe that μw i (g) does not make sense for g = 2 since all
surfaces of genus 2 are hyperelliptic and the hyperelliptic involution fixes points.
30 2 Number of Conjugacy Classes of Symmetries
Proof. Let S be a genus g Riemann surface such that Aut(S) acts fixed point freely
and let G be a 2-group of automorphisms of S generated by representatives of the
conjugacy classes of its symmetries. Let us write S = H/Γ and G = Λ/Γ, where
Γ has signature (g; −) and Λ is a proper NEC group. Since the automorphisms in
G act fixed point freely, the group Λ contains no reflection and no elliptic element;
hence s(Λ) = (h; −; [−]; {−}) for some integer h > 2.
By the Hurwitz–Riemann formula, 2π|G|(h − 2) = 4π(g − 1) = 2r+1 πu, which
implies that |G| divides 2r because G is a 2-group and u is odd. The product of
two symmetries generates a dihedral group of index ≤ 2r−2 and so Corollary 2.1.3
yields μw i (g) ≤ 2
r−1
.
Assume that g ∈ / {3, 5}. If this bound is attained then |G| = 2r , h − 2 = u
and, by Lemma 2.1.2, no element of G has order greater than two. So G = Zr2 .
Moreover, since G is generated by the cosets Γdi , where d1 , . . . , dh form a set of
canonical generators of Λ, it follows that h ≥ r and so u ≥ r − 2.
Conversely, if u ≥ r − 2 then u + 2 ≥ 4 since otherwise g = 3 or 5. So
we may take a maximal NEC group Λ with signature (u + 2; −; [−]; {−}). Let
{d1 , . . . , du+2 } be a set of canonical generators of Λ. Take G = Zr2 with generating
basis {z1 , . . . , zr }, and let θ : Λ → G be the epimorphism induced by the assign-
ment θ(di ) = zi for 1 ≤ i ≤ r and θ(dj ) = z1 for r + 1 ≤ j ≤ u + 2. Then
ker θ is a surface Fuchsian group and S = H/ ker θ is a Riemann surface of genus
g = 2r−1 u + 1 with exactly 2r−1 conjugacy classes of symmetries whose full group
Aut(S) acts fixed point freely.
If the bound were attained for g = 3 or g = 5 then, with the above notations,
s(Λ) = (3; −; [−]; {−}), which is not a maximal signature. Indeed, according to the
list of normal pairs of NEC signatures given in [12], for each Λ with the above sig-
nature, there exists an NEC group Λ with signature s(Λ ) = (0; +; [2, 2, 2]; {(−)})
containing Λ as a normal subgroup of index 2. Up to automorphisms in Λ and
Λ , there is a unique embedding of Λ in Λ , given by d1 = x1 c, d2 = cx2 and
d3 = x2 cx3 x2 , see [12, Proposition 4.8], where {x1 , x2 , x3 , c} is a set of canonical
generators of Λ . Using this embedding it is easy to see that any smooth epimor-
phism θ : Λ → G, where G = Z22 if g = 3 and G = Z32 if g = 5, can be extended
to a smooth epimorphism θ : Λ → G where G = Z32 if g = 3 and G = Z42
if g = 5. Hence ker θ = ker θ and so the Riemann surface H/ ker θ = H/ ker θ
admits automorphisms with fixed points, namely, the images under θ of the elliptic
elements of Λ . This is a contradiction and so μw w
i (3) < 2 and μi (5) < 4.
Let us consider now a maximal NEC group Λ with signature (4; −; [−]; {−})
and define the epimorphisms θ1 : Λ → Z2 = σ by θ1 (di ) = σ for 1 ≤ i ≤ 4 and
θ2 : Λ → Z22 = σ1 , σ2 by θ2 (d1 ) = θ2 (d2 ) = σ1 and θ2 (d3 ) = θ2 (d4 ) = σ2 .
The group Aut(Sj ) of the Riemann surface Sj = H/ ker θj acts fixed point freely
and has one conjugacy class of symmetries if j = 1 and two if j = 2. This yields
μwi (3) = 1 because S1 has genus 3, and μi (5) ≥ 2 because S2 has genus 5.
w
2.4 Symmetries of Surfaces Admitting a Fixed Point Free Symmetry 31
the image under θ of the glide reflection c1 c2 c3 is one of them. Hence the bound
2r+1 is achievable when u ≥ r − 2.
Recall from Theorem 2.2.1 that 2r+1 is also the upper bound for the number
of conjugacy classes of symmetries with fixed points. As a consequence of
Theorems 2.2.1 and 2.4.1 we get the following.
Corollary 2.4.2. The maximum number of non-conjugate symmetries (of any type)
that a Riemann surface of genus g may admit is 2r+1 , where 2r−1 is the largest
power of 2 dividing g − 1.
Theorem 3.1.1. With the above notations, the number of ovals of a symmetry σ with
fixed points is
σ = [ C(G, θ(c)) : θ(C(Λ, c)) ],
where c runs over all non-conjugate canonical reflections of Λ whose images under
θ are conjugate in G to σ.
Proof. Let us write σ = Γσ /Γ for some NEC subgroup Γσ of Λ. On the one
hand, Fix(σ) is homeomorphic to the boundary of S/σ, whilst on the other hand
the number of ovals of Fix(σ) coincides with the number of empty period cycles of
s(Γσ ), because the Riemann surfaces S/σ and H/Γσ are isomorphic.
Hence, we have to count the reflections of Λ belonging to Γσ but non-conjugate
there. Observe that Γσ = θ−1 (σ). So, as σ has fixed points, it is conjugate to
θ(ci ) for some canonical reflection ci of Λ; without loss of generality we can as-
sume that θ(ci ) = σ because conjugate symmetries have the same number of ovals.
Now, given w ∈ Λ, its conjugate cw i belongs to Γσ if and only if w belongs to
θ−1 (C(G, θ(ci ))), the inverse image of the centralizer of θ(ci ) in G, which we shall
denote by Ci . In particular we see that Ci normalizes Γσ and so, for v, w ∈ Ci , the
−1
reflections cvi and cw
i of Γσ are conjugate in Γσ if and only if w v ∈ C(Λ, ci )Γσ .
As a consequence, the conjugates of ci give rise to
[Ci : wC(Λ, cj )Γw−1 ] = [Cj : C(Λ, cj )Γ] = [C(θ(Λ), θ(cj )) : θ(C(Λ, cj ))]
In this section we shall find a bound for the maximum number of ovals that k non-
conjugate symmetries of a Riemann surface can admit. For non-conjugate separating
symmetries such a bound was found by Natanzon [96] using methods different from
ours. Our result is more general as we find here the bound without any assumption
3.2 Total Number of Ovals of Non-Conjugate Symmetries 35
Proof. Let us write S = H/Γ and G = Λ/Γ where Γ is a surface Fuchsian group
and Λ is a proper NEC group containing Γ as a normal subgroup. Let θ : Λ → G be
the canonical epimorphism. We write the signature of Λ as
g − 1 ≥ (4m + 4 − 8 + s)|G|/8.
Let us show now that we can assume that = 0. Suppose that = 0 and consider
an NEC group Λ with signature
which has one empty period cycle less than s(Λ). By a little abuse of the notations,
for the sake of technical simplicity, we denote in the same way as in the group Λ
some of the canonical generators of Λ ; namely those generators which correspond
to “pieces” of the signature of Λ in the signature of Λ and for the sake of termino-
logical convenience we shall refer to these generators of Λ as old generators. To be
more precise, this means that the hyperbolic generators of Λ are a1 , b1 , . . . , ag , bg
or d1 , . . . , dg , according to whether sign(s ) = “+” or sign(s ) = “−”, the elliptic
generators are x1 , . . . , xr , the generators corresponding to the first non-empty pe-
riod cycle are e1 , c0 , c1 , c2 , c3 , c10 , c11 , . . . , c1s1 , the generators corresponding to
the remaining non-empty period cycles are ei , ci0 , ci1 , . . . , cisi , whilst the genera-
tors corresponding to empty period cycles are em+1 , cm+1 , . . . , em+−1 , cm+−1 .
Furthermore, according to this convention, c0 , c1 , c2 and c3 , are the new generators,
whilst the remaining are the old ones.
We may assume, without loss of generality, that θ(c10 ) = z and we shall consider
the following two cases separately:
θ (e1 ) = θ(e1 · · · em+ )θ(e2 · · · em+−1 )−1 , θ (c0 ) = θ (e1 c1s1 e−1
1 ),
θ (c1 ) = θ (c3 ) = z, and θ (c2 ) = θ(cm+ ).
Then Γ = ker θ is a surface Fuchsian group. Indeed, by Theorem 1.2.2 its signa-
ture has no proper periods, by Theorem 1.2.5 it has no link periods, and finally, by
Theorem 1.2.1, its sign is “ + ”. Let S = H/Γ . As Area(Λ) = Area(Λ ) both S
and S have the same genus. Let Σ be the sum of the number of ovals of represen-
tatives of all conjugacy classes of symmetries with fixed points of the surface S .
We shall prove now that Σ ≤ Σ .
As the images under θ of all, except c10 , old canonical reflections corresponding
to non-empty period cycles and their neighbours coincide with their images under
θ it follows, from Theorem 3.1.1 and either (1) or (2) of Lemma 1.3.1, that each of
these reflections contributes with at least as many ovals to Σ as to Σ. Similarly, by
Theorem 3.1.1 and part (3) of Lemma 1.3.1, old reflections corresponding to empty
38 3 Counting Ovals of Symmetries of Riemann Surfaces
Using again Theorems 1.2.1, 1.2.2 and 1.2.5 it follows straightforwardly that Γ =
ker θ is a surface Fuchsian group and hence S = H/Γ is a Riemann surface. Its
genus coincides with that of S by the Hurwitz–Riemann formula. Let Σ denote
again the sum of the numbers of ovals of representatives of all the conjugacy classes
of symmetries with fixed points of the surface S . As in the precedent case we shall
prove that Σ ≤ Σ .
Also here each old canonical reflection but c10 contributes to Σ with as many
ovals as it contributes to Σ. The new reflection c2 contributes to Σ with no less ovals
than c10 contributes to Σ. Moreover, cm+ contributes to θ(cm+ ) with either
|G|/4 or |G|/2 ovals according to θ(em+ ) = 1 or θ(em+ ) = 1. In the first case
we see that Σ ≤ Σ as also c3 contributes with |G|/4 ovals to Σ . If θ(em+ ) = 1,
then θ (e1 ) = θ(e1 ). Hence, in this case θ (c0 ) = θ(c10 ) and therefore both c1 and
c3 contribute with |G|/4 ovals to Σ . Thus, again Σ ≤ Σ .
This way we have substituted Λ by another NEC group Λ whose signature has
one period cycle less that s(Λ) and Σ ≤ Σ . After repeating the argument we can
assume from the very beginning that Λ has no empty period cycles, i.e., its signature
has the form
The next step is to show that, actually, we can assume that m = 1, that is, Λ has
just one period cycle. Suppose that m > 1. Observe first that we can assume that
θ(c1s1 ) = z and θ(c20 ) = z. Let Λ be an NEC group with signature
Observe that s(Λ ) has one period cycle less than s(Λ). Moreover, the reflections of
Λ corresponding to the first period cycle are c10 , . . . , c1s1 , c0 , c20 , . . . , c2s2 , c1 , c2
and it is easily seen that Area(Λ) = Area(Λ ). Let us define an epimorphism θ :
Λ → G via the following assignment: except on e1 , the epimorphisms θ and θ
coincide when acting on all old canonical generators and, furthermore, we define
θ (e1 ) = θ(e1 )θ(e2 ), θ (c0 ) = θ (c1 ) = z and θ (c2 ) = θ (e−1
1 )θ(c10 )θ (e1 ).
Using once more Theorems 1.2.1, 1.2.2 and 1.2.5 we realize that Γ = ker θ is a
surface Fuchsian group. Then S = H/Γ is a Riemann surface of genus g. Let Σ
be the sum of the numbers of ovals of representatives of all the conjugacy classes of
symmetries with fixed points of the surface S . Let us check that Σ ≤ Σ . Indeed,
all old canonical reflections, but c10 and c20 , contribute to Σ with at least as many
ovals as they contribute to Σ. Let us denote by wiΣ the contribution of ci0 to Σ. Then
wiΣ = qi /4ki ,
where qi is the order of the centralizer of θ(ci0 ) in G and ki is the order of
the product θ(ci0 ci1 )ni1 /2 θ(e−1 i (cisi −1 cisi )
nisi /2
ei ) for i = 1, 2. In particular
wi ≤ qi /4. On the other hand, since the elements θ (c10 c11 )n11 /2 θ (e1 c1 c2 e−1
Σ
1 )
and θ (c1s1 −1 c1s1 )n1s1 /2 θ (c1s1 c0 ) have order 2, we see that c10 and c1s1 contribute
with no less ovals to Σ than c10 contributes to Σ. Similarly, it is easy to check that
c20 and c2s2 contribute to Σ with no less ovals than c20 contributes to Σ. This proves
our claim Σ ≤ Σ .
Therefore, after repeating the process, we conclude that it is sufficient to deal
with the case of an NEC group Λ with signature
s(Λ) = (g ; ±; [m1 , . . . , mr ]; {(n1 , . . . , ns )}).
Observe that the Hurwitz–Riemann formula yields
s ≤ 8(g − 1)/|G| + 4. (3.2)
Let c0 , . . . , cs denote the corresponding canonical reflections. We may assume that
θ(c0 ) is a central symmetry of S and so in particular θ(c0 ) = θ(cs ). Consider
c0 , . . . , cs−1 as s points on a circle labelled by the symbols θ(c0 ), . . . , θ(cs−1 ) re-
spectively. By Lemma 3.2.2, at least for k − 1 indices in range 0 ≤ i1 < · · · <
ik−1 ≤ s − 1, we have θ(cij −1 ) = θ(cij +1 ), where the indices are taken modulo s.
Now, if nij > 2 or nij +1 > 2 then θ(cij ) is not central in G, and so
|C(G, θ(cij ))| ≤ |G|/2. Therefore cij contributes to Σ with at most |G|/8 ovals.
If both nij = nij +1 = 2 then |θ(C(Λ, cij ))| ≥ 8 and thus also now cij contributes
with at most |G|/8 ovals to Σ. The remaining canonical reflections contribute to Σ
with no more than |G|/4 ovals. Consequently, using inequality (3.2) we get
(k − 1)|G| (s − k + 1)|G| s|G| (1 − k)|G|
Σ≤ + = +
8 4 4 8
(1 − k)|G| (9 − k)|G|
≤ 2(g − 1) + |G| + = 2(g − 1) + .
8 8
This completes the proof.
40 3 Counting Ovals of Symmetries of Riemann Surfaces
Remark 3.2.4. We will see in Theorem 3.2.6 that the bound obtained in
Theorem 3.2.3 is sharp for k ≥ 9. However, the proof of Theorem 3.2.3 works
for k ≥ 3, as is easy to see.
Proof. As we are looking for the number of ovals of {σ1 , . . . , σk } and conjugate
symmetries have the same number of ovals, we can assume, using Sylow theorem,
that {σ1 , . . . , σk } generate a 2-subgroup G of Aut(S). Clearly, k ≤ |G|/2 and,
moreover, |G| ≥ 2r by the definition of r.
Write S = H/Γ and G = Λ/Γ for a surface Fuchsian group Γ and a proper NEC
group Λ containing it as a normal subgroup. Write the signature of Λ as (3.1) in the
proof of Theorem 3.2.3. As s + ≥ k ≥ 9, it follows from [22, Theorem 2.4.7]
and [41] that s(Λ) is a maximal NEC signature. Hence, by [22, Theorem 5.1.2],
there exists a maximal NEC group Λ isomorphic to Λ, say via ϕ : Λ → Λ . Let
S = H/Γ , where Γ = ϕ(Γ). Then Aut(S ) = Λ /Γ and ϕ induces a group
isomorphism ϕ : Λ/Γ → Λ /Γ . Therefore τ1 = ϕ(σ
1 ), . . . , τk = ϕ(σ
k ) are non-
conjugate symmetries of S .
The group generated by each σi can be written as σi = Λi /Γ for some proper
NEC subgroup Λi of Λ containing Γ as a subgroup of index 2. Here σi is the
number of period cycles of s(Λi ). Thus σi = τi because τi = ϕ(Λi )/Γ .
Therefore G ∼ = Aut(S ) is a 2-group and, by Theorem 3.2.3,
The next result shows that the bound obtained in Corollary 3.2.5 is attained for
infinitely many values of g. We present here the proof given in [50].
Theorem 3.2.6. Let k ≥ 9 be an integer and let r be the smallest positive integer
satisfying k ≤ 2r−1 . Then, for each t ≥ k − 3 there exists a Riemann surface S
of genus g = 2r−2 t + 1 having k non-conjugate symmetries σ1 , . . . , σk with fixed
points such that
Proof. Let G = Zr2 = z1 ⊕ · · · ⊕ zr , and let Λ be a maximal NEC group with
s(Λ) = (0; +; [−]; {(2, .2s
. ., 2)}), where s = t + 2 ≥ k − 1. Let {a1 , . . . , a2r−1 }
be all elements of order 2 in G which have odd length with respect to the set
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 41
In the same vein as Theorem 3.1.1 we have the following result, see [49].
Theorem 3.3.2. Let S be a Riemann surface represented as H/Γ for some surface
Fuchsian group Γ. Let Λ be an NEC group and let θ : Λ → Aut(S) be a group
epimorphism with ker θ = Γ. Then
(1) S = [Aut(S) : θ(C(Λ, c))], where c runs over a set of representatives of
conjugacy classes of canonical reflections of Λ.
(2) 2S ≤ k |Aut(S)|, where k is the number of non-conjugate canonical reflec-
tions of Λ.
(3) Suppose that every period cycle of Λ has an even link period. Then
4S ≤ t |Aut(S)|, where t is the number of even link periods of s(Λ).
We shall use Theorem 3.3.2 to estimate an upper bound for the function ν.
Throughout the rest of this section we write the genus g surface as S = H/Γ for
some surface Fuchsian group Γ and denote G = Aut(S). Moreover, we denote by
θ : Λ → G a group epimorphism with ker θ = Γ, where Λ is an NEC group.
(3.3.3) A first bound for S in terms of Λ and g.
Let us write the signature of Λ as
and therefore
2(g − 1)
|G| ≤ .
ηg + 3/4 + m − 2 + r/2 + ( + s)/4
|G|( + s) (g − 1)( + s)
S ≤ ≤ . (3.4)
2 ηg + 3/4 + m − 2 + r/2 + ( + s)/4
We are ready to present Gromadzki’s proof, see [49], of the inequality ν(g) ≤
12(g −1) for most values of the genus g ≥ 2. More precisely, we have the following.
ν(2) = 24, ν(3) = 36, ν(5) = 72, ν(7) = 126 and ν(9) = 100.
Proof. Using the previous notations, it is enough to study the case ηg + 3/4 + m+
r/2 < 2 since otherwise, by (3.4), we have S ≤ 4(g − 1). In particular ηg < 2.
If ηg = 1 then m = 0 and, since ≥ 1, we get r = 0 and = 1, i.e.,
Area(Λ) = 0, an absurdity. Thus g = 0 and so 3 + 4m + 2r < 8. In particular
< 3 and if = 2 then m = r = 0, i.e., Area(Λ) = 0, an absurdity. Suppose that
= 1, which implies 2m + r ≤ 2, and in particular either m = 0 or m = 1. In the
first case r = 2 and either m1 > 2 or m2 > 2, because Area(Λ) > 0. But in such a
situation Area(Λ) ≥ π/3 and so
|G| 4π(g − 1)
S ≤ ≤ ≤ 12(g − 1).
2 Area(Λ)
where r ≤ 1.
44 3 Counting Ovals of Symmetries of Riemann Surfaces
|G| 4π(g − 1)
S ≤ ≤ = 10(g − 1).
4 4π/10
|G| 20
S ≤ ≤ (g − 1).
2 3
For n = 5 and m = 3, |G| = 30(g − 1). Let c0 and c1 be the canonical reflections
of Λ and let e be the connecting generator. Notice that c0 , e(c0 c1 )2 ∈ C(Λ, c0 ) and
θ(c0 ) = θ(e(c0 c1 )2 ). Furthermore, θ(e(c0 c1 )2 ) = 1, because |θ(e)| = 3 while
|θ((c0 c1 )2 )| = 5. Hence, |θ(C(Λ, c0 ))| ≥ 4 and therefore
|G|
S ≤ < 8(g − 1).
4
For n = 6 we have S ≤ |G|/4 ≤ 6(g − 1). Finally, S ≤ |G|/2 < 11(g − 1)
for n ≥ 7. This finishes the analysis of the case s = 1.
Now let s = 2. Again r = 1, because Area(Λ) > 0, and so
4π(g − 1)
S ≤ |G| = ≤ 12(g − 1).
Area(Λ)
Assume then that m = 2. If both n1 , n2 are odd then Area(Λ) ≥ π/3 and again by
part (2) in Theorem 3.3.2,
4π(g − 1)
S ≤ |G| = ≤ 12(g − 1).
Area(Λ)
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 45
If n1 is even and n2 is odd then Area(Λ) ≥ π/6. Therefore, by (3) in Theorem 3.3.2,
S ≤ |G|/2 ≤ 12(g − 1) because Area(Λ) > 0. In such a case Area(Λ) ≥ π/4
and this implies S ≤ |G|/2 ≤ 8(g − 1). Thus, also the case s = 2 is finished.
The case s = 3 is more complicated and will be considered later on. Assume
now that s ≥ 4. Then for r = 1, Area(Λ) ≥ π(s − 2)/2. Thus, |G| ≤ 8(g − 1)/
(s − 2) and therefore
s|G| 4s
S ≤ ≤ (g − 1) ≤ 8(g − 1).
2 s−2
If all ni are odd then Area(Λ) ≥ 2π/3. Consequently, |G| ≤ 6(g − 1) and
S ≤ 3(g − 1). Thus, assume that some ni is even and observe that, as in this
case Area(Λ) ≥ π(s − 4)/2, one has |G| ≤ 8(g − 1)/(s − 4) and therefore, for
s > 5, and by (3) in Theorem 3.3.2,
2s
S ≤ (g − 1) ≤ 10(g − 1).
s−4
8n
|G| = (g − 1).
n−2
|G| 15
S ≤ ≤ (g − 1).
2 2
In the last case |G| = 30(g − 1). However, the centralizer C(Λ, c0 ) contains both
c0 and (c2 c0 )2 (c1 c2 )(c0 c1 ). Furthermore, θ(c0 ) = θ((c2 c0 )2 (c1 c2 )(c0 c1 )), because
otherwise Γ would contain an orientation reversing element.
Moreover, since θ(c0 c2 ) = 1, it follows that θ((c2 c0 )2 (c1 c2 )(c0 c1 )) = 1. Thus
|θ(C(Λ, c0 ))| ≥ 4 and therefore S ≤ (15/2)(g − 1), as all reflections of Λ are
conjugate.
Assume now that some of the integers k, , m is even. As before we can show,
using part (3) in Theorem 3.3.2, that S ≤ 12(g − 1) for k ≥ 3. Thus, suppose
that k = 2 and recall that we have assumed ≤ m. Observe first that for ≥ 8,
Area(Λ) ≥ π/4 and so
3|G|
S ≤ ≤ 12(g − 1).
4
12m
|G| = (g − 1).
m−3
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 47
6m
S ≤ (g − 1) < 11(g − 1).
m−3
whenever m ≥ 12. Thus, we just have to analyze the cases m ∈ {6, 8, 10}.
Here we have c0 c1 , (c1 c2 )3 ∈ C(Λ, c1 ) and their images under θ are dis-
tinct, since otherwise θ(c0 c2 ) = θ(c1 c2 )4 , which is impossible, as the first of
these two elements has order m while the second one has order 3. Furthermore,
θ(c1 ) ∈ θ(c0 c1 ), θ(c1 c2 )3 . So |θ(C(Λ, c1 ))| ≥ 8. Similarly, one can show that
|θ(C(Λ, c0 ))| ≥ 8. Thus
2|G| |G| 6m
S ≤ + = (g − 1) ≤ 12(g − 1).
8 4 m−3
40m
|G| = (g − 1).
3m − 10
|G| |G|
S ≤ + < 12(g − 1).
8 4
So it remains to deal only with the cases = 3 and = 4.
We shall see below that the group G of all automorphisms of a Riemann surface
which has more than 12(g − 1) ovals is a factor group of some group G with a
certain specific presentation which makes it finite. In virtue of the Hurwitz–Riemann
formula, the order of G imposes restrictions on the genera of the corresponding
Riemann surfaces. In most cases the calculation of the order is easily done. In one
particular case, however, the use of a Computer Algebra System has been necessary.
We acknowledge that all the results obtained have been checked by using the GAP
Program (Groups, Algorithms and Programming) developed by J. Neubüsers group
at Aachen [46].
48 3 Counting Ovals of Symmetries of Riemann Surfaces
16m
|G| = (g − 1).
m−4
If |θ(C(Λ, c1 ))| < 8 then, by either (1) or (2) in Lemma 1.3.3, θ(c0 c1 ) = θ(c1 c2 )2 ,
and therefore θ(c0 c2 ) = θ(c1 c2 )3 . Thus θ(c0 c2 )4 = 1, which implies that m
divides 4. Hence Area(Λ) ≤ 0, which is false. Therefore |θ(C(Λ, c1 ))| ≥ 8.
Now if m is even, similar arguments show that |θ(C(Λ, c0 ))| ≥ 8. So
2|G| |G| 8m
S ≤ + = (g − 1) ≤ 12(g − 1)
8 4 m−4
for m ≥ 12. For m = 10, |θ(C(Λ, c2 ))| ≥ 8, since otherwise θ(c0 c2 )5 = θ(c1 c2 )2
and so θ(c0 c2 )4 = θ(c1 )θ(c2 c0 )θ(c1 ), which implies θ(c0 c2 )5 = 1, an absurdity.
Similarly, |θ(C(Λ, c2 ))| ≥ 8 for m = 6, 8. Therefore, for m = 8, 10,
3
S ≤ |G| ≤ 12(g − 1).
8
Therefore, we just have to deal with the case m = 6.
Observe that |G| = 48(g − 1) and S ≤ 18(g − 1), as we already know that
|θ(C(Λ, ci ))| ≥ 8 for i = 0, 1, 2. If |θ(C(Λ, c1 ))| > 8 for each index i = 0, 1, 2,
then |θ(C(Λ, ci ))| ≥ 12 and S ≤ 3|G|/12 = 12(g − 1). Assume then that
|θ(C(Λ, ci ))| = 8 for some i = 0, 1, 2.
If |θ(C(Λ, c0 ))| = 8 then, by part (1) of Lemma 1.3.3, θ(c0 c1 )θ(c0 c2 )3 is an
element of order 2, and so G is a factor group of the group with presentation
which can be shown to have order 192. So the only values that g can attain in this
case are g = 2, 3 or 5. But later on we shall see that there exists a Riemann surface
of genus 5 with 72 ovals.
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 49
|G| |G|
S ≤ + < 11(g − 1)
8 4
and therefore the only cases left are m = 5 and m = 7. Assume first that m = 7.
Then |θ(C(Λ, c0 ))| ≥ 8. Indeed, if this were not the case then by (2) of Lemma
1.3.3,
Therefore θ(c1 c2 ) and θ(c0 c2 ) are conjugate, which is impossible as they have dis-
tinct orders. Consequently,
2|G| 4m
S ≤ ≤ (g − 1) < 10(g − 1).
8 m−4
Finally, let us assume that m = 5. Then, similarly as above, we can show that
|θ(C(Λ, c0 ))| ≥ 8. We already know that |θ(C(Λ, c1 ))| ≥ 8 and we shall show that
|θ(C(Λ, c1 ))| ≥ 16. Indeed, if this is not the case then θ(c0 c1 )θ(c1 c2 )2 has either
order 2 or order 3. However, one can easily show that in the first case θ(c1 c2 )2 = 1
while in the second one θ(c1 c2 ) = 1, and both equalities are false. Moreover, if
|θ(C(Λ, c1 ))| = 16 then S ≤ |G|/16 + |G|/8 = 15(g − 1) and G is a factor
group of the group with presentation
which can be shown to have order 320. Therefore g equals either 2, 3 or 5 in this
case. Thus, assume that |θ(C(Λ, c1 ))| > 16. Then |θ(C(Λ, c1 ))| ≥ 20 and we shall
look now for the order of θ(C(Λ, c0 )). If |θ(C(Λ, c0 ))| = 8 then S ≤ |G|/20 +
|G|/8 = 14(g − 1) and G is a factor group of the group G with presentation
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )4 , (c0 c2 )5 , ((c0 c1 )(c2 c0 )2 (c2 c1 )2 (c0 c2 )2 )2 ,
Let first m = 2m for some m ≥ 4. We shall show that |θ(C(Λ, ci ))| ≥ 8 for
i = 0, 1. Indeed, if |θ(C(Λ, c0 ))| < 8 then, by Lemma 3.3.1, θ(c0 c1 ) = θ(c0 c2 )4 ,
which implies θ(c2 c1 ) = θ(c0 c2 )3 and so θ(c0 c2 )9 = 1, a contradiction. Now if
|θ(C(Λ, c1 ))| < 8 then, again by Lemma 3.3.1, θ(c0 c1 ) = θ(c2 c1 )θ(c2 c0 )m θ(c1 c2 ).
Using the defining relations for Λ we can easily show that
θ(c0 c2 )2 = θ(c1 )θ(c0 c2 )m +1 θ(c1 ).
However, the last implies that θ(c0 c2 )4 = θ(c1 )θ(c0 c2 )2 θ(c1 ) and so θ(c0 c2 )6 = 1,
which is false. Thus, for m ≥ 12,
2|G|
S ≤ ≤ 12(g − 1).
8
Therefore, in case of even m, it remains to deal with the cases m = 8 and m = 10.
Assume first that m = 8. We know that |θ(C(Λ, ci ))| ≥ 8 for i = 0, 1, which
implies S ≤ 2|G|/8 = 24(g − 1). If |θ(C(Λ, c0 ))| = 8 then G is a factor group
with presentation
of the group G
Consequently, G = G, as the last has order 96. Thus, here S has genus 2. Further-
more, (c0 c1 )(c2 c1 )(c0 c2 )4 (c1 c2 ) represents in G an element of order 2 and so, in
virtue of Lemma 3.3.1, |θ(C(Λ, c1 ))| = 8. Hence, there exists a Riemann surface of
genus g = 2 with 2|G|/8 = 24 ovals.
Assume therefore that |θ(C(Λ, c0 ))| > 8, which implies that |θ(C(Λ, c0 ))| > 12,
since otherwise (θ(c0 c1 )θ(c0 c2 )4 )3 = 1 and from the defining relations for Λ we
deduce that θ(c1 c2 ) = 1, a contradiction. Henceforth, |θ(C(Λ, c0 ))| ≥ 16.
Now, if |θ(C(Λ, c1 ))| = 8 then G is a factor group of the group G with
presentation
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )3 , (c0 c2 )8 , ((c0 c1 )(c2 c1 )(c0 c2 )4 (c1 c2 ))2 ,
which can be shown to have order 384. Thus, S has genus g = 2, 3 or 5 as |G| =
96(g − 1). We have already shown that there exists a Riemann surface of genus
g = 2 with 24 ovals. If g = 5 then G = G, while (c0 c1 )(c0 c2 )4 represents in G
an element of order 4 and thus, in virtue of Lemma 3.3.1, |θ(C(Λ, c0 ))| = 16. So
there exists a Riemann surface of genus g = 5 with |G|/8 + |G|/16 = 18(g −
1) = 72 ovals. Finally, we shall show that there exists a Riemann surface of genus
g = 3 with 18(g − 1) = 36 ovals. Indeed, one can show that the group G with
presentation
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )3 , (c0 c2 )8 , [((c0 c1 )(c1 c2 ))2 , ((c1 c2 )(c0 c1 ))2 ]
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 51
has order 192. Moreover, (c0 c1 )(c0 c2 )4 and (c0 c1 )(c2 c1 )(c0 c2 )4 (c1 c2 ) represent
in G elements of order 4 and 2, respectively, and thus, in virtue of Lemma 3.3.1,
|θ(C(Λ, c0 ))| = 16 and |θ(C(Λ, c1 ))| = 8. Consequently, the corresponding surface
has |G|/16 + |G|/8 = 18(g − 1) = 36 ovals.
Hence we can assume that |θ(C(Λ, c1 ))| ≥ 12. Indeed, if |θ(C(Λ, c1 ))| ≥ 16
then
2|G|
S ≤ = 24 = 12(g − 1).
16
Thus, let |θ(C(Λ, c1 ))| = 12. Then
|G| |G|
S ≤ + = 28 = 14(g − 1)
16 12
with presentation
and G is a factor group of the group G
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )3 , (c0 c2 )8 , ((c0 c1 )(c2 c1 )(c0 c2 )4 (c1 c2 ))3 ,
which was checked, using the mentioned GAP program, to have order 4896. But
using GAP once more we see that (c0 c1 )(c0 c2 )4 represents in G an element of order
9. Therefore |θ(C(Λ, c0 ))| ≥ 36 and so
|G| |G|
S = + < 21 < 11(g − 1).
12 36
The orders of the possible factor groups of G which may be smooth factors of Λ
are 1632, 288 and 96. However, in a factor group G of G of order 288 the element
4
(c0 c1 )(c0 c2 ) would still represent an element of order 9 and therefore also in this
case we would have S < 11(g − 1). In those factor groups G of G with orders 96
and 1632, the product (c0 c1 )(c0 c2 )4 would represent an element of order 3, which
we already showed to be impossible.
This finishes the discussion of the case m = 8 and so we shall assume now that
m = 10. Here |G| = 60(g − 1) and, as |θ(C(Λ, ci ))| ≥ 8 for i = 0, 1, we have
2|G|
S ≤ = 15(g − 1).
8
So we can assume that g = 2, 5. If |θ(C(Λ, ci ))| > 8 then |θ(C(Λ, ci ))| ≥ 12 for
i = 0, 1, which implies S ≤ 2|G|/12 = 10(g − 1). So let |θ(C(Λ, c0 ))| = 8.
with presentation
Then G is a factor group of the group G
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )3 , (c0 c2 )10 , ((c0 c1 )(c0 c2 )5 )2 .
In this case G has order 240. We have assumed that g = 2, 5 and we shall see
that also g = 3. Otherwise, if g = 3 then |G| = 120. Now c0 c1 and c1 c2 rep-
resent in G elements generating a subgroup H which either equals G or it has
52 3 Counting Ovals of Symmetries of Riemann Surfaces
order 60. However, in the former case Γ would contain an orientation reversing
element, which is impossible. So H has order 60. Clearly c0 c1 and c1 c2 represent
in H elements of order 2 and 3, respectively, whose product is an element of order
10. But then it is easy to show that H contains a normal subgroup of order 2, which
is obviously impossible. If |θ(C(Λ, c1 ))| = 8 then G is a factor group of the group
G with presentation
c0 , c1 , c2 | c20 , c21 , c22 , (c0 c1 )2 , (c1 c2 )3 , (c0 c2 )10 , ((c0 c1 )(c2 c1 )(c0 c2 )5 (c1 c2 ))2
But then, using the defining relations for Λ, one can show as before that θ(c1 c2 ) and
θ(c0 c2 )2 are conjugate, which implies that θ(c0 c2 )6 = 1, a contradiction. Hence,
|θ(C(Λ, c0 ))| ≥ 8 and therefore
|G| 3m
S ≤ = (g − 1) ≤ 9(g − 1)
8 m−6
Therefore, the equality S = 21(g − 1) is equivalent to the fact that G is a factor
with presentation
group of the group G
which can be shown to have order 1008. Now c0 c1 and c1 c2 represent elements
which either equals G
generating a subgroup of G or has order 504. However, in the
former case, Γ would contain an orientation reversing element, which is an absurd.
So G = G and this proves that there exists a Riemann surface of genus g = 7 with
21(g − 1) = 126 ovals.
3.3 Total Number of Ovals of all Symmetries of a Riemann Surface 53
|G| 2m
S ≤ = (g − 1) ≤ 14(g − 1).
12 m−6
with presentation
has order 3, or equivalently, if G is a factor group of the group G
which can be shown to have order 336. As before we claim that G = G and therefore
S has genus 3.
Finally, if |θ(C(Λ, c0 ))| > 12 then |θ(C(Λ, c0 ))| ≥ 16 and this implies
|G|
S ≤ < 11(g − 1).
16
This finishes the proof of the first part of the theorem. Now we will show that the
bound 12(g − 1) is attained for infinitely many values of g.
Let us consider the group Ω = Z2 ∗ Z3 = x, y | x2 , y 3 and let M be the
subgroup of Ω generated by A = [x, y] and B = [x, y −1 ]. A straightforward com-
putation gives
Therefore, B = A2 and so
Clearly, the factor group M/K is a metabelian group of order 2k 2 , provided that k is
even. We claim that if, in addition, k is a multiple of 4 then K is a normal subgroup
of Ω. Indeed, it suffices to show that W x , W y ∈ K for any element W ∈ L. For
−1
W = Ak it is clear that W x = (Ak ) ∈ K and a straightforward computation
shows that
Therefore, if 4 divides k then W y ∈ K. One can deal similarly with the remaining
elements of L.
It is straightforward to check that in the factor group G = Ω/K, the commutator
[A, B] is conjugate to (xy)6 , whilst [A, [A, B]] and [B, [A, B]] are conjugate to
(xy)6 (yx)6 . Therefore, G has the following presentation
and denote by a, b and c the images of c0 , c1 and c2 , respectively, under the canon-
ical projection from Λ onto G = Λ/Γ. Observe that x1 = c0 c1 and x2 = c1 c2
are canonical elliptic generators of Δ of orders 2 and 3, respectively, and so we can
assume that x = ab and y = bc. Finally, (ab)(ac)6 and (ab)(bc)(ac)6 (cb) are ele-
ments of order 2 as both are conjugate to (xy)6 x. Therefore, by (2) of Lemma 1.3.3,
|θ(C(Λ, ci ))| = 8 for i = 0, 1 and so, by Theorem 3.3.2, S has 12(g − 1) ovals.
This completes the proof.
3.4 Total Number of Ovals of a Couple of Symmetries 55
In this section we shall find, using Theorem 3.1.1, a bound for the total number
of ovals of a pair of symmetries of a Riemann surface of genus g, in terms of g
and the order and number of points fixed by their product. The first precedent of
this kind of result goes back to Natanzon [102], who classified pairs of commuting
symmetries. Afterwards, Bujalance and Costa studied in [19] pairs of symmetries
of p-hyperelliptic Riemann surfaces. Some of the results we present here have been
recently showed by Kozłowska-Walania [69].
Let S = H/Γ be a Riemann surface and let θ : Λ → Aut(S) be a smooth
epimorphism with ker θ = Γ. Recall that if a symmetry σ of S and the image
θ(ci ) of a canonical reflection ci of Λ are conjugate then we denote the index
[C(G, θ(ci )) : θ(C(Λ, ci ))] by wi and we call it the contribution of ci to σ.
Lemma 3.4.1. Let Dn = Λ/Γ be a dihedral group of automorphisms of a Riemann
surface S = H/Γ generated by two non-central symmetries σ and τ , and let C =
(n1 , . . . , ns ) be a period cycle of the signature of the NEC group Λ.
(1) If n is odd then the reflections corresponding to C contribute to σ and τ
with at most 2 ovals in total.
(2) If n is even then the reflections corresponding to C contribute to σ and τ
with at most t ovals in total, where t is the number of even link periods if s ≥ 1
and some ni is even, and with at most 2 ovals in total in the remaining cases.
Proof. Let θ : Λ → Dn be the canonical epimorphism. The case of odd n is an
immediate consequence of Theorem 3.1.1, since all canonical reflections c corre-
sponding to the period cycle C are conjugate, the centralizer C(Dn , θ(c)) has order
two and c ∈ C(Λ, c).
Now, for even n, the centralizer of any non-central element of Dn has order 4.
Since ci ∈ C(Λ, ci ), necessarily the contribution wi of θ(ci ) satisfies wi ≤ 2, and
since σ and τ are non-conjugate, we may assume that either s ≥ 2 or s = 1 and n1
is even. If c belongs to two odd link periods then we can suppose that c contributes
neither to σ nor to τ . If c belongs to an even link period n and the product cc
has order n then (cc )n /2 ∈ C(Λ, c). Now θ((cc )n /2 c) = 1, because ker θ is a
Fuchsian group, and therefore θ(C(Λ, c)) has order 4.
We give now a new proof of an upper bound, first estimated by Bujalance, Costa
and Singerman in [21], for the total number of ovals of two symmetries of a Riemann
56 3 Counting Ovals of Symmetries of Riemann Surfaces
surface. The original proof used a deep method due to Hoare [60], while the proof
here is based on Theorem 3.1.1 and follows ideas of Kozłowska-Walania in [69].
Theorem 3.4.2. Let σ and τ be two symmetries of a Riemann surface S of genus g
whose product has order n. Then
⎧
⎪ 2(g − 1)
⎨ + 4 if n is odd;
σ + τ ≤ n
⎪
⎩ 4g + 2 if n is even.
n
Proof. Let t = σ + τ denote the total number of ovals of σ and τ and let
Dn = σ, τ . Hence Dn = Λ/Γ for some surface Fuchsian group Γ and an NEC
group Λ with signature
where n1 , . . . , n are odd and each period cycle Ci = (ni1 , . . . , nisi ) with either
si ≥ 2 or si = 1 and ni1 is even. Let s = s1 + · · · + sk .
Observe first that either r = 0 or s + = 1. Otherwise s(Λ) would have a
unique link period, say n0 , and so, by Corollary 1.2.7, the signature of the canonical
Fuchsian group of Λ would be s(Λ+ ) = (h ; n0 ). As Λ+ /Γ = Zn , the relation
x1 [a1 , b1 ] · · · [ah , bh ] = 1 in Λ+ would give θ(x1 ) = 1 for the canonical smooth
epimorphism θ : Λ → Dn , which is impossible because Γ = ker θ is a surface
Fuchsian group.
Let us first analyze the case of odd n. We know that t ≤ 2k + 2 + 2m, by
Lemma 3.4.1, and so
2π(g − 1)
= Area(Λ) ≥ 2π(k + + m − 2) ≥ π(t − 4),
n
where η = 2 if the sign of s(Λ) is “+” and η = 1 otherwise. Using that either r = 0
or s + = 1, it is easy to see that in all cases (1)–(5), we have
r +m 3
ηh + k + + + ≥ .
2 2 3 2
Thus
2π(g − 1) t
= Area(Λ) ≥ π −1 ,
n 2
which gives
4(g − 1) 4g
t≤ +2< + 2.
n n
Therefore we can assume that Λ is an NEC group with signature
Observe now that for n = 2 the desired bound obviously holds in virtue of Harnack–
Weichold theorem. So let n = 2. The images of the elements xi = ci−1 ci generate
Zn and they satisfy the relation x1 · · · xs = 1. So, if one of the link periods is odd,
then either there is another odd link period or a link period ni ≥ 6. In the first case
we have
2π(g − 1) s−2 2 π 4
= Area(Λ) ≥ 2π −1 + + = s−2− .
n 4 3 2 3
4g 4 4 4g
t≤s−2≤ − + < + 2.
n n 3 n
In the second case one gets, similarly,
4g 4 4g
t≤s−1≤ − +2< + 2.
n n n
Therefore, we can assume that n ≥ 4 and all link periods of s(Λ) are even. If there
are more than three link periods ni ≥ 4 then
2π(g − 1) s−4 3 π(s − 2)
= Area(Λ) ≥ 2π −1 + + ≥
n 4 2 2
and, consequently,
4g 4 4g
t=s≤ − +2< + 2.
n n n
58 3 Counting Ovals of Symmetries of Riemann Surfaces
If all link periods are equal to 2 except n1 and n2 then n1 = n2 = n and so the
bound holds.
Therefore we can assume that all link periods are equal 2 except ni , ni , ni , and
that 4 ≤ ni ≤ ni ≤ ni . Now, if ni ≥ 6 then
2π(g − 1) s−3 5 π(s − 2)
= Area(Λ) ≥ 2π −1 + + ≥
n 4 4 2
4g 4 4g
t≤s≤ − +2< + 2.
n n n
On the other hand, if ni = 4 then also n = 4 and arguing as above we get
4g 4 4g
t≤ − +3= + 2.
n n n
Consequently,
4g 4 3 4 4g 3 3 4g
t≤ − + + ≤ + + ≤ + 2.
n n 2 ρ n ρ 2 n
Our next result shows that the precedent upper bounds are sharp in all cases.
Theorem 3.4.3. The upper bounds occurring in Theorem 3.4.2 are attained for all
arithmetically admissible values of g and n.
Proof. Let first n be an odd divisor of g − 1. Consider an NEC group Λ with
signature
s(Λ) = (0; +; [−]; {(−), (g−1)/n
. . . . . . .+. .2, (−)})
As every period cycle contributes with one oval to each of the symmetries σ and τ ,
we deduce that
2(g − 1)
σ + τ = + 4.
n
3.4 Total Number of Ovals of a Couple of Symmetries 59
Now let n be an even divisor of 4g. Consider an NEC group Λ with signature
• If s is odd then
Theorem 3.4.2 provides bounds of the form [2(g − 1)/n] + 4 and [4g/n] + 2,
where [ · ] denotes the integer part function, for the total number of ovals of two
symmetries. We now study such bounds in detail. For instance, we show that the
first bound is attained only for n dividing g − 1. As announced above the following
Theorems 3.4.4, 3.4.5 and 3.4.7 are due to Kozłowska-Walania [69].
Theorem 3.4.4. Let σ and τ be two symmetries of a genus g Riemann surface S,
whose product στ has odd order n. If n does not divide g − 1 then
2(g − 1)
σ + τ ≤ + 3.
n
Even more, for each odd integer n ≥ 3 this bound is attained for infinitely many
values of g.
Proof. Let t = σ + τ denote the total number of ovals of σ and τ and write
σ, τ = Dn = Λ/Γ for some surface Fuchsian group Γ and an NEC group Λ with
signature
s(Λ+ ) = (h ; n0 ); but there is no smooth epimorphism from a Fuchsian group with
this signature onto the cyclic group Zn , see the proof of Theorem 3.4.2.
So Λ has either a proper period or at least two link periods. Then
2π(g − 1) 1
= Area(Λ) > 2π k++m−2+ ≥ π(2(k++m) − 3) ≥ π(t−3)
n 2
Then ker θ is a surface Fuchsian group and the Riemann surface S = H/ ker θ has
genus n(m + 1) by the Hurwitz–Riemann formula. Theorem 3.1.1 yields that both
σ and τ have m + 2 ovals and so
2(g − 1)
σ + τ = 2m + 4 = + 3.
n
In contrast to the previous theorem, the bound [4g/n] + 2 for even n in
Theorem 3.4.2 without any divisibility conditions on the pair (n, g) cannot be
improved.
Theorem 3.4.5. For each even integer n > 4 there are infinitely many values of g
for which n does not divide 4g and such that there exists a Riemann surface of genus
g admitting two symmetries, whose product has order n, and with [4g/n] + 2 ovals
in total.
Proof. Let Λ be an NEC group with signature
and which maps the reflections corresponding to the unique non-empty period cycle
of s(Λ) alternatively to τ and (στ )n/2−1 σ. The orbit space S = H/ ker θ is a
3.4 Total Number of Ovals of a Couple of Symmetries 61
σ + τ = 2g + 2 − q − q ≥ g + 3.
Suppose that σ and τ do not commute. Then the order n of στ is greater than 2. By
Theorem 3.4.2 we get, for even n,
4g
g + 3 ≤ σ + τ ≤ + 2 ≤ g + 2,
n
a contradiction. For odd n,
2(g − 1) 2(g − 1)
g + 3 ≤ σ + τ ≤ +4≤ +4
n 3
and so g ≤ 1, which is not our case.
Proof. The exceptional case was considered by Natanzon [108], who proved that
two symmetries with g > 2 and 1 ovals, respectively, always commute.
On the other hand, for g = 2 and {q, q } = {1, 2}, we can choose n = 8
in Theorem 3.4.2 to obtain a Riemann surface of such genus 2 with two non-
commuting symmetries having one and two ovals.
So in what follows let {q, q } = {1, g} and we assume that q ≤ q . We define
s = g − q, s = g − q and we distinguish several cases according to the value of
q + q − g (mod 4).
62 3 Counting Ovals of Symmetries of Riemann Surfaces
and θ(x) = στ , θ(e) = τ σ. Again, the genus g surface H/ ker θ admits two non-
commuting symmetries σ and τ having g + 1 − q and g + 1 − q ovals, respectively.
If q + q − g ≡ 1 (mod 4) and g < q + q − 1 consider an NEC group Λ with
signature
s(Λ) = (h; −; [2, 4]; {(2, . s. ., 2, 4, 2, .s. ., 2, 4)})
Once more the genus g surface S = H/ ker θ admits two non-commuting symme-
tries σ and τ having g + 1 − q and g + 1 − q ovals, respectively.
Chapter 4
Symmetry Types of Some Families
of Riemann Surfaces
−1
σ1 : Σ → Σ ; z → z̄ ; σ2 : Σ → Σ ; z → .
z̄
Our goal in this section is to prove that each symmetry of Σ is conjugate either to
σ1 or to σ2 . Before doing this, we recall without proof some well known facts about
semilinear endomorphisms of complex vector spaces.
Definition and Proposition 4.1.1. Let E be a complex vector space.
(1) A map f : E → E is a semilinear endomorphism if
n
f (uj ) = aij ui , for 1 ≤ j ≤ n.
i=1
n
uj = cij vi , for 1 ≤ j ≤ n.
i=1
that satisfies MfA (E) = A, where E is the standard basis of C2 . Since σ 2 is the
identity map and det(A) = 1, either fA2 or −fA2 is the identity endomorphism. This
implies, by part (4) in Proposition 4.1.1, that AĀ = ±I. We now distinguish two
cases:
Case 1: AĀ = I, that is, fA2 is the identity. Since fA is not a dilatation, there exists
a vector u ∈ C2 such that {u, fA(u)} is a basis of C2 . Thus also
c11 z + c12
ϕ(z) = .
c21 z + c22
The equality C = AC̄ means that the antianalytic automorphism ϕ◦σ1 , with matrix
C, coincides with the antianalytic automorphism σ ◦ ϕ, with matrix AC̄. Therefore
ϕ−1 ◦ σ ◦ ϕ = σ1 , that is, σ is conjugate to complex conjugation.
Case 2: AĀ = −I, that is, −fA2 is the identity. Any non-zero vector u ∈ C2 is not
an eigenvector of fA . Otherwise fA (u) = λu for some λ ∈ C and this would imply
satisfies MfA (B1 ) = C −1 AC̄ by part (3) in Proposition 4.1.1. Hence, the analytic
automorphism
c11 z + c12
ϕ(z) =
c21 z + c22
of Σ satisfies σ(ϕ(z)) = ϕ(σ2 (z)). So in this case σ is conjugate to the antipo-
dal map.
Finally, let Δ be the closed disk of radius 1 centered at the origin of C. It is easy
to see that the map
i,
[(z + i)/(iz + 1)]σ1 if z =
Δ → Σ/
σ1 ; z →
∞ if z = i,
is a homeomorphism.
On the other hand, let S2 be the unit sphere in R3 and let πN : S2 → Σ be
the extension of the stereographic projection from the north pole (0, 0, 1) ∈ S2 that
maps this point to ∞. Let τ = πN −1
◦ σ2 ◦ πN . Obviously the orbit spaces S2 /
τ
and Σ/
σ2 are homeomorphic. Moreover, for each point p ∈ S2 let us denote by
[p]∼ ∈ P2 (R) the associated projective point, and let [p]τ be its class modulo τ . A
straightforward computation shows that the map
P2 (R) → S2 /
τ ; [p]∼ → [p]τ
is a homeomorphism, and so Σ/
σ2 is the real projective plane.
The closed disk is an orientable surface with one boundary component, that is,
the species of the symmetry σ1 is +1. The real projective plane is a non-orientable
surface with empty boundary; that is, σ2 has species 0. Consequently, Theorem 4.1.2
yields the following.
Corollary 4.1.3. The symmetry type of the Riemann sphere is {+1, 0}.
Let S be a Riemann surface of genus one. It is well known (see [92, Chap. II]) that,
up to analytic isomorphism, S is the orbit space S = C/L where
f : S1 → S2 ; z + L1 → (az + b) + L2 .
(2) The index [L2 : aL1 ] of aL1 as a subgroup of L2 is the degree of f . In partic-
ular, f is an isomorphism if and only if L2 = aL1 .
Proof. (1) The map f is unramified because the Euler characteristic of both S1 and
S2 equals 0. In particular, f is a local homeomorphism. Let πi : C → Si ; z →
z + Li be the corresponding universal covering, for i = 1, 2. The domain of
the covering f ◦ π1 : C → S2 is simply connected and so it is isomorphic, as a
covering, to π2 . Therefore there exists a holomorphic function F : C → C such
that
π2 ◦ F = f ◦ π1 .
We shall prove that F (z) = az + b where a ∈ C \ {0} and b ∈ C satisfies
f (0 + L1 ) = b + L2 . In such a case, we will have
F : C → C ; z → F (z + ) − F (z).
k
L2 = (zi + aL1 ).
i=1
f : S → S ; z + L → (az + b) + L
Proposition 4.2.4. (1) Every torus C/L1 where L1 is a square lattice is isomor-
phic to C/L(i) where L(i) = Z + Zi.
(2) Every torus C/L2 where L2 is a hexagonal lattice is isomorphic to C/L(eπi/3 )
where L(eπi/3 ) = Z + Zeπi/3 .
Proof. (1) Let us write L1 = Zω1 + Zω2 . We may assume, after relabeling the
generators if necessary, that ω2 = iω1 . Then f : C/L1 → C/L(i) ; z + L1 →
ω1−1 z + L(i) is a well defined isomorphism. The proof of part (2) is analogous and
we omit it.
Another consequence of Proposition 4.2.1 is the following (see [3, Chap. 9] and
[92, Chap. III]): each torus is analytically isomorphic to S = C/L(ω), where
L(ω) = Z + Zω and the complex number ω can be chosen to satisfy the following
conditions
|ω| ≥ 1 if Re(ω) ≥ 0,
Im(ω) > 0, −1 < 2Re(ω) ≤ 1 with (4.1)
|ω| > 1 if Re(ω) < 0,
where Re(ω) and Im(ω) stand, respectively, for the real and imaginary parts of ω.
In the sequel we will assume that any lattice L is of the form L = L(ω) where ω
satisfies these conditions.
The next proposition shows that, as said above, there are very few automorphisms
of the complex torus C/L which fix 0 + L.
f : z + L → az + L,
Proof. We may assume that L = L(ω) for some ω satisfying conditions (4.1). Let
us determine the set
Thus |r| = |s| and rs = 0, a contradiction. So ω = eiθ with π/3 ≤ θ < 2π/3 and
so −1 < 2 cos θ ≤ 1. Observe that rs ≤ 0 since otherwise 1 ≥ r2 +2rs cos θ+s2 >
r2 − rs + s2 = (r − s)2 + rs ≥ rs ≥ 1, a contradiction. So rs ≤ 0 and hence
The unique integer solutions (r, s) with s = 0 to this inequality are (0, ±1) and
±(1, −1). They correspond to ±ω and ±(1 − ω). However,
Let L be a lattice of the form L = L(ω) = Z+Zω where ω satisfies conditions (4.1).
Our first goal is to characterize the symmetric tori C/L(ω) in terms of the complex
number ω. For each symmetric torus we will also determine its symmetries and the
species of each of them.
The next proposition can be seen as the counterpart, for antianalytic maps, of
Proposition 4.2.1. To prove it we adapt some of the already used arguments.
σ : S → S ; z + L → (az̄ + b) + L.
π ◦ g = (π ◦ f ) ◦ κ = σ ◦ (π ◦ κ) = σ ◦ p,
Consider the lattice K = κ(L) and let T be the torus T = C/K. The covering
p : C → S factorizes throughout T as p = κ̄ ◦ π where
π : C → T ; z → z + K and κ̄ : T → S ; z + K → z̄ + L.
τ ◦ π = σ ◦ (κ̄ ◦ π ) = σ ◦ p = π ◦ g.
gk : C → C ; z → g(z + k) − g(z),
σ : S → S ; z + L → z̄ + L
76 4 Symmetry Types of Some Families of Riemann Surfaces
σ : S → S ; z + L → ω z̄ + L
because ω ω̄ = 1 ∈ L.
(1) Then
f ◦ σ ◦ f −1 : S → S ; z + L → (ac2 z̄ + bc + d − ac2 d)
¯ + L.
f ◦ σ ◦ f −1 (z + L) = (az̄ + bc + d − ad)
¯ + L.
To begin with, we classify the symmetries of those tori C/L(ω) with |ω| > 1.
Proposition 4.2.10. Let ω ∈ C satisfy conditions (4.1) with |ω| > 1 such that
Re(ω) = 0 or Re(ω) = 1/2. Let L = Z + Zω and S = C/L.
σ1 : S → S ; z + L → z̄ + L; σ2 : S → S ; z + L → (z̄ + 1/2) + L;
The surfaces S/
σ1 and S/
σ3 are closed annuli, while S/
σ2 and S/
σ4
are Klein bottles.
4.2 Symmetry Types of Tori 77
(2) If Re(ω) = 1/2 then each symmetry of S is conjugate to one of the following:
σ1 : S → S ; z + L → z̄ + L; σ3 : S → S ; z + L → −z̄ + L.
Both surfaces S/
σ1 and S/
σ3 are Möbius strips.
(3) The symmetries above are pairwise non-conjugate.
Proof. Given a symmetry σ of S there exist, by part (3) in Proposition 4.2.7, com-
plex numbers a, b with a ∈ {1, −1} such that
σ : S → S ; z + L → (az̄ + b) + L.
1
f ◦ σ ◦ f −1 : z + L → az̄ + ab̄ + b + L.
2
We may assume, after translating by an element of L, that (ab̄ + b)/2 lies in the
fundamental parallelogram RL = {r + sω : 0 ≤ r, s < 1} of L. Since ab̄ + b
belongs to L (by part (1) in Proposition 4.2.7), we see that ab̄ + b is one of the four
vertices of RL , that is,
ab̄ + b ∈ {0, 1, ω, 1 + ω}.
We distinguish cases according to whether Re(ω) = 0 or Re(ω) = 1/2.
(1) Re(ω) = 0. If a = 1 then ab̄ + b = 2Re(b), which is a real number. So
ab̄ + b = 0 or 1. In the first case, σ is conjugate to σ1 ; in the second one, σ is
conjugate to σ2 .
If a = −1 then ab̄ + b = 2Im(b)i, which is purely imaginary. So ab̄ +
b = 0 or ω. In the first case σ is conjugate to σ3 ; in the second one, σ is
conjugate to σ4 .
Next we must determine the topology of the orbit spaces S/
σi for 1 ≤ i ≤
4. Since S has genus one, each surface S/
σi is either an annulus or a Möbius
strip or else a Klein bottle. Therefore the topology of each S/
σi is determined
by the number (two, one or zero, respectively) of its boundary components, that
is, by the number of ovals of the fixed point set Fix(σi ) of each symmetry σi .
Let z = r+sω with 0 ≤ r, s < 1 be a point of the fundamental parallelogram
RL . It is straightforward to check that a point z + L is fixed by σ1 : z + L →
z̄ + L if and only if 2sω ∈ L, that is, 2s ∈ Z. But s ∈ [0, 1) and so either
s = 0 or s = 1/2. This means that Fix(σ1 ) consists of the projection under
the covering map π : C → S of the two horizontal segments [0, 1) × {0} and
[0, 1) × {ω/2} of R. It is clear that the projections are two disjoint ovals in S;
so Fix(σ1 ) has two ovals and S/
σ1 is indeed an annulus.
An analogous argument shows that Fix(σ3 ) consists of the projection under
π of the two vertical segments {0} × [0, ω) and {1/2} × [0, ω). The projections
are also disjoint and therefore S/
σ3 is an annulus.
78 4 Symmetry Types of Some Families of Riemann Surfaces
On the other hand, both σ2 and σ4 are fixed point free and so both S/
σ2
and S/
σ4 are Klein bottles.
(2) Re(ω) = 1/2. If a = 1 then the same argument as above shows that σ is con-
jugate to σ1 or σ2 . But now σ1 and σ2 are conjugate. Indeed, the automorphism
g : S → S ; z + L → (z + Im(ω)i/2) + L satisfies g ◦ σ2 ◦ g −1 = σ1 , as is
easy to check.
If a = −1 then ab̄ + b = 2Im(b)i ∈ iR. As Re(ω) = 1/2, the unique
vertex of the fundamental parallelogram RL that lies in the imaginary axis is 0.
So ab̄ + b = 0 and therefore σ is conjugate to σ3 : z + L → −z̄ + L.
Next we show that S/
σ1 and S/
σ3 are Möbius strips, or equivalently,
their boundaries are connected and non-empty. We study just the first case. Let
us write z = r + sω with 0 ≤ r, s < 1. It is straightforward to check that a
point z + L is fixed by σ1 : z + L → z̄ + L if and only if s(ω − ω̄) ∈ L, that
is, 2sIm(ω)i ∈ L. Since 2Im(ω)i = 2ω − 1 belongs to L and there is no point
of L in the open segment joining 0 and 2Im(ω)i, we see that s = 0. Therefore
Fix(σ1 ) consists of the projection under π : C → S of the horizontal segment
[0, 1) × {0} of the fundamental parallelogram R. This shows that S/
σ1 is
indeed a Möbius strip.
(3) We have already proved that σ1 (respectively σ3 ) in (1) is not conjugate to
σ2 (respectively σ4 ) because S/
σ1 and S/
σ3 are closed annuli and S/
σ2
and S/
σ4 are Klein bottles. On the other hand, it follows from part (2) in
Proposition 4.2.9 that σ1 and σ3 in (1) and (2) are not conjugate. By the same
reason σ2 and σ4 in (1) are not conjugate.
The classification of the symmetries of the tori C/L(ω) where L(ω) is neither
square nor hexagonal and with |ω| = 1 is the following.
Proposition 4.2.11. Let ω ∈ C satisfy conditions (4.1) with |ω| = 1 and such that
0 < Re(ω) < 1/2. Let L = Z + Zω and S = C/L. Then each symmetry of S is
conjugate to one of the following:
σ1 : S → S ; z + L → ω z̄ + L; σ2 : S → S ; z + L → −ω z̄ + L,
Proof. The last part is an immediate consequence of part (2) in Proposition 4.2.9.
For the first one we repeat the same arguments as in the beginning of the proof of
Proposition 4.2.10 to get that any symmetry of S is conjugate to
1
σ : z + L → az̄ + ab̄ + b + L where ab̄ + b ∈ {0, 1, ω, 1 + ω}.
2
Here a = ±ω by part (6) in Proposition 4.2.7. Let us write ω = eiθ with π/3 < θ <
π/2. We distinguish cases according to whether a = ω or a = −ω.
4.2 Symmetry Types of Tori 79
If a = ω then
ab̄ + b = eiθ b̄ + b = eiθ/2 eiθ/2 b̄ + e−iθ/2 b ∈ Reiθ/2 .
σ : z + L → ω z̄ + L or σ : z + L → (ω z̄ + (1 + ω)/2) + L,
respectively. But these two symmetries are conjugate via the automorphism z+L →
z + (ω − 1)/4 + L, as is easy to check.
If a = −ω then, writing β = b̄eiθ/2 , we get
ab̄ + b = −eiθ b̄ + b = −eiθ/2 b̄eiθ/2 − be−iθ/2
= ei(π+θ/2) β − β̄ = ei(π+θ/2) 2Im(β)i ∈ Rei(θ−π)/2 .
σ(z + L) = −ω z̄ + L = σ2 (z + L).
To finish this section we classify the symmetries of the special tori, that is, those
tori S = C/L where the lattice L is either square or hexagonal. By Proposition 4.2.4
the lattice L may be assumed to be L = Z + Zi if L is a square lattice and L =
Z + Zeπi/3 if L is a hexagonal lattice.
Moreover, S/
σ1 is a closed annulus, S/
σ2 is a Klein bottle and S/
σ3 is
a Möbius strip. In particular, the symmetries σ1 , σ2 and σ3 are pairwise non-
conjugate.
80 4 Symmetry Types of Some Families of Riemann Surfaces
Proof. Repeating the same arguments as in the beginning of the proof of Proposi-
tion 4.2.10 we get that any symmetry of S is conjugate to
1
σ : z + L → az̄ + ab̄ + b + L where ab̄ + b ∈ {0, 1, i, 1 + i}.
2
σ1 : S → S ; z + L → z̄ + L; σ2 : S → S ; z + L → ω z̄ + L.
Proof. The same arguments as in the beginning of the proof of Proposition 4.2.10
yields that any symmetry of S is conjugate to
1
z + L → az̄ + ab̄ + b + L where ab̄ + b ∈ {0, 1, ω, 1 + ω}.
2
4.2 Symmetry Types of Tori 81
Hyperelliptic Riemann surfaces have attracted the attention of geometers since long
time ago. The study of their symmetries began with the works of Natanzon [98, 99]
at the end of the seventies of the past century, and Maskit [78] in 1995. They both
restricted themselves to the case of symmetries whose number of ovals is maximal.
Later on, the authors of this monograph classified all symmetries of each hyper-
elliptic surface without any assumption on the number of their ovals, [14]. This
work is too extensive even to survey it here. Hence we present in this section the
main ingredients to understand why the computation of the symmetry types of hy-
perelliptic surfaces is achievable. We will also explain an illustrative example in
detail. It is worth mentioning that the results are also described in an algebraic way.
That is, the surfaces are given by means of defining polynomial equations and the
symmetries are described as the composite of complex conjugation with birational
transformations.
A Riemann surface is called cyclic p-gonal if it is a cyclic p-covering of the
Riemann sphere. The 2-gonal surfaces are, therefore, the hyperelliptic ones. Sym-
metries on trigonal and cyclic p-gonal surfaces, for p an odd prime, have been
studied by Costa and Izquierdo in [35, 36]. The proofs in these two articles are
based on the theory of NEC groups, in contrast to the more geometric approach we
follow here with hyperelliptic surfaces.
We first recall the notion of hyperellipticity.
Definition 4.3.1. A Riemann surface S of genus g ≥ 2 is hyperelliptic if any of the
following equivalent conditions holds:
S = {(x, y) ∈ C2 : y 2 = PS (x)},
4.3 Symmetry Types of Hyperelliptic Riemann Surfaces 83
where
πS : S → Σ ; (x, y) → x
is a meromorphic function of degree 2. Its 2g + 2 branch points are thus the roots of
PS together with ∞ if δ = 0. They constitute what we call (by abuse of language)
the branch point set of S, which we denote by BS . With the above notations,
{e1 , . . . , e2g+2 } if δ = 1,
BS =
{e1 , . . . , e2g+1 , ∞} if δ = 0.
f
S - T
πS πT
? ?
fˆ-
Σ Σ
ax + b
m(x) = with {a, b, c, d} ⊂ C and det m := ad − bc = 0
cx + d
be a Möbius transformation such that m(BS ) = BT . Then the formulae of its lift-
ings, say f and f ◦ ρS , depend on whether ∞ ∈ BS or not and whether m fixes ∞
or not. These formulae, as they appear in [30], are the following.
(3) If ∞ ∈
/ BS and m(∞) = ∞ then
a g+1
a b
f (x, y) = x+ , y· .
d d d
(4) If ∞ ∈
/ BS and m(∞) = ∞ then
ax + b y · cg+1
f (x, y) = , PT (a/c) .
cx + d (cx + d)g+1
Note that, algebraically, AutΣ (S) is nothing else but the factor group Aut(S)/
ρ.
4.3 Symmetry Types of Hyperelliptic Riemann Surfaces 85
Apart from the analytic methods used in Sects. 4.1 and 4.2 to compute the symmetry
types of the Riemann sphere and tori, respectively, for surfaces of higher genus the
most common method is the combinatorial study of NEC groups. However, the nice
properties of hyperelliptic surfaces allow to develop a geometric method which turns
out to be easier than the combinatorial one. The goal of this subsection is to explain
this geometric approach.
Assume that the hyperelliptic surface S admits a symmetry σ. Our purpose here
is to compute the species of σ in terms of both its formula and the equation of S.
This is summarized in Theorems 4.3.4 and 4.3.5 below, which are based on results
in [57, Sect. 6], adapted to our point of view.
In the same way as analytic isomorphisms induce Möbius transformations, the
symmetry σ : S → S induces an antianalytic Möbius transformation σ : Σ → Σ,
which also has order 2. We proved in Theorem 4.1.2 that if σ fixes points then it is
conjugate to complex conjugation x → x̄; otherwise it is conjugate to the antipodal
map x → −1/x̄.
is complex conjugation. In this case the polynomial PS
Let us assume first that σ
defining the surface S has real coefficients since its roots are permuted by σ . Then
one of the liftings of σ is σ : (x, y) → (x̄, ȳ) and so its fixed point set consists
exactly of the points with real coordinates of the surface S = {y 2 = PS (x)}. It is
then easy to see that if PS has 2k > 0 real roots then Fix(σ) consists of exactly k
ovals. Moreover, the separating character of Fix(σ) depends on the number of real
roots of PS or, more precisely, since ∞ may be a branch point, on the number of
branch points of S fixed by σ . Indeed, if 2k is the number of branch points fixed by
then
σ ⎧
⎪
⎪ g + 1 if k = g + 1;
⎨
−k if 0 < k < g + 1;
sp(σ) =
⎪
⎪ 1 if k = 0 and g is even;
⎩
2 if k = 0 and g is odd.
is σ ◦ ρ : (x, y) → (x̄, −ȳ). If σ
The other lifting of σ fixes no branch point
then PS is always positive on R; it follows that in this case Fix(σ ◦ ρ) is empty,
that is, sp(σ ◦ ρ) = 0. If σ
fixes some branch point we claim that the species of
σ ◦ ρ coincides with the species of σ. To show this we use the obvious fact that if
α and β are symmetries of two different Riemann surfaces S and T respectively,
and f ◦ α = β ◦ f for some isomorphism f : S → T , then the species of α and β
coincide.
86 4 Symmetry Types of Some Families of Riemann Surfaces
Proof. Let r be the greatest real root of PS and let m be the real Möbius transforma-
tion m : Σ → Σ ; x → −1/(x − r). Consider the hyperelliptic Riemann surface T
whose branch point set is BT = m(BS ). Clearly, γ : T → T ; (x, y) → (x̄, ȳ) is a
symmetry of T . Denoting by f a lifting of m it is easy to check that f ◦ σ ◦ ρ = γ ◦ f
and so sp(σ ◦ ρ) = sp(γ ). It remains to see that sp(γ) = sp(σ). Since both sym-
metries γ and σ have the same formula (x, y) → (x̄, ȳ), it follows from the above
that sp(γ) = sp(σ) if and only if the number of branch points of T fixed by γ
coincides with the number of branch points of S fixed by σ . But this is clear be-
◦ m−1 =
cause m ◦ σ γ and so the branch points of T fixed by γ are the images by
m of the branch points of S fixed by σ .
Theorem 4.3.4. Suppose that the branch point set of the hyperelliptic surface S
. Then S admits the symmetries
of genus g is preserved by complex conjugation σ
defined by
.
Let 2k be the number of branch points of S fixed by σ
(1) If k > 0 then
g + 1 if k = g + 1,
sp(σ) = sp(σ ◦ ρ) =
−k if k < g + 1.
(2) If k = 0 then
1 if g is even,
sp(σ) = sp(σ ◦ ρ) = 0.
2 if g is odd;
Remark 4.3.6. Observe that in case (3) we do not know which one is the lifting with
is complex conjugation (see Theorem 4.3.4). When
non-zero species, except if σ
dealing with polynomial equations we can exhibit explicit formulae of both liftings
and so we can decide which one fixes points.
4.3.2 An Example
Let S be a symmetric hyperelliptic Riemann surface of genus g and let AutΣ (S)
be the group of Möbius transformations preserving its branch point set BS . This is
a finite group of isometries of the Riemann sphere and so it has a nice geometric
interpretation. For instance, AutΣ (S) can be the group of (orientation preserving or
reversing) isometries of a regular cube, and it is easy to describe how its elements
permute the 2g + 2 branch points of S. This allows us to describe the distribution of
the branch points on the sphere and, in particular, to find out how many of them are
fixed by a given symmetry, say σ . Once this is achieved, we can use Theorem 4.3.5
to compute the species of its liftings σ and σ ◦ ρ (provided they are involutions).
The distribution of the branch points of S also provides an algebraic equation of
S. With this equation at hand, the formulae of the automorphisms of S can be explic-
itly computed as liftings of the Möbius transformations in AutΣ (S). This way we
get a presentation by means of generators and defining relations of the full group
88 4 Symmetry Types of Some Families of Riemann Surfaces
Theorem 4.3.7. Let S be a hyperelliptic Riemann surface of genus g such that the
induced group AutΣ (S) of Möbius transformations is dihedral of order 2n ≥ 4
generated by two reflections σ1 and σ 2 . Assume that the two fixed points of the
2 ◦ σ
rotation σ 1 are branch points of S. Then there exist non-negative integers r, p1
and p2 with 2r + p1 + p2 = 2g/n ≥ 3 such that S admits an algebraic equation of
the form
r
p1
p2
y2 = x · (xn − wjn )(xn − w̄jn ) · (xn − λnj ) · (xn + μnj ),
j=1 j=1 j=1
where the roots of the polynomial on the right hand side are simple and λnj and μnj
are positive real numbers for all j. Moreover, the full group Aut(S) of automor-
phisms of S is dihedral of order 4n generated by the symmetries
σ1 : S → S ; (x, y) → (x̄, ȳ) and σ2 : S → S ; (x, y) → x̄e2πi/n , ȳeπi/n .
2 : Σ → Σ ; x → x̄e2πi/n .
1 : Σ → Σ ; x → x̄ and σ
σ
In particular, the branch point set BS of S consists of the orbits of points lying in
this fundamental region.
(1) If w lies in the interior of F then its orbit under the action of
2 consists
σ1 , σ
of the 2n roots of the polynomial (xn − wn )(xn − w̄n ); observe that in this
case wn is a complex number with positive imaginary part.
(2) If λ = 0, ∞ is fixed by σ1 (and hence it lies in the boundary of F) then its orbit
consists of the n roots of the polynomial xn − λn ; observe that λn is a positive
real number.
(3) If ν = 0, ∞ is fixed by σ2 (and hence it lies in the boundary of F) then its orbit
consists of the n roots of the polynomial xn − ν n ; observe that ν n is a negative
real number.
(4) Finally, if α = 0 or ∞ then its orbit is {α} itself.
Let r be the number of branch points lying in the interior of F and, for i = 1, 2, let
pi be the number of branch points lying in F ∩ Fix( σi ) and different from 0 and ∞.
Since we are assuming that both 0 and ∞ are also branch points, the total number
of branch points of S is 2rn + p1 n + p2 n + 2. This yields the equality 2r + p1 +
p2 = 2g/n in the statement of the theorem. In addition, the above also shows that a
defining algebraic equation y 2 = PS (x) of S has the form given in the statement of
the theorem, because the roots of PS (x) are precisely the finite branch points of S.
The condition 2r + p1 + p2 ≥ 3 is necessary to assure that AutΣ (S) has no
more automorphisms than those in
2 and so AutΣ (S) is indeed dihedral of
σ1 , σ
order 2n. In fact, if 2r + p1 + p2 = 1 or 2 then there exist Möbius transformations
in AutΣ (S) \
2 . Explicitly, if p1 = 1 and r = p2 = 0 then the reflection
σ1 , σ
defined as Σ → Σ ; x → λ21 /x̄ belongs to AutΣ (S) \
2 ; the case p2 = 1
σ1 , σ
and r = p1 = 0 is analogous. In case r = 1 and p1 = p2 = 0 the same happens
to the reflection Σ → Σ ; x → |w1 |2 /x̄. In case r = 0, p1 = p2 = 1 the trans-
formation to be considered is Σ → Σ ; x → λ1 μ1 /x and finally, in case p1 = 2
and r = p2 = 0 (the case p2 = 2 and r = p1 = 0 is analogous) the transformation
Σ → Σ ; x → λ1 λ2 /x works.
With the algebraic equation of S at hand it is immediate to check that both maps
σ1 : S → S ; (x, y) → (x̄, ȳ) and σ2 : S → S ; (x, y) → (x̄ · e2πi/n , ȳ · eπi/n )
1 and σ
are liftings of σ 2 respectively. Since σ1 and σ2 have order two and their
composite σ2 ◦ σ1 : S → S ; (x, y) → (x · e2πi/n , y · eπi/n ) has order 2n, we see
that they generate a dihedral group of order 4n which therefore coincides with the
full group Aut(S). In particular, they are representatives of the unique two conju-
gacy classes of symmetries of Aut(S). Let us compute their species. To do this we
use Theorem 4.3.5 and for that we need to calculate the number of branch points
fixed by each σi for i = 1, 2. We can use either the defining algebraic equation of S
90 4 Symmetry Types of Some Families of Riemann Surfaces
or the geometric description of how the branch points are distributed in the above
fundamental region F. Let us choose this last option.
Assume first that n is even. Then the branch points fixed by σ 1 are 0, ∞, the
p1 branch points lying in F ∩ Fix( σ1 ), say λ1 , . . . , λp1 , and the images of these
σ2 ◦ σ
last under the rotation ( 1 )n/2 : Σ → Σ ; x → −x, that is, −λ1 , . . . , −λp1 .
This makes a total of 2p1 + 2 branch points lying on Fix( σ1 ). Hence, according to
Theorem 4.3.5, we have
−(p1 + 1) if p1 + 1 < g + 1;
sp(σ1 ) =
g+1 if p1 + 1 = g + 1.
Observe that this last equality holds just for n = 2 and p2 = 0; in fact, if p1 = g then
the equality 2r+p1 +p2 = 2g/n becomes 2r+p2 = g(2/n−1) which forces n = 2
and r = p2 = 0. As to the branch points fixed by σ 2 , they are 0, ∞, the p2 branch
points lying in F ∩ Fix( σ2 ), say ν1 , . . . , νp2 , and the images of these last under the
σ2 ◦ σ
rotation ( 1 )n/2 : Σ → Σ ; x → −x, namely, −ν1 , . . . , −νp2 . This makes a
total of 2p2 + 2 branch points lying on Fix( σ2 ). Hence, according to Theorem 4.3.5,
−(p2 + 1) if p2 + 1 < g + 1;
sp(σ2 ) =
g+1 if p2 + 1 = g + 1.
Again this last equality holds just for n = 2 and r = p1 = 0. Consequently, the
symmetry type of S for n even is the one given in the statement of the theorem.
Assume now that n is odd and let us keep the above meaning of λ1 , . . . , λp1
1 are 0, ∞, λ1 , . . . , λp1
and ν1 , . . . , νp2 . In this case the branch points fixed by σ
and the images of ν1 , . . . , νp2 under the rotation ( 1 )(n−1)/2 : Σ → Σ ; x →
σ2 ◦ σ
−x · e−πi/n . This makes a total of p1 + p2 + 2 branch points lying on Fix( σ1 ).
Hence, according to Theorem 4.3.5,
As to the branch points fixed by σ 2 , they are 0, ∞, ν1 , . . . , νp2 and the images of
σ2 ◦ σ
λ1 , . . . , λp1 under the rotation ( 1 )(n+1)/2 : Σ → Σ ; x → −x · eπi/n . This
makes a total of p1 + p2 + 2 branch points lying on Fix( σ2 ). Hence, according to
Theorem 4.3.5,
sp(σ2 ) = −(p1 + p2 + 2)/2.
Therefore, the symmetry type of S for n odd is
In this chapter we study the symmetries of those surfaces S whose group Aut+ (S)
of analytic automorphisms is large. As said in the introduction, the symmetries of
these surfaces are topologically determined by Aut+ (S). We will explain in detail
results that originally are due to the authors of this monograph in collaboration
with Broughton [8, 9] and to Turbek alone [124]. More precisely, in this chapter we
classify the symmetries of the Macbeath–Singerman, the Accola–Maclachlan and
the Kulkarni surfaces. Our choice comes from the fact that the results are complete
for these surfaces. However, the interested reader is invited to see also [15, 16, 94].
Many arguments are closely related to Gromadzki’s work in [48].
Let us write the compact Riemann surface S as H/Γ for some surface Fuchsian
group Γ and Aut+ (S) = Δ/Γ, where Δ is a Fuchsian group containing Γ as a
normal subgroup. For the surfaces we shall deal with in this chapter, the group
Δ has triangle signature. We shall find general formulae for the number of ovals
of each symmetry of such surfaces. These formulae appeared in [48], see also the
recent paper [52] for a refined version. Following Theorem 3.1.1 and Lemma 1.3.3
we divide this study into several cases according to the parity of the three proper
periods of Δ.
For the reader’s convenience, let us introduce some technical notions that we will
use in the sequel. Given an automorphism ϕ of a group G, two elements x, y ∈ G
are said to be ϕ-conjugate, and denoted by
x ∼ϕ y if x = wyϕ(w)−1
for some w ∈ G. This notion was introduced in the context of low dimension topol-
ogy by Reidemeister in [109]. Observe that if ϕ is the identity then this notion
coincides with the usual conjugacy.
Recall also that the isotropy group Isotr(ϕ) of ϕ is the subgroup consisting of all
elements of G fixed by ϕ:
Theorem 5.1.1. Let S be a symmetric Riemann surface of the first type whose group
Aut+ (S) of analytic automorphisms is generated by a pair (a, b) of elements of
order k and = 2 + 1, respectively, whose product has order m = 2m + 1.
(1) The surface S has exactly one conjugacy class of symmetries with ovals. The
number of its ovals is N/M , where N is the order of the isotropy group of ϕ in
Aut+ (S) and
k
2 a (ab)−m b ak b− (ab)m if k = 2k,
M=
(ab)−m b ak if k = 2k + 1.
(2) The surface S has a symmetry without ovals if and only if there is an element
x ∈ Aut+ (S) such that ϕ(x) = x−1 and x is not ϕ-conjugate to 1.
Proof. Let us write S = H/Γ for some surface Fuchsian group Γ and G =
Aut+ (S) = Δ/Γ, where Δ is a Fuchsian group with signature s(Δ) = [k , , m ].
By the proof of Theorem 1.5.10 there exists an NEC group Λ with signature s(Λ) =
(k , , m ) containing Δ and Γ as normal subgroups so that G = Aut(S) = Λ/Γ.
By Remark 1.5.11 we know that the smooth epimorphism θ : Λ → G = G Z2 =
a, b t is given by θ(c0 ) = at, θ(c1 ) = t, θ(c2 ) = tb, where {c0 , c1 , c2 } is a set
of canonical generating reflections of Λ.
Let σ be a symmetry of S and write σ = Γσ /Γ where Γσ is a proper NEC
group. If Γσ contains no conjugate to a canonical reflection then σ fixes no oval by
Lemma 1.5.9. If Γσ contains a conjugate of a canonical reflection then σ fixes κ > 0
ovals, that is, the signature of Γσ has κ > 0 period cycles. Observe that c0 , c1 and c2
are pairwise conjugate since and m are odd. Therefore, the surface S has exactly
one conjugacy class of symmetries with ovals. In addition, Theorem 3.1.1 yields
Since t and at are conjugate, the centralizers C(G, t) and C(G, at) have the same
order. Now, the restriction to G of conjugation by t is precisely the automorphism ϕ.
It follows that the order of C(G, t) is twice the order of the isotropy group of
ϕ in G. Let us calculate now the order of θ(C(Λ, c0 )). Assume that k = 2k.
5.1 Some General Results 93
Then, using Lemma 1.3.3 and noting that the images in G of the generators of
C(Λ, c0 ) have order 2, we obtain that θ(C(Λ, c0 )) has order 4M , where M is the
order of the element
θ (c0 c1 )k (c2 c0 )m (c1 c2 ) (c1 c0 )k (c2 c1 ) (c0 c2 )m = ak (ab)−m b ak b− (ab)m .
This completes the first part of the proof for k even. The first part of the proof for
k odd is similar and we omit it.
Now, any symmetry σ has the form xt for some x ∈ G and, in fact, ϕ(x) = x−1
because (xt)2 = 1. Observe that σ fixes no oval if and only if σ is not conjugate
to t. But xt is conjugate to t if and only x ∼ϕ 1, as is easy to see. This completes
the proof of the theorem.
Theorem 5.1.2. Let S be a symmetric Riemann surface of the first type whose group
Aut+ (S) of analytic automorphisms is generated by a pair (a, b) of elements of
order k = 2k and = 2, respectively, whose product has order m = 2m.
(1) The surface S admits three symmetries with ovals, σ1 , σ2 and σ3 , such that
each symmetry with ovals of S is conjugate either to σ1 or to σ2 or else to
σ3 . Furthermore, σ1 ∼ σ2 , σ2 ∼ σ3 and σ1 ∼ σ3 , respectively, if and only if
a ∼ϕ 1, b ∼ϕ 1 and a ∼ϕ b±1 , respectively.
Let M1 , M2 and M3 be the orders of ak (ab)m , ak b and (ab)m b , respectively,
and let N1 , N2 and N3 be the orders of the isotropy groups of ϕa , ϕ and ϕb in
Aut+ (S), respectively. Then,
• If σ1 , σ2 and σ3 are pairwise non-conjugate then σi has Ni /(2Mi ) ovals
for i = 1, 2, 3.
• If σ1 , σ2 and σ3 are pairwise conjugate then each of them has N1 /(2M1 )+
N2 /(2M2 ) + N3 /(2M3 ) ovals.
• If (u, v, w) is a permutation of (1, 2, 3) such that σu ∼ σv and σw is non-
conjugate to σu and σv then σu has Nu /2(Mu ) + Nv /(2Mv ) ovals and
σw has Nw /(2Mw ) ovals.
(2) The surface S has a symmetry without ovals if and only if there is an element
x ∈ Aut+ (S) such that ϕ(x) = x−1 and x is ϕ-conjugate neither to 1, nor to
a±1 nor to b±1 .
Theorem 5.1.3. Let S be a symmetric Riemann surface of the first type whose group
Aut+ (S) of analytic automorphisms is generated by a pair (a, b) of elements of
order k = 2k and = 2, respectively, whose product has order m = 2m + 1.
(1) The surface S admits two symmetries with ovals, σ1 and σ2 , such that each
symmetry with ovals of S is conjugate either to σ1 or to σ2 . Furthermore σ1 ∼
σ2 if and only if a ∼ϕ 1.
94 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
(2) The surface S has a symmetry without ovals if and only if there is an element
x ∈ Aut+ (S) such that ϕ(x) = x−1 and x is not ϕ-conjugate to 1.
Proof. Write S = H/Γ for some surface Fuchsian group Γ. Then G = Aut+ (S) =
Δ/Γ, where Δ is a Fuchsian group with signature [k, k, m] containing Γ as a
normal subgroup. By the proof of Theorem 1.5.10, there exists an NEC group
Λ with signature s(Λ) = (0; +; [k]; {(m)}) containing Γ and Δ as normal sub-
groups and, by Remark 1.5.11, the group Γ is the kernel of the epimorphism
θ:Λ→G = G Z2 = a, b t induced by the assignment
Now c0 and c1 are conjugate in Λ and therefore the surface S has exactly one
conjugacy class of symmetries with ovals. In order to count them, we have to
find the number of empty period cycles in the signature of the subgroup Γ0 =
θ−1 (t) = Γ, c0 . By Theorem 3.1.1, the number of ovals of this symmetry equals
t)) : θ(C(Λ, c0 ))]. Now the order of C(G,
[C(G, t)) is 2N , where N is the order of
the isotropy group of ϕ in G and, by Lemmata 1.3.1 and 1.3.2,
c0 ⊕ (c0 c1 )m/2 ∗ e(c0 c1 )m/2 e−1 if m is even,
C(Λ, c0 ) =
c0 ⊕ e(c0 c1 )(m−1)/2 if m is odd.
5.2 Symmetry Types of Macbeath–Singerman Surfaces 95
So for even m, the order of θ(C(Λ, c0 )) is four times the order of (ab)m/2 (ba)m/2 ,
whilst for odd m, it is twice the order of (ab)(m+1)/2 a. This completes the proof of
the first part. The proof of the second part is analogous to that of Theorem 5.1.1.
has the projective special linear group PSL(2, 7) of order 168 = 84(3 − 1) as a
Hurwitz automorphism group. This surface is the so called Klein quartic. Macbeath
proved in [72] the existence of a unique Riemann surface of genus 7 on which
the group PSL(2, 8) of order 504 = 84(7 − 1) acts as a Hurwitz group. Later on,
Macbeath characterized in [74] the values of the positive integers q for which the
group PSL(2, q) acts as a Hurwitz group on some Riemann surface. These integers
are usually called H-numbers. Macbeath’s characterization is the following, see [74,
Theorem 8].
Theorem 5.2.1. A positive integer q is an H-number if and only if either q = 7 or
q ≡ ±1 (mod 7) is prime, or q = p3 for some prime number p such that p ≡ ±2
(mod 7) or p ≡ ±3 (mod 7).
A first result about the symmetric character of surfaces with large group of ana-
lytic automorphisms is the following.
Theorem 5.2.2. Let p be a prime H-number and let S be a Riemann surface ad-
mitting PSL(2, p) as a Hurwitz automorphism group. Then S is symmetric.
U AU −1 = A−1 and U BU −1 = B −1 .
Proposition 5.2.3. With the notations in the proof of Theorem 5.2.2, the element
u ∈ PSL(2, k) has order 2. In particular, if u ∈ PSL(2, k) \ PSL(2, p) then the
semidirect product PSL(2, p)u defined by
Proof. The argument is easier if u ∈ PSL(2, p). Indeed, it suffices to check that
u2 g = gu2 for each g ∈ PSL(2, p), because PSL(2, p) is a simple group. Moreover,
since a and b generate PSL(2, p), it is enough to show this equality for g = a and
g = b. Recall that ugu−1 = g −1 for these values of g and so
Then
−1 β 1 β2
Q CQ = ,
β3 β4
where
β1 = q1 q4 α1 + q3 q4 α2 − q1 q2 α3 − q2 q3 α4 ,
β2 = q2 q4 α1 + q42 α2 − q22 α3 − q2 q4 α4 .
while
β1 + τ β2 β2
(Q−1 CQ)X = .
β3 + τ β4 β4
Then τ = 0 and again u2 = v = 1.
The last part follows easily from [37, Theorem 7.5].
Aut(S) = PSL(2, p) Z2 .
and let
Isotr(ϕu ) = {g ∈ PSL(2, p) : ϕu (g) = g}
be its isotropy group.
(2.1) If u ∈ PSL(2, p) then the order of Isotr(ϕu ) coincides with the order of
the centralizer in PSL(2, p) of the class (mod ± I) of the matrix
0 1
M= ∈ SL(2, p).
−1 0
(2.2) If u ∈ PSL(2, k)\PSL(2, p) then the order of Isotr(ϕu ) is half the order
of the centralizer in PGL(2, p) of the class (mod ±I) of the matrix
0 1
N= ∈ GL(2, p),
−δ 0
where δ ∈ Fp is a non-square in Fp .
Proof. (1) This part is an immediate consequence of Remark 1.5.11 and
Proposition 5.2.3.
(2.1) Let U ∈ SL(2, p) be a representative of u and consider the Fp -linear endo-
morphism
2 2 x x
f : Fp → Fp ; → U ,
y y
that satisfies f 2 = ε id for some ε = ±1. There exists a vector ω ∈ F2p such
that
B = {ω, εf (ω)}
is a basis of F2p . The matrix of f with respect to B is
0 1
Mf (B) = .
ε 0
Cm → Isotr(ϕu ) ; g → qgq −1
Let n be the class of N (mod ± I) and let q be the class (mod ± I) of Q. Let
Cn be the centralizer of n in PGL(2, p). The result follows because the map
Cn → Isotr(ϕu ) ; g → qgq −1
is two-to-one.
Proof. (1) By part (2.1) in Proposition 5.2.4 the order of Isotr(ϕu ) is half the
number of matrices C ∈ SL(2, p) such that CM = ±M C, where
0 1
M= .
−1 0
Writing C as
x y
C= ,
z t
the condition CM = ±M C, together with det C = 1, leads to
⎧ 2 ⎧ 2
⎨ x + y2 = 1 ⎨ x + y 2 = −1
either x −t =0 or x + t = 0.
⎩ ⎩
z +y =0 z −y = 0
Clearly,
card(Γ1 ) + card(Γ2 )
| Isotr(ϕu )| = .
2
100 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
Suppose now that p ≡ 1 (mod 4). Then there exists ξ ∈ Fp such that ξ 2 = −1
and so the linear isomorphism
Assume now that p ≡ 3 (mod 4). We choose a point (x0 , y0 ) ∈ Γ2 and consider
the linear map
is a bijection. So | Isotr(ϕu )| = p + 1.
(2) By part (2.2) in Proposition 5.2.4, the order of Isotr(ϕu ) is 1/4 the number of
matrices D ∈ GL(2, p) such that DN = ±N D, where
x y 0 1
D= and N = ,
z t −δ 0
that satisfy
card(H1 ) + card(H2 ) + card(H3 ) + card(H4 )
| Isotr(ϕu )| = .
4
5.2 Symmetry Types of Macbeath–Singerman Surfaces 101
card(H1 ) + card(H2 )
| Isotr(ϕu )| = .
2
| Isotr(ϕu )| = card(H1 ) = p + 1,
Remark 5.2.6. The above proof of Proposition 5.2.5 shows in particular that for an
involution u ∈ PSL(2, p), we have
for some generating pair (a, b). Let M (a, b) be the order of the element
(1) Suppose that Aut(S) = PSL(2, p) ⊕ Z2 . Then S admits exactly two conjugacy
classes of symmetries. A representative of the first one fixes no oval while the
number of ovals fixed by a representative of the second one is
(p − 1)/(2M (a, b)) if p ≡ 1 (mod 4),
(p + 1)/(2M (a, b)) if p ≡ 3 (mod 4).
(2) Suppose now that Aut(S) = PGL(2, p). Then S admits exactly one conjugacy
class of symmetries. The number of ovals fixed by any of them is
(p + 1)/(2M (a, b)) if p ≡ 1 (mod 4),
(p − 1)/(2M (a, b)) if p ≡ 3 (mod 4).
Proof. The formulae are consequence of Theorem 5.1.1 and Proposition 5.2.5, tak-
ing into account that Aut(S)\Aut+ (S) contains either one or two conjugacy classes
of involutions according to Aut(S) being PGL(2, p) or PSL(2, p)⊕Z2 respectively,
see [38].
Remark 5.2.8. It is well known that Computer Algebra Systems like CAYLEY,
MAGMA , GAP, MATLAB , MAPLE and MATHEMATICA find a presentation of the pro-
jective special linear group of the form
very quickly, and they compute the order M (a, b) of the element a(ab)4 bab2 (ab)3
associated to a given presentation of the required form. This enables us, by applying
Corollary 5.2.7, to effectively calculate the number of ovals of each symmetry of a
Macbeath–Singerman surface.
For example, if p = 13 there exists, up to conformal equivalence, a unique
Macbeath–Singerman surface S with Aut(S) = PSL(2, 13) ⊕ Z2 . By Corollary
5.2.7 it admits a fixed point free symmetry and, moreover, M (a, b) = 6; hence
it admits also a symmetry with exactly one oval. On the other hand, there exist
two non-isomorphic Macbeath–Singerman surfaces S1 , S2 such that Aut(Si ) =
PGL(2, 13) for i = 1, 2. In this case all symmetries of each surface Si are conju-
gate and each of them has exactly one oval, because M (a, b) = 7.
Recall that a symmetry σ of a Riemann surface S is said to be separating if
S \ Fix(σ) is disconnected. Otherwise it is said that σ is non-separating. Although
we have just remarked that M (a, b) can be calculated with the aid of a Computer Al-
gebra System, in order to demonstrate the non-separating character of the involved
symmetries, we will also need to show that M (a, b) ≥ 3. This is the goal of the next
proposition.
Proposition 5.2.9. Let p ≥ 7 be a prime number and let a presentation of the
projective linear group be given by PSL(2, p) = a, b | a2 , b3 , (ab)7 , . . .. Then the
order of the element a(ab)4 bab2 (ab)3 is at least 3.
5.2 Symmetry Types of Macbeath–Singerman Surfaces 103
that is, m = 0. Therefore, it is enough to prove that m ∈ {−2, 0, 2}. To that end we
first express m as a polynomial with respect to the trace γ of C. More precisely, we
will show that the following equality holds:
B 2 = −(B + I).
Write
M = A(AB)4 BAB 2 (AB)3 = AC 4 DB 2 C 3 ,
where
D = BA.
Moreover, since
DC = BAAB = −B 2 = B + I
and
BDB = B 2 AB = −(B + I)AB = −(DB + C),
we have
M = (γ 3 − 2γ)B + (γ 2 − 1)A (γ 3 − 2γ)DB + (γ 2 − 1)I − γD
= −(γ 3 − 2γ)2 (DB + C) + (γ 3 − 2γ)(γ 2 − 1)B + γ(γ 3 − 2γ)(D + A)
+(γ 3 − 2γ)(γ 2 − 1)(γC − I) + (γ 2 − 1)2 A − γ(γ 2 − 1)AD. (5.2)
Notice that
tr(AD) = tr(ABA) = tr(BA2 ) = tr(−B) = 1
and
Thus tr(DB + C) = 0. We now deduce equality (5.1) after comparing the traces of
both sides in equality (5.2):
f (T ) = T 3 + T 2 − 2T − 1. (5.3)
Since C 2 = γC − I we get
Consequently,
I = C 7 = C 6 C = (γ 6 − 5γ 4 + 6γ 2 − 1)C − (γ 5 − 4γ 3 + 3γ)I.
5.2 Symmetry Types of Macbeath–Singerman Surfaces 105
m (T ) = T 7 − 5T 5 + 8T 3 − 5T − m.
m (T ) = (T 4 − T 3 − 2T 2 + T + 2)f (T ) − rm (T )
Proof. Let S+ and S− be the connected components of S \ Fix(σ), and let v denote
a common fixed point of σ and τ . Since τ commutes with σ, either it interchanges
S+ and S− or it maps both of them to themselves. By analyzing the analytic map h
locally around v we see that τ and σ behave like symmetries with respect to circles
meeting at right angles, and this implies that S+ and S− are both mapped by τ to
themselves.
Consequently h(S+ ) = S− and h(S− ) = S+ . Therefore Fix(h) ⊂ Fix(σ) and
so Fix(h) ⊂ Fix(τ ). Thus Fix(h) ⊂ Fix(τ ) ∩ Fix(σ) and, since the converse is
evident, we get
Fix(h) = Fix(τ ) ∩ Fix(σ).
Finally, as the sets Fix(σ) and Fix(τ ) have no common connected component, h
fixes at most two points on each oval of σ, as desired.
The third result we will use in the proof of Theorem 5.2.12 is a classical formula
due to Macbeath [75, Theorem 1] for counting the number of points fixed by an
automorphism. We include a proof here for the sake of completeness. To state it we
introduce some notations. Let S be a Riemann surface of genus ≥ 2 and let G be a
group of automorphisms of S. Let Λ be a Fuchsian group and let θ : Λ → G be a
106 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
group epimorphism whose kernel is the surface Fuchsian group Γ that uniformizes
S, that is, S = H/Γ. Let m1 , . . . , mr be the proper periods in the signature of Λ
corresponding to the elliptic canonical generators x1 , . . . , xr of Λ. Given h ∈ G, let
δi (h) = 1 if h is conjugate to a power of gi = θ(xi ) and δi (h) = 0 otherwise.
Theorem 5.2.11. If h ∈ G is not the identity then the number of points of S fixed
by h is given by the formula
r
δi (h)
| Fix(h)| = |NG (h)| ,
i=1
mi
We have seen in Corollary 5.2.7 that all symmetries of S with fixed points are
conjugate. So it suffices to prove that σ is non-separating. To obtain a contradiction,
suppose that σ is separating. Clearly τ = aσ ∈ Aut(S) \ Aut+ (S) and it is in fact
a symmetry as
τ 2 = aσaσ = aσaσ −1 = aa−1 = 1.
Moreover, στ = τ σ, because
p+1
| Fix(a)| ≤ .
M (a, b)
2(p + 1) 2(p + 1)
p − 1 ≤ |C(PSL(2, p), a)| = 2| Fix(a)| ≤ ≤ ,
M (a, b) 3
a contradiction because p ≥ 7.
This was done, independently, by Hall [58] and Conder [33] and, to explain the
employed method, we need the following:
(5.2.14) Factorization of f1 (T ) in Fp [T ].
Let us see first that the polynomial
f1 (T ) = T 3 + T 2 − 2T − 1 ∈ Fp [T ]
P (T ) = 1 + T + T 2 + T 3 + T 4 + T 5 + T 6 ∈ Z[T ] and
Q(T ) = T 2k − T k+j − T k−j + 1 ∈ Z[T ]
1 = R1 (T )P (T ) + R2 (T )Q(T ).
f3 (T ) = f1 (T 2 ) = T 6 + T 4 − 2T 2 − 1 = (T 2 − μ1 )(T 2 − μ2 )(T 2 − μ3 ).
W. Hall and M. Conder defined the invariant δ(p) = deg(gcdFp [T ] (f2 , f3 )), and it
follows that δ(p) = 2κ(p), where
f1 (T ) = T 3 + T 2 − 2T − 1.
We use the quadratic reciprocity law to decide how many roots μj are squares in
F29 . Indeed, 29 ≡ 2 (mod 3) is not a square in F3 and so μ1 is not a square in F29 .
On the other hand, 29 ≡ 1 (mod 7) is a square in F7 and so μ2 is a square in F29 .
Finally 29 ≡ 7 ≡ −4 (mod 11) is not a square in F11 . Thus −μ3 is not a square in
F29 . Since −1 is a square in F29 , we conclude that μ3 is not a square in F29 .
Therefore κ(29) = 1, that is, δ(29) = 2 and 29 ≡ 1 (mod 4). Hence d(29) = 1
and s(29) = 2. Let S1 , S2 and S3 be non-isomorphic Riemann surfaces on which
the group PSL(2, 29) acts as a Hurwitz group of automorphisms. Then, up to re-
ordering,
Let (a1 , b1 ), (a2 , b2 ) and (a3 , b3 ) be generating pairs of PSL(2, 29) with
In the 1960’s, Accola [1] and Maclachlan [77] proved independently that for every
integer g ≥ 2 there is a Riemann surface Xg of genus g whose automorphism
group has order 8g + 8. It is called the Accola–Maclachlan surface. This result is
interesting because 8g + 8 is the largest order of an automorphism group that can
be uniformly constructed for every g. Much later, Kulkarni considered the question
of uniqueness of these surfaces. In other words, is Xg the only surface of genus g
whose automorphism group has order 8g + 8? It turns out, see [70], that this is so
for g ≡ 0, 1, 2 (mod 4) and g sufficiently large. However, for large enough g ≡ 3
(mod 4), Kulkarni also proved that, in addition to the Accola–Maclachlan surface,
there exists exactly one other surface of genus g whose automorphism group has
order 8g + 8. We shall call it the Kulkarni surface and denote it by Yg . Moreover, he
found the following presentations for the groups of analytic automorphisms of Xg
and Yg :
Proof. It suffices to see, by Theorem 1.5.10, that the assignment a → a−1 , b → b−1
induces, in both cases, an automorphism of the group G of analytic automorphisms
of either Xg or Yg . For Xg this is equivalent to saying that a−1 has order 2(g + 1),
b−1 has order 4, a−1 b−1 has order 2 and b−2 is central in G. Since b−2 = b2 , just
the third claim requires some care. Here, |a−1 b−1 | = |ba| = |ab| = 2.
As to the surface Yg , it is enough to check that b−2 a−1 b−2 a−g = 1, which
follows immediately from the last relator in the presentation of Aut+ (Yg ).
Our next goal is to calculate the number of ovals of the symmetries of the surfaces
Xg and Yg . To that end we will apply Theorem 5.1.2. We start with the Accola–
Maclachlan surface Xg .
5.3 Symmetry Types of Accola–Maclachlan and Kulkarni Surfaces 111
(3) The element ag+2 b has order 2 if g is odd and order 4 if g is even.
(4) The elements ag+1 b2 and ab3 have order 2.
(5) Let ϕ be the automorphism of the group Aut+ (Xg ) induced by the assignment
ϕ : a → a−1 , b → b−1 .
Then
and
{b, b3 } for even g,
Z= 3 g+1 g+1 3
{b, b , a b, a b } for odd g.
Then X = Y ∪ Z.
(8) The ϕ-conjugates to 1 in Aut+ (Xg ) are the elements of the form a2i b2j .
The ϕ-conjugates to a±1 in Aut+ (Xg ) are the odd powers of a.
The ϕ-conjugates to b±1 in Aut+ (Xg ) are the odd powers of b.
In particular, ab2 ∈ X and it is ϕ-conjugate neither to 1, nor to a±1 , nor
to b±1 .
Proof. (1) It is easily seen that for different pairs (i, j) = (k, ) with 0 ≤ i, k ≤
2g + 1 and 0 ≤ j, ≤ 3, the products ai bj and ak b are distinct elements in
Aut+ (Xg ). Since the order of this last group is 8(g + 1), we are done.
112 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
It is easy to check that this equality also holds for negative values of i.
(3) For odd g we get
This implies a2i = b1−2j , which is false. Finally, suppose that a ∼ϕ b±1 , i.e.,
there exists w = ai bj ∈ Aut+ (Xg ) such that
or, equivalently, bj a2i−1 bj = a−1 . For even j this means a2i−1 = a−1 , that is,
a2i = 1. Hence either i = 0 or i = g + 1. This way x ∈ {1, b2 , ag+1 , ag+1 b2 }.
If j is odd then
or, equivalently, bai b = a−i b2(1−j) . If i is even then a−i b2 = a−i b2(1−j) , that
is, j = 0 or 2. If i is odd then a−i = a−i b2(1−j) and so j = 1 or 3. Therefore
| Isotr(ϕb )| = 4(g + 1).
(7) Let x = ai bj with even j. Then ϕ(x) = a−i b−j = b−j a−i = x−1 and so
x ∈ X. Hence Y ⊂ X. For odd j, the element x−1 = b−j a−i equals ϕ(x) =
a−i b−j if and only if ai bj = bj ai , that is, ai b2 = bai b.
If i is odd then bai b = a−i , by (2), and the equality ai b2 = a−i cannot hold.
If i is even then ai b2 = bai b = a−i b2 , that is, either i = 0 or i = g + 1.
Therefore X = Y ∪ Z.
(8) An element x is ϕ-conjugate to 1 if there exists w = ai bj ∈ Aut+ (Xg ) such
that
x = wϕ(w)−1 = ai bj (a−i b−j )−1 = ai bj bj ai = a2i b2j .
Analogously, x is ϕ-conjugate to a±1 if there exists w = ai bj ∈ Aut+ (Xg )
such that
x = wa±1 ϕ(w)−1 = ai bj a±1 bj ai = a2i±1 .
Finally, x is ϕ-conjugate to b±1 if there exists w = ai bj ∈ Aut+ (Xg ) such that
(2) The surface Xg admits a fixed point free symmetry. There is one conjugacy class
of fixed point free symmetries if g is even and two classes if g is odd.
Proof. (1) The existence of exactly three pairwise non-conjugate symmetries with
fixed points follows immediately from Theorem 5.1.2 and part (5) in Propo-
sition 5.3.2. Moreover, by part (1) in Theorem 5.1.2 and parts (3) and (6) in
Proposition 5.3.2 we get
| Isotr(ϕa )| 2 for odd g,
σ1 = =
2|ag+2 b| 1 for even g.
| Isotr(ϕ)| | Isotr(ϕb )|
σ2 = = 1 and σ3 = = g + 1.
2|ag+1 b2 | 2|ab3 |
(2) The first statement follows straightforwardly from Theorem 5.1.2 and part (8)
in Proposition 5.3.2. The second one is an easy exercise in group theory, see [9].
(3) The orders of ag+2 b, ag+1 b2 and ab3 are 2, 2 and 4, respectively.
(4) Let ϕ be the automorphism of the group Aut+ (Yg ) induced by the assignment
ϕ : a → a−1 , b → b−1 .
Then a ϕ 1, b ϕ 1 and a ϕ b±1 .
5.3 Symmetry Types of Accola–Maclachlan and Kulkarni Surfaces 115
Then
Therefore,
For the second it suffices to check that (ag+1 b2 )2 = 1. Using (2) once more,
2 2
(ag+1 b2 )2 = ag+1 b(bag+1 )b2 = ag+1 ba(g −1)/2 b3 = ag+1 a(g −1)(g−1)/4 = 1,
or, equivalently, bj a2i−1 bj = a−1 . For j = 0 this means a2i−1 = a−1 , that
is, a2i = 1. Hence either i = 0 or i = g + 1. For j = 1 the condition is
1 = ba2i−1 ba, that is, ba2i−1 = ba or, equivalently, either i = 1 or i = g + 2.
For j = 2 we must solve the equation b2 a2i−1 b2 = a−1 with 0 ≤ i ≤ 2g + 1,
i.e.,
a2i−1 = b2 a−1 b2 = (b2 ab2 )−1 = ag ,
whose solutions are i = (g + 1)/2 and i = 3(g + 1)/2. Finally, for j = 3 we
have b3 a2i−1 b3 = a−1 or, equivalently,
This leads us to distinguish two cases, according to the parity of the exponent j.
First, for odd j, we have a−i = bai b. By (2) this forces i to be odd, and hence
a contradiction. If i is even then = i(g − 1)/2 is also even, and this implies
Finally, suppose that x ∼ϕ b±1 , and let w = ai bj ∈ Aut+ (Yg ) such that
x = wb±1 ϕ(w)−1 , that is, there exists k ∈ Z with
For odd k this means that a(g+1)/2 b = ai bai and, by (2), the exponent i = 2
is even. Thus,
For j = 2 the equality ϕ(x) = x−1 means a−i b−2 = b−2 a−i , that is, ai b2 =
b2 ai and this is equivalent to
Corollary 5.3.5. (1) The Kulkarni surface Yg admits exactly three pairwise non-
conjugate symmetries σ1 , σ2 and σ3 with fixed points. The number of ovals of
each symmetry is
g+1
σ1 = 2, σ2 = 1 and σ3 = .
4
(2) The surface Yg admits a fixed point free symmetry if and only if g ≡ 3 (mod 8).
In such a case, all fixed point free symmetries of Yg are pairwise conjugate.
Proof. (1) The existence of exactly three pairwise non-conjugate symmetries with
fixed points follows immediately from Theorem 5.1.2 and part (4) in Propo-
sition 5.3.4. Moreover, by part (1) in Theorem 5.1.2 and parts (3) and (5) in
Proposition 5.3.4 we get
| Isotr(ϕa )| 8
σ1 = = = 2.
2|ag+2 b| 4
120 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
(2) The first statement is an immediate consequence of Theorem 5.1.2 and parts (6)
and (7) in Proposition 5.3.4. The second one is an easy exercise in group theory,
see [9].
Our goal now is to find out the separating or non-separating character of the sym-
metries of the Accola–Maclachlan and the Kulkarni surfaces. Obviously, the fixed
point free symmetries are non-separating, so we just focus on symmetries with fixed
points. As we observed in Theorem 1.5.3, Hurwitz proved in his classical article [59]
that given a symmetry σ of a genus g Riemann surface the following conditions
hold:
1 ≤ k + ε ≤ g + 1; k ≡ g + 1 (mod (2 − ε)),
where k is the number of ovals of σ and ε = 0 or 1 according to σ being separating
or not. In particular,
(1) If k = g + 1 then ε = 0, and if k = g then ε = 1.
(2) If k = 1 and g is odd then ε = 1.
(3) If g ≡ 3 (mod 8) and k = (g + 1)/4 then ε = 1.
Therefore, with the notations throughout this section, we get the following.
and so, if we denote Z2 = t, the group Aut(Xg ) is the image of Λ under the
epimorphism θ : Λ → Aut(Xg ) induced by the assignment
c0 → at = σ1 ; c1 → t = σ2 ; c2 → tb = σ3 . (5.4)
s(Λ) = (h; ±; [2, . v. ., 2]; {(−), . t. ., (−), (2, .r.1., 2), . . . , (2, .r.s., 2)})
for some non-negative integers h, v, t and some even positive integers r1 , . . . , rs be-
cause, as ker θ is a surface Fuchsian group, the canonical reflections corresponding
to a non-empty period cycle of s(Λ) must be mapped, alternatively, to σ and τ . This
implies that r = r1 + · · · + rs ≤ 4, because τ has two ovals.
On the other hand, the hyperbolic area of a fundamental region of Λ is
But, as σ and τ have g + 3 ovals in total, we deduce from Theorem 3.1.1 that
g + 3 ≤ 2t + r. After substituting, we get
2ηh + v + 2s ≤ r/2 ≤ 2.
s1 = (0; +; [−]; {(−), . t. ., (−)}) and s2 = (0; +; [−]; {(−), . t. ., (−), (2, 2, 2, 2)}),
with t = (g + 3)/2 in the first case and t = (g − 1)/2 in the second one.
Suppose first that s(Λ) = s1 , and let c0 , . . . , ct−1 be a set of canonical reflections
of Λ. Since θ is surjective, it maps some canonical reflection to τ . In fact there is
only one ci with this property; otherwise σ ≤ 2(t − 2) = g − 1, a contradiction.
Thus, θ(cj ) = σ for j = i and so θ maps all connecting canonical generators of Λ
onto (0, 0) ∈ Z2 ⊕ Z2 . Hence, applying Theorem 1.2.1 to the kernel of p ◦ θ, where
p : Z2 ⊕ Z2 → Z2 ⊕ Z2 /τ is the canonical projection, we conclude that τ is a
separating symmetry.
Let us assume now that s(Λ) = s2 . Since ker θ is a surface group, the canonical
reflections c0 , c1 , c2 , c3 and c4 of the non-empty period cycle in s2 are mapped by
θ alternatively to σ and τ . Thus they contribute with two ovals to σ and also
with two ovals to τ . Consequently, the reflections in the empty period cycles
must contribute with g − 1 ovals to σ, and this forces θ(ei ) = (0, 0) for every
connecting canonical generator ei of Λ. Using again Theorem 1.2.1, we deduce that
τ is a separating symmetry.
By combining Proposition 5.2.10 and Theorem 5.2.11 we obtain now a formula
that yields the non-separating character of some symmetries of Xg and Yg .
Proposition 5.3.11. Let G = Aut+ (S) with S = Xg or S = Yg , and consider the
presentation of G given in the introduction of this section. Let σ, τ be two distinct
commuting symmetries of S having a common fixed point such that σ is separating.
For each x ∈ G let δx be either 1 if σ ◦ τ is conjugate to some power of x, or 0
otherwise. Then
δa δb δab
|C(G, σ ◦ τ )| + + ≤ 2σ.
2g + 2 4 2
This is the formula in Theorem 5.2.11. Indeed, since σ and τ are commuting sym-
metries, h has order 2, and so NG (h) = C(G, h). Moreover, as we observed in
5.3 Symmetry Types of Accola–Maclachlan and Kulkarni Surfaces 123
(5.3.8), the proper periods in the signature of a Fuchsian group Δ admitting a group
epimorphism Δ → G whose kernel is the Fuchsian group Γ that uniformizes S are
2g + 2, 4 and 2, and we are done.
σ ◦ τ = σ2 ◦ (σ3 ◦ σ2 ◦ σ3 ) = b2 ,
τ ◦ σ = (σ3 ◦ σ2 ◦ σ3 ) ◦ σ2 = tb2 t = b−2 = b2 ,
a contradiction.
(2) We now apply Proposition 5.3.11 with σ = σ3 and τ = σ2 ◦ σ3 ◦ σ2 . As before
it is easily seen that σ and τ are commuting symmetries with a common fixed
point. Moreover, σ ◦ τ = b2 and δa = δab = 0, δb = 1. By Proposition 5.3.4,
|C(Aut+ (Yg )(σ ◦ τ ))| = 4(g + 1) and so, using Proposition 5.3.11 once more,
a contradiction.
Proof. We prove just part (1) since the proof of (2) is essentially the same. Let
K = σ1 , σ2 be the subgroup of Aut(Xg ) generated by σ1 and σ2 . From (5.4), we
see that σ1 σ2 = (at)t = a is an element of order 2(g + 1). Hence K is the dihedral
group of 4g + 4 elements. We shall use the obvious equalities
containing the surface group Γ that uniformizes Xg . Let us denote by c0 , c1 and c2
the generating reflections of Γ1 satisfying
vi = {(σ1 σ2 )i , (σ1 σ2 )i σ1 }, 0 ≤ i ≤ 2g + 1.
Analogously,
Finally,
Φ(c2 )(vi ) = σ1 (σ2 σ1 )2g (vi ) = {σ1 (σ2 σ1 )2g−i , σ1 (σ2 σ1 )2g−i σ1 }
= {(σ1 σ2 )2g−i , (σ1 σ2 )2g−i σ1 } = v2g−i .
Therefore,
Consequently, the set of vertices of the Schreier graph S, with the loops deleted,
admits a bipartition
We summarize the results of this section in terms of species (see Definition 1.5.4)
as follows.
Theorem 5.3.14. (1) The species of representatives of the three conjugacy classes
of the symmetries with fixed points of the Accola–Maclachlan surface Xg and
the Kulkarni surface Yg are displayed in the following table:
Surface g σ1 σ2 σ3
Xg odd +2 −1 +(g + 1)
Xg even +1 −1 +(g + 1)
Yg g ≡ 3 (mod 4) −2 −1 −(g + 1)/4
(2) The surface Xg admits one conjugacy class of fixed point free symmetries if g
is even and two classes if g is odd.
(3) The surface Yg admits a fixed point free symmetry if and only if g ≡ 3 (mod 8).
In such a case, all fixed point free symmetries of Yg are pairwise conjugate.
It is well known that the Accola–Maclachlan and the Kulkarni surfaces admit a
nice description by means of defining algebraic equations. Our goal in this sec-
tion is to provide algebraic formulae for the symmetries and automorphisms of
these surfaces when they are described by polynomial equations. We start with the
Accola–Maclachlan surfaces.
126 5 Symmetry Types of Riemann Surfaces with a Large Group of Automorphisms
1 iy
a : (x, y) → x · e πi/(g+1)
, y , b : (x, y) → , , σ : (x, y) → (x̄, ȳ).
x xg+1
Proof. Let AutΣ (Xg ) be the group of Möbius transformations which preserve the
branch point set of Xg . This group has order 8(g + 1) because Aut(Xg ) has order
16(g + 1). Observe that the 2g + 2 branch points of Xg are the (2g + 2)-th roots of
unity, that is, ekπi/(g+1) for k = 0, . . . , 2g + 1. The set of these points is preserved
i : Σ → Σ defined, for i = 1, 2, 3, by
under the action of the reflections σ
1 : x → x̄,
σ 2 : x → 1/x̄ and
σ 3 : x → x̄ · eπi/(g+1) .
σ
σ σ3 ◦ σ
4 := ( 1 )g+1 ◦ σ
2 : Σ → Σ ; x → −1/x̄.
5.4 Algebraic Formulae for the Symmetries 127
Finally, it is easy to see that the points (x, y) of the form x = r · eπi/(2g+2) , y = is
with r, s ∈ R such that s2 = r2g+2 + 1, are points in Xg which are fixed by ab2 σ.
Therefore aσ is the fixed point free symmetry:
2 1 if g is even,
sp(aσ) = 0 and sp(ab σ) =
2 if g is odd,
Now we shall find defining equations for the symmetries with fixed points of the
Kulkarni surface Yg , where g ≡ 3 (mod 4) is large enough. To that end we recall
that Kulkarni proved in [70] that Yg is a (2g + 2)-cyclic covering of the Riemann
sphere. In fact, it admits the following defining equation
First we shall find explicit formulae for the elements of the group G. Second,
observe that if C(x, y) denotes the field of rational functions of Yg , the symmetries
of Yg are nothing else but the√field automorphisms of C(x, y) of order 2 that fix the
real numbers but map i = −1 to −i. In particular, since the polynomial defin-
ing Yg has real coefficients, the automorphism σ of C(x, y) over R induced by the
assignment
σ(x) = x; σ(y) = y; σ(i) = −i
(8g+8):1
Yg / Yg /G
(2g+2):1 2:1
2:1 / Yg /H.
Yg /K
Let Δ be a triangle Fuchsian group with signature s(Δ) = [2g + 2, 4, 2]. Denote
by x1 , x2 and x3 the elliptic canonical generators of Δ of orders 2g + 2, 4 and 2,
respectively. The kernel of the epimorphism Δ → G induced by the assignment
x1 → a ; x2 → b ; x3 → ab,
is the surface Fuchsian group Γ which uniformizes the surface Yg . In this way we
get a diagram
H / H/Δ
Yg = H/Γ / Yg /G = (H/Γ)/Δ/Γ.
The upper horizontal arrow ramifies over three branching points with ramification
indices 2g+2, 4 and 2, and so the same holds true for the lower horizontal arrow. We
denote by R0 , R∞ , R ∈ Yg /G the ramification points of the projection Yg → Yg /G,
with ramification indices 2g + 2, 4 and 2, respectively.
As to the covering Yg → Yg /K, it is easy to see that we may assume, without
loss of generality, that the automorphism a generating K is defined by (x, y) →
5.4 Algebraic Formulae for the Symmetries 129
Yg → Yg /K = Σ ; (x, y) → x,
that ramifies over the points p−1 , p1 and p0 corresponding to x = −1, x = 0 and
x = 1 in Σ. Using elementary methods, as in [43] or [92, pp. 73, 74], it is easily
checked that the respective ramification indices are
2g + 2 2g + 2 2g + 2
= 2g + 2, = g + 1, = 2g + 2.
gcd(2g +2, g +2) gcd(2g +2, g −1) gcd(2g +2, 1)
Of course, since both Yg /G and Yg /H are covered by the sphere Yg /K, they are
also spheres. To understand the covering Yg /H → Yg /G let us denote by r1 , . . . , rk
its ramification indices. Then, by the Hurwitz–Riemann formula,
k
2g(Yg /H) − 2 = [G : H] 2g(Yg /G) − 2 + (1 − 1/ri )
i=1
and, since g(Yg /H) = g(Yg /G) = 0 and each ri divides the degree of the map
Yg /H → Yg /G, which equals 2, we get −2 = 2(−2 + k/2), i.e., k = 2. Clearly,
the point R0 , that corresponds to a fixed point of the automorphism a, is unbranched
with respect to Yg /H → Yg /G, and so both R and R∞ are branching points whose
ramification index equals 2. Therefore there exist exactly two points q0 , q1 ∈ Yg /H
lying above R0 and there exists a unique point q∞ ∈ Yg /H lying above R∞ .
To finish the analysis of the combinatorics of these coverings let us denote by
s1 , . . . , s the ramification indices of Yg /K → Yg /H. By the Hurwitz–Riemann
formula,
2g(Yg /K) − 2 = [H : K] 2g(Yg /H) − 2 + (1 − 1/si )
i=1
α1 x α1 α1
b2 K(x) = and = −1, = 1.
α2 x + α3 α2 + α3 α2 − α3
It follows easily that bH(z) = 1−z, that is, bH(x2 ) = 1−x2 . Hence b(x2 ) = 1−x2 ,
and to calculate b(x) we must find a square root of 1 − x2 in C(x, y). Note that in
this field,
y 2g0
v(x, y) =
xg0 (x + 1)g0
5.4 Algebraic Formulae for the Symmetries 131
We can write
x2 − 1
y 8g0 = x4g0 (x + 1)4g0
x2
and so
x2
b(y)8g0 = b(x)4g0 (b(x) + 1)4g0 .
x2 − 1
After multiplying the respective sides of both equalities we get
y 2 b(y)2
x(x + 1)(−ixv 2 )(1 − ixv 2 )
is a 4g0 -th root of unity. Among the different possibilities we choose the simplest
one, namely,
Therefore both C(x, v 2 , u)|C(x) and C(x, v)|C(x) are degree 2 subextensions of
the cyclic Galois extension C(x, y)|C(x) and so they coincide. Hence, u ∈ C(x, v),
that is, there exist Ai ∈ C(x) with 0 ≤ i ≤ 3 such that
We are going to prove that (A0 , A2 ) = (0, 0). Otherwise we deduce, from the two
first equations above, that A21 = A23 v 4 , and so A1 = εA3 v 2 for some ε = ±1. After
substituting these values in the third and fourth equations, and multiplying the first
one by v 2 , we get
⎧
⎨ A1 (εA0 + A2 v 2 ) = 0,
2 2 4 2 2
A0 + A2 v + 2εA1 v = h,
⎩
A21 + A0 A2 = /2.
Consequently,
2
h= + A22 v 4 =⇒ 4v 4 A42 − 4hA22 + 2 = 0.
4A22
−ε
v2 = ∈ C(x),
4A22 + x3 (x + 1)
−x3 (x + 1)
2A1 A3 v 4 = h = −x3 (x+1)v 4 =⇒ 2A1 A3 = −x3 (x+1) =⇒ A3 = .
2A1
x6 (x + 1)2 v 4
= A21 + =⇒ 4A41 − 4A21 + x6 (x + 1)2 v 4 = 0.
4A21
± x3 (x + 1) ix2 (x + 1)(−1 ± x)
A21 = = .
2 2
This forces the sign to be “−” and so A21 = −ix2 (x + 1)2 /2. Since c = (1 + i)/2
satisfies c2 = i/2 we get, up to sign, A1 (x) = cix(x + 1). Therefore,
−x3 (x + 1) −x2
A3 = = = cx2 .
2A1 2ci
Theorem 5.4.6. Let g be a positive integer with g ≡ 3 (mod 4) and let C(x, y),
where y 2g+2 = (x − 1)xg−1 (x + 1)g+2 , be the field of rational functions of the
Kulkarni surface Yg of genus g. Let ξ be a primitive (2g + 2)-th root of unity, c =
(1 + i)/2 and g0 = (g + 1)/4. Consider the rational functions
y 2g0
v(x, y) = and u = cix(x + 1)v + cx2 v 3 .
xg0 (x + 1)g0
Then the automorphism group of the field extension C(x, y)|C is generated by the
automorphisms a and b defined by
u
a(x) = x, a(y) = ξy; b(x) = −ixv 2 , b(y) = .
y
Proof. The groups Aut(C(x, y)|C) and a, b coincide because they both have
8(g + 1) elements. The other claims are checked straightforwardly.
Henceforth,
xv 2
γ 2 (x) = γ(ixv 2 ) = σ(b(ix)b(v)2 ) = = x.
v2
Consequently,
a contradiction.
(2.4) For j = 3 we get x = b3 σb−2 (x) = b3 σb2 (x) = b3 σ(−x) = b3 (−x)
= −ixv 2 , a contradiction again.
(3) Suppose finally that τ and γ are conjugate. Let φ = ak bj ∈ Aut+ (Yg ) such
that
σb = φ−1 aσφ = b−j a−k aσak bj =⇒ bj σb1−j = a1−k σak .
Evaluating at x we get bj σb1−j (x) = a1−k σak (x) = x ; however, we have just
proved that this is false.
This way we have shown that the symmetries σ, τ = a◦σ and γ = σ ◦b are pairwise
analytically non-conjugate.
To finish this section we must prove the following.
Proposition 5.4.8. With the above notations, σ, τ = a ◦ σ and γ = σ ◦ b are
symmetries with fixed points of the Kulkarni surface Yg .
Proof. In case g ≡ 7 (mod 8) the surface Yg admits no fixed point free symmetry.
So in what follows we suppose that g ≡ 3 (mod 8). The symmetry σ fixes points,
because the equation
y 2g+2 = (x − 1)xg−1 (x + 1)g+2
admits a real solution for y whenever x ≥ 1.
As to the symmetry τ, note that τ (x) = x and τ (y) = a(y) = ξy where ξ =
eπi/(g+1) . Therefore, since ξ is unitary,
5.4 Algebraic Formulae for the Symmetries 137
¯
τ ((1 + ξ)y) = (1 + ξ)ξy = (1 + ξ −1 )ξy = (1 + ξ)y.
and κ is a negative real number. Indeed, as the argument of ξ equals π/(g + 1), we
have that arg(1 + ξ) = π/2(g + 1) and so arg(1 + ξ)2g+2 = π.
Notice that
¯
τ (ω) = τ (1 + ξ)τ (y) = (1 + ξ)ξy = (1 + ξ −1 )ξy = (1 + ξ)y = ω.
Thus, τ fixes both x and ω and τ (i) = −i, that is, τ is complex conjugation
with respect to the new coordinates. Moreover, it fixes points because the equation
ω g+2 = κ(x−1)xg−1 (x+1)g+2 admits a real solution for ω whenever −1 ≤ x ≤ 1.
To prove that γ also fixes points we will repeat the strategy above, that is, we
will find a new equation of the function field of Yg so that γ acts as complex conju-
gation with respect to the new coordinates. However, the process is now much more
involved and it requires many computations.
Recall that we defined g0 = (g + 1)/4 and consider the auxiliary functions
y2 1 1 v q
q= ; v = q g0 ; t=q+ ; r=v+ and s = + .
x(x + 1) q v q v
b(y 2 ) u2
b(q) = = 2
b(x)b(x + 1) y (−ixv )(1 − ixv 2 )
2
a(y)2 ξ2 y2
a(q) = = = ξ2 q
a(x)(1 + a(x)) x(1 + x)
which implies ag+1 (q) = ξ 2g+2 q = q and so a(t) = t. Therefore C(t) ⊂ Fix(M ).
Thus, to prove the equality it is enough to check that
Clearly, [C(q) : C(t)] = 2 because the irreducible polynomial Irr(q, C(t)) of q over
C(t) is T 2 − tT + 1. Also, [C(q, x) : C(q)] = 2 since x2 = 1/(1 − q g+1 ); finally
[C(x, y) : C(x, q)] = 2 because y 2 = x(x + 1)q. Hence,
as desired.
(5.4.10) Let us prove now that C(x, q) = Fix(ag+1 ). Indeed, a(x) = x and
Recall that b(v) = 1/v and b(x) = −ixv 2 . Hence, a direct computation gives
√
2(i − 1)v v14 − 1 (−ixv 2 )
b(f ) = = if.
2
(5.4.12) For later purposes let us prove that f 4 = −r2 (r2 − 4).
Of course,
(i − 1)4 (v 4 − 1)4 x4 −(v 4 − 1)4 x4
f4 = =
4v 4 v4
and (v 4 − 1)x2 = −1, that is,
2
−(v 4 − 1)2 (v 4 − 1)x2
4 −(v 4 − 1)2 2 1 2
f = = = − v −
v4 v4 v2
1 2
1
1
= − v 2 + 2 + 4 = − v 2 + 2 + 2 v 2 + 2 − 2 = −r2 (r2 − 4).
v v v
5.4 Algebraic Formulae for the Symmetries 139
[C(x, y) : C(t)] 8 8
[C(t, f ) : C(t)] = = = = 4.
[C(x, y) : C(t, f )] [C(x, y) : Fix(ag+1 )] |ag+1 |
Hence r2 (r2 − 4) is fixed by both ag+1 and b, that is, r2 (r2 − 4) ∈ C(t).
3
(5.4.14) The sum ψ0 = i=0 bi (y) is fixed by b and σ but ag+1 (ψ0 ) = −ψ0 .
The first assertion is evident since b has order 4. As to the third, recall that b
and ag+1 commute, by part (9) in Proposition 5.3.4. Moreover, a(y) = ξy, where
ξ = eπi/g+1 , and so ag+1 (y) = −y. Hence,
3 3 3
ag+1 (ψ0 ) = ag+1 bi (y) = bi (ag+1 (y)) = bi (−y) = −ψ0 .
i=0 i=0 i=0
For the last assertion, observe that 1 − v 4 = 1/x2 and y 2 = qx(x + 1). This way,
F2 (T ) = T 2 + 2f T + f 2 − r(s − 2).
Finally, we get
We should express the coefficients of Irr(ζ, C(t)) as rational functions of the vari-
able t. Notice that
1 1 v q 1
r=v+ = q g0 + g0 and s = + = q g0 −1 + g0 −1 .
v q q v q
[n/2]
1 n n−2k 2
fn (t) = t (t − 4)k ,
2n−1 2k
k=0
1 1
fn+1 (t) + fn−1 (t) = q n+1 + n+1 + q n−1 + n−1
q q
1
1
= q n+1 + n−1 + q n−1 + n+1
q q
1 1
1
= q qn + n + qn + n
q q q
= tfn (t).
Thus the sequence {fn (t)}n satisfies the second order recurrence relation
t + √t2 − 4 n
t − √t2 − 4 n
fn (t) = α +β .
2 2
α+β=2
Since f0 (t) = 2 and f1 (t) = t we get , that is, α = β = 1 and so
α−β=0
√ √
(t + t2 − 4)n + (t − t2 − 4)n
fn (t) = ,
2n
which yields, by Newton binomial, the formula in the statement.
As we pointed out, r = fg0 (t) and s = fg0 −1 (t). Moreover, g0 is odd because
g = 4g0 − 1 ≡ 3 (mod 8), and so
5.4 Algebraic Formulae for the Symmetries 143
(g0 −1)/2
1 g0 g0 −2k 2
r = r(t) = fg0 (t) = t (t − 4)k ∈ R[t], and
2g0 −1 2k
k=0
(g0 −1)/2
1 g0 − 1 g0 −1−2k 2
s = s(t) = fg0 −1 (t) = t (t − 4)k ∈ R[t].
2g0 −2 2k
k=0
Then we get the irreducible polynomial of ζ over C(t) or, equivalently, a new equa-
tion of the function field of Yg , after substituting these values of r(t) and s(t) in the
formula of (5.4.18), that is, P (ζ, t) = 0, where
2 2
P (ζ, t) = ζ 2 − r(s − 2) + r2 (r2 − 4) + 16r3 (r2 − 4)(s − 2)ζ 2 ∈ R[ζ, t].
Hence, to show that the symmetry γ = σ ◦ b fixes points, it suffices to see that the
polynomial P (ζ, t) has infinitely many real zeros.
For each θ ∈ (π/2, π) define v(θ) = e4θi/g+1 ∈ R. Then q(θ) = v(θ)(g+1)/4 =
θi
e and this implies
1
r(θ) = q(θ) + = eθi + e−θi and
q(θ)
v(θ) q(θ)
s(θ) = + = e(g−3)θi/(g+1) + e(3−g)θi/(g+1) .
q(θ) v(θ)
√
Thus −2 < r(θ) < 0 and −2 < s(θ) < 2 because g ≥ 7. Then r(θ)(s(θ)−2) > 0
and so there exists a real number ζ1 (θ) such that ζ1 (θ)2 = r(θ)(s(θ) − 2). Hence,
P (ζ1 (θ), t(θ)) = r4 (θ)(r2 (θ) − 4) r2 (θ) − 4 + 16(s(θ) − 2)2 < 0,
are simple. The Riemann surface SB with function field C(x, y), where y 2 = f (x),
is a hyperelliptic surface of genus (deg f −2)/2 = g. The automorphism of the field
extension C(x)|C defined by x → ξ 2 x induces a bijection on the set of roots of f ,
because (ξ 2k bi )n = bni . Thus, there exists an automorphism a of the field extension
C(x, y)|C satisfying a(x) = ξ 2 x.
(6.1.1) We look for a suitable choice of the set B such that the surface SB admits
an anticonformal automorphism τ with τ 2 = a. This implies, in particular, that τ is
not a symmetry. Since the covering
SB → Σ = SB /a ; (x, y) → x
ramifies over x = 0 and x = ∞, the automorphism τ must switch these two points.
Hence, there should exist a complex number α such that τ (x) = 1/(αx). In fact,
1 αx
ξ 2 x = a(x) = τ 2 (x) = τ = .
αx ᾱ
n n
−1 ¯ n −1 ¯ n −b¯i 1
τ (x −
n
bni ) = − bi = n − bi = n n
x + n and
ξx x x b¯i
n
1 −1 1 1 1 1
τ xn + n = + n = n − n = n n (xn − bni )
¯
bi ξx bi bi x bi x
1
f (x) = (x −
n
bni ) n
x + n .
i=1
b¯i
6.1 Compact Riemann Surfaces Without Symmetries 147
¯
Moreover, let c ∈ C be a square root of the product n
i=1 (−bi /bi ) . Then
1
2
τ (y) = τ (f (x)) = τ (x − bi ) x + n
n n n
i=1
b¯i
¯ n
−b i f (x) cy 2
= = .
i=1
bi x2n xg+1
Henceforth, the field extension C(x, y)|R associated to the surface SB admits the
anticonformal automorphism τ defined by
−1 cy
τ (i) = −i; τ (x) = ; τ (y) = .
ξx xg+1
or, equivalently, j(g + 1) must be divisible by n but not by 2n, and 2j must be a
multiple of 2n. This is impossible because g + 1 is even.
The third statement is evident because ρ commutes with a. Finally, τ 2n = an = 1
and, as n is even,
|τ |
n = |a| = |τ 2 | = =⇒ |τ | = 2n,
gcd(2, |τ |)
and |aρ| = lcm(|a|, |ρ|) = n, where lcm stands for the least common multiple.
(6.1.3) Let H = ρ be the subgroup generated by the hyperelliptic involution
and let G+ be the subgroup of the group G = τ ⊕ H = Z2n ⊕ Z2 consisting of
the conformal elements of G. We claim that G = Aut(SB ) unless the factor group
148 6 Appendix
Aut+ (SB )/H contains a subgroup containing aH and which is isomorphic either
to the dihedral group Dn or to the cyclic group Zkn for some integer k ≥ 2.
Indeed, suppose that G Aut(SB ). Then, the difference Aut(SB ) \ G contains
some conformal automorphism, and so also Zn = G+ /H Aut+ (SB )/H. But
this last factor group is a finite group of automorphisms of the Riemann sphere
SB /H and so Aut+ (SB )/H is isomorphic either to the symmetric group S4 , or to
the alternating groups A4 , A5 , or to the dihedral group Dm or to the cyclic group
Zm , for some multiple m of n.
Assume there exists a multiple m of n such that either Zm = Aut+ (SB )/H or
Dm = Aut+ (SB )/H. In the first case
and so Aut+ (SB )/H, which contains aH, is isomorphic to Zkn for some integer
k ≥ 2. In the second case also Dn ≤ Dm = Aut+ (SB )/H, and this proves our
claim in these two cases. Thus we may suppose that Aut+ (SB )/H A4 , S4 or
A5 , and so n = 2, 4, because S4 , A4 and A5 contain no element of order ≥ 6.
Assume first that n = 2. The Sylow 2-subgroups of A4 and A5 are isomorphic to
the Klein group Z2 ⊕ Z2 , while the Sylow 2-subgroups of S4 are isomorphic to the
dihedral group D4 . In either case the Sylow 2-subgroup of Aut+ (SB )/H containing
aH is isomorphic to Z2 and it is contained either in Z4 or in D2 .
Finally, if n = 4 then Aut+ (SB )/H = A4 or A5 , because these last groups have
no element of order 4. Thus Aut+ (SB )/H S4 and so it contains the dihedral
group D4 . This completes the proof of our claim.
(6.1.4) We are ready to determine explicit conditions on the set B = {b1 , . . . , b2 }
which imply that Aut(SB ) = τ ⊕ H = Z2n ⊕ Z2 . Notice that if this is the case
then the surface SB admits no symmetry. Indeed, the anticonformal automorphisms
of SB are of the form τ j ρ for some odd integer j. If one of them were a sym-
metry, its square aj = τ 2j would be the identity, that is, n would divide j, a
contradiction.
Hence, our task now is to find conditions on the set B such that the factor group
Aut+ (SB )/H contains no subgroup containing aH and which is isomorphic either
to the dihedral group Dn or to the cyclic group Zkn for some integer k ≥ 2.
First assume that Aut+ (SB )/H Zkn . In this case x = 0 and x = ∞ are
fixed by a conformal automorphism of SB of order kn. Thus, if ω = e2πi/kn , the
map x → ωx is an automorphism of order kn of the field extension C(x)|C which
lifts to C(x, y). Therefore, by (6.1.1), for each factor xn − bni of f, the polynomials
kn
xkn − bkn i and x
kn
− (−1)k /b¯i are also factors of f . Hence, for some divisor k
of , the surface SB should admit an equation of the form
/k
(−1)k+1
y2 = (xkn − bkn
i ) x
kn
+ kn
. (6.1)
i=1 b¯i
6.1 Compact Riemann Surfaces Without Symmetries 149
Assume now that Aut+ (SB )/H Dn . Then SB admits a conformal automorphism
of order 2 that interchanges x = 0 and x = ∞. Therefore the automorphism x →
1/x of C(x)|C induces a bijection of the set B. Hence, for each factor xn − bni of
f there exists an index j such that anj = 1/ani or −1/a¯j n = a¯i n . Consequently,
must be even and the surface SB should admit an equation of the form
/2
2 1 1 n
y = (x −
n
bni ) x − n
n
x + n (xn + b¯i ).
n
(6.2)
bi ¯
bi
i=1
−1 cy
τ (i) = −i; τ (x) = ; τ (y) = .
ξx xg+1
(3) Suppose that a defining equation for SB cannot be expressed in the form (6.1)
or (6.2). Then τ ⊕ ρ = Z2n ⊕ Z2 is the full group of automorphisms of SB .
In particular, this surface admits no symmetry.
References
151
152 References
46. GAP-Groups, algorithms and programming. Version 3 Release 4.4 (1997) (Lehrstuhl D für
Mathematik, RWTH, Aachen, Germany).
47. L. Greenberg: Maximal Fuchsian groups. Bull. Am. Math. Soc. 69, 569–573 (1963).
48. G. Gromadzki: Groups of Automorphisms of Compact Riemann and Klein Surfaces.
University Press, WSP, Bydgoszcz (1993).
49. G. Gromadzki: On a Harnack-Natanzon theorem for the family of real forms of Riemann
surfaces. J. Pure Appl. Alg. 121, 253–269 (1997).
50. G. Gromadzki: On ovals on Riemann surfaces. Rev. Mat. Iberoam.16(3), 515–527 (2000).
51. G. Gromadzki: Symmetries of Riemann surfaces from a combinatorial point of view. London
Math. Soc. Lecture Note Series 287, 91–112. Cambridge University Press, Cambridge (2001).
52. G. Gromadzki: On Singerman symmetries of a class of Belyi Riemann surfaces. J. Pure Appl.
Alg. 213(10), 1905–1910 (2009).
53. G. Gromadzki, M. Izquierdo: Real forms of a Riemann surface of even genus. Proc. Amer.
Math. Soc. 126(12), 3475–3479 (1998).
54. G. Gromadzki, M. Izquierdo: On ovals of Riemann surfaces of even genera. Geometriae Ded-
icata 78, 81–88 (1999).
55. G. Gromadzki, E. Kozłowska-Walania: On fixed points of doubly symmetric Riemann sur-
faces. Glasg. Math. J. 50(3), 371–378 (2008).
56. G. Gromadzki, E. Kozłowska-Walania: On ovals of non-conjugate symmetries of Riemann
surfaces. Int. J. Math. 20(1), 1–13 (2009).
57. B. H. Gross, J. Harris: Real algebraic curves. Ann. Sci. École Norm. Sup.14, 157–182 (1981).
58. W. Hall: Automorphisms and coverings of Klein surfaces. Ph. D. Thesis. Southampton
University (1977).
59. A. Harnack: Über die Vieltheiligkeit der ebenen algebraischen Kurven. Math Ann. 10,
189–198 (1876).
60. A. H. M. Hoare: Subgroups of NEC groups and finite permutation groups. Quarterly J. Math.
Oxford (2) 41, 45–59 (1990).
61. A. H. M. Hoare, D. Singerman: The orientability of subgroups of plane groups. London Math.
Soc. Lecture Note Series 71, 221–227 (1982).
62. J. Huisman, M. Lattarulo: Imaginary automorphisms on real hyperelliptic curves. J. Pure
Appl. Algebra 200(3), 318–331 (2005).
63. M. Izquierdo, D. Singerman: Pairs of symmetries of Riemann surfaces. Ann. Acad. Sci. Fenn.
Math. 23(1), 3–24 (1998).
64. L. Keen: On Fricke moduli. Advances in the theory of Riemann surfaces. Ann. Math. Stud.
66, 205–224 (1971).
65. F. Klein: Über die Transformationen siebenter Ordnung der elliptischen Functionen. Math.
Ann. 14, 428–471 (1879).
66. B. Köck, D. Singerman: Real Belyi theory. Q. J. Math. 58, 463–478 (2007).
67. B. Köck, E. Lau: A note on Belyi’s theorem for Klein surfaces. Q. J. Math. 61, 103–107
(2010).
68. E. Kozłowska-Walania: On p-hyperellipticity of doubly symmetric Riemann surfaces. Publi-
cacions Matematiques 51, 291–307 (2007).
69. E. Kozłowska-Walania: On commutativity and ovals for a pair of symmetries of a Riemann
surface. Colloq. Math. 109, 61–69 (2007).
70. R. S. Kulkarni: A note on Wiman and Accola-Maclachlan surfaces. Ann. Acad. Sci. Fenn. 16,
83–94 (1991).
71. A. G. Kurosch: Gruppentheorie. Berlin (1953).
72. A. M. Macbeath: On a curve of genus 7. Proc. Lond. Math. Soc. (3) 15, 527–542 (1965).
73. A. M. Macbeath: The classification of non-euclidean plane crystallographic groups. Can.
J. Math. 19, 1192–1205 (1967).
74. A. M. Macbeath: Generators of the linear fractional groups. Number Theory Proc. Symposia
in Pure Mathematics 12. American Mathematical Society, Providence, 14–32 (1969).
75. A. M. Macbeath: Action of automorphisms of a compact Riemann surface on the first
homology group. Bull. Lond. Math. Soc. 5, 103–108 (1973).
154 References
76. A. M. Macbeath, D. Singerman: Spaces of subgroups and Teichmüller space. Proc. Lond.
Math. Soc. 31(3), 211–256 (1975).
77. C. Maclachlan: A bound for the number of automorphisms of a compact Riemann surface.
J. Lond. Math. Soc. 44, 265–272 (1969).
78. B. Maskit: Remarks on m-symmetric Riemann surfaces. Lipa’s legacy. Contemp. Math. 211,
433–445 (1995).
79. C. L. May: Automorphisms of compact Klein surfaces with boundary. Pac. J. Math. 59,
199–210 (1975).
80. C. L. May: Cyclic automorphism groups of compact bordered Klein surfaces. Houst. J. Math.
3, 395–405 (1977).
81. C. L. May: A bound for the number of automorphisms of a compact Klein surface with bound-
ary. Proc. Am. Math. Soc. 63, 273–280 (1977).
82. C. L. May: Large automorphism groups of compact Klein surfaces with boundary I. Glasg.
Math. J. 18, 1–10 (1977).
83. C. L. May: Maximal symmetry and fully wound coverings. Proc. Am. Math. Soc. 79, 23–31
(1980).
84. C. L. May: The species of Klein surfaces with maximal symmetry of low genus. Pac. J. Math.
111, 371–394 (1984).
85. C. L. May: A family of M ∗ -groups. Can. J. Math. 38, 1094–1109 (1986).
86. C. L. May: Nilpotent automorphism groups of bordered Klein surfaces. Proc. Am. Math. Soc.
101, 287–292 (1987).
87. C. L. May: Supersolvable M ∗ -groups. Glasg. Math. J. 30, 31–40 (1988).
88. C. L. May. Complex doubles of bordered Klein surfaces with maximal symmetry. Glasg.
Math. J. 33, 61–67 (1991).
89. C. L. May: The Groups of Real Genus 4. Mich. Math. J. 39, 219–228 (1992).
90. C. L. May: Finite groups acting on bordered surfaces and the real genus of a group. Rocky Mt.
J. Math. 23, 707–724 (1993).
91. A. D. Mednykh: Hyperbolic Riemann surfaces with the trivial group of automorphisms. De-
formations of mathematical structures (Łdź/Lublin, 1985/87), 115–125, Kluwer Acad. Publ.,
Dordrecht (1989).
92. R. Miranda: Algebraic Curves and Riemann Surfaces. Graduate Studies in Mathematics. 5.
American Mathematical Society (1995).
93. B. Mockiewicz: Real Genus 12, Rocky Mt. J. Math 34(4), 1391–1398 (2004).
94. G. Nakamura: The existence of symmetric Riemann surfaces determined by cyclic groups.
Nagoya Math. J. 151, 129–143 (1998).
95. S. M. Natanzon: On the order of a finite group of homeomorphisms of a surface into itself and
the number of real forms of a complex algebraic curve. Dokl. Akad. Nauk SSSR 242, 765–768
(1978). Soviet Math. Dokl. 19(5), 1195–1199 (1978).
96. S. M. Natanzon: Automorphisms of the Riemann surface of an M -curve. (Russian) Funkt-
sional. Anal. i Prilozhen.12(3), 82–83 (1978). Functional Anal. Appl. 12, 228–229 (1978).
97. S. M. Natanzon: Moduli spaces of real curves. Trudy Moskov. Mat. Obshch. 37, 219–253
(1978). Trans. Moscow Math. Soc. 1, 233–272 (1980).
98. S. M. Natanzon: Lobachevskii geometry and automorphisms of complex M -curves. Geomet-
ric methods in problems of analysis and algebra, (Yaroslav), 130–151 (1978). Selecta Math.
Soviet. 1(1), 81–99 (1981).
99. S. M. Natanzon: Automorphisms and real forms of a class of complex algebraic curves. Funk-
tsional Anal. i Priloz. 13(2), 89–90 (1979). Funct. Anal. Appl. 13, 148–150 (1979).
100. S. M. Natanzon: On the total number of ovals of real forms of complex algebraic curves.
Uspekhi Mat. Nauk. 35(1), 207–208 (1980). Russ. Math. Surveys 35(1), 223–224 (1980).
101. S. M. Natanzon: On the total number of ovals of four complex-isomorphic real algebraic
curves. Uspekhi Mat. Nauk. 35(4), 184 (1980). Russ. Math. Surveys 35(4), 177 (1980).
102. S. M. Natanzon: Topological classification of pairs of commuting antiholomorphic involu-
tions of Riemann surfaces. Uspekhi Mat. Nauk. 41(5), 191–192 (1986). Russ. Math. Surveys
41(5), 159–160 (1986).
References 155
103. S. M. Natanzon: Uniformization of spaces of real meromorphic functions. Dokl. Akad. Nauk.
SSSR 287, 1058–1061 (1986). Sov. Math. Dokl. 33, 487–490 (1986).
104. S. M. Natanzon: Real meromorphic functions on real algebraic curves. Dokl. Akad. Nauk.
SSSR 297, 40–43 (1987). Sov. Math. Dokl. 36, 425–427 (1988).
105. S. M. Natanzon: Finite groups of homeomorphisms of surfaces and real forms of complex
algebraic curves. Trudy Moskov. Mat. Obshch. 51, 3–53 (1988). Trans. Moscow Math. Soc.
51, 1–51 (1988).
106. S. M. Natanzon: Klein surfaces. Uspekhi Mat. Nauk. 45(6), 47–90 (1990). Russ. Math. Sur-
veys 45(6), 43–108 (1990).
107. S. M. Natanzon: Geometry and algebra of real forms of complex curves. Math. Zeit. 243,
391–407 (2003).
108. S. M. Natanzon: Moduli of Riemann surfaces, real algebraic curves, and their superanalogs.
Translated from the 2003 Russian edition by Sergei Lando. Translations of Mathematical
Monographs, 225. American Mathematical Society, Providence (2004).
109. K. Reidemeister: Automorphismen von Homotopiekettenringen. Math. Ann. 112(1), 586–593
(1936).
110. M. Seppälä: Complex algebraic curves with real moduli. J. Reine Angew. Math. 387, 209–220
(1988).
111. M. Seppälä: Real algebraic curves in the moduli space of complex curves. Compos. Math. 74,
259–283 (1990).
112. M. Seppälä: Moduli spaces of real algebraic curves.Topics on Riemann surfaces and Fuchsian
groups. London Math. Soc. Lecture Note Series 287, 133–153, Cambridge University Press,
Cambridge (2001).
113. M. Seppälä, T. Sorvali, Geometry of Riemann surfaces and Teichmüller spaces, North-
Holland Mathematics Studies, 169, Amsterdam (1992).
114. R. J. Sibner: Symmetric Fuchsian groups. Amer. J. Math. 90, 1237–1259 (1968).
115. D. Singerman: Non-Euclidean crystallographic groups and Riemann surfaces. Ph. D. Thesis,
University of Birmingham (1969).
116. D. Singerman: Subgroups of Fuchsian groups and finite permutation groups. Bull. Lond.
Math. Soc. 2, 319–323 (1970).
117. D. Singerman: Finitely generated maximal Fuchsian groups. J. Lond. Math. Soc. 6, 29–38
(1972).
118. D. Singerman: Symmetries of Riemann Surfaces with Large Automorphism Group. Math.
Ann. 210, 17–32 (1974).
119. D. Singerman: On the structure of non-euclidean crystallographic groups. Proc. Camb. Phil.
Soc. 76, 233–240 (1974).
120. D. Singerman: Symmetries and pseudosymmetries of hyperelliptic surfaces. Glasg. Math. J.
21, 39–49 (1980).
121. D. Singerman: Mirrors on Riemann surfaces. Contemp. Math. 184, 411–417 (1995).
122. M. Suzuki: Group Theory I. Grundlehren der mathematischen Wissenschaften 247. Springer,
Berlin (1982).
123. P. Turbek: An explicit family of curves with trivial automorphism groups. Proc. Am. Math.
Soc. 122(2), 657–664 (1994).
124. P. Turbek: The full automorphism group of the Kulkarni surface. Rev. Mat. Univ. Complut.
Madrid 10(2), 265–276 (1997).
125. P. Turbek: Algebraic curves, Riemann surfaces and Klein surfaces with no non-trivial auto-
morphisms or symmetries. Proc. Edinb. Math. Soc. 45(1), 141–148 (2002).
126. E. Tyszkowska: On Macbeath-Singerman symmetries of Belyi surfaces with PSL(2, p) as a
group of automorphisms. Cent. Eur. J. Math. 1(2), 208–220 (2003).
127. G. Weichold: Über symmetrische Riemannsche Flächen und die Periodizitätsmodulen der
zugehörigen Abelschen Normalintegrale erstes Gattung. Leipziger Dissertation (1883).
128. H. C. Wilkie: On non-euclidean crystallographic groups. Math. Zeit. 91, 87–102 (1966).
Index
H
A Hurwitz automorphism group, 95
abstractly orientable group, 21 Hurwitz group, 95
orientation in an, 21 Hurwitz-Riemann formula, 6
abstractly oriented group, 21 hyperelliptic involution, 82
Accola-Maclachlan surface, xviii, 110 hyperelliptic Riemann surface, 82
area
of a signature, 5
of an NEC group, 5 I
asymmetric surface, xix isotropy group, 91
automorphism
group of a Riemann surface, 16
of a Riemann surface, 16 K
of the hyperbolic plane, 2 Klein quartic, xvii, 95
Klein surface, xii, 11
Kulkarni surface, xviii, 110
B
bordered surface NEC group, 11
L
lattice, 69
C hexagonal, 71
canonical Fuchsian group, 2 square, 71
157
158 Index
R
real algebraic curve, xi T
purely imaginary, xii Teichmüller space, 14
real form, xii topological type of a symmetry, xii, 15
real structure of a Riemann surface, 15 topologically equivalent actions, 12
Lecture Notes in Mathematics
For information about earlier volumes
please contact your bookseller or Springer
LNM Online archive: springerlink.com
Vol. 1816: S. Albeverio, W. Schachermayer, M. Tala- Vol. 1837: S. Tavaré, O. Zeitouni, Lectures on Probabil-
grand, Lectures on Probability Theory and Statistics. ity Theory and Statistics. Ecole d’Eté de Probabilités de
Ecole d’Eté de Probabilités de Saint-Flour XXX-2000. Saint-Flour XXXI-2001. Editor: J. Picard (2004)
Editor: P. Bernard (2003) Vol. 1838: A.J. Ganesh, N.W. O’Connell, D.J. Wischik,
Vol. 1817: E. Koelink, W. Van Assche (Eds.), Orthogonal Big Queues. XII, 254 p, 2004.
Polynomials and Special Functions. Leuven 2002 (2003) Vol. 1839: R. Gohm, Noncommutative Stationary Pro-
Vol. 1818: M. Bildhauer, Convex Variational Problems cesses. VIII, 170 p, 2004.
with Linear, nearly Linear and/or Anisotropic Growth Vol. 1840: B. Tsirelson, W. Werner, Lectures on Probabil-
Conditions (2003) ity Theory and Statistics. Ecole d’Eté de Probabilités de
Vol. 1819: D. Masser, Yu. V. Nesterenko, H. P. Schlick- Saint-Flour XXXII-2002. Editor: J. Picard (2004)
ewei, W. M. Schmidt, M. Waldschmidt, Diophantine Vol. 1841: W. Reichel, Uniqueness Theorems for Vari-
Approximation. Cetraro, Italy 2000. Editors: F. Amoroso, ational Problems by the Method of Transformation
U. Zannier (2003) Groups (2004)
Vol. 1820: F. Hiai, H. Kosaki, Means of Hilbert Space Vol. 1842: T. Johnsen, A. L. Knutsen, K3 Projective Mod-
Operators (2003) els in Scrolls (2004)
Vol. 1821: S. Teufel, Adiabatic Perturbation Theory in Vol. 1843: B. Jefferies, Spectral Properties of Noncom-
Quantum Dynamics (2003) muting Operators (2004)
Vol. 1822: S.-N. Chow, R. Conti, R. Johnson, J. Mallet- Vol. 1844: K.F. Siburg, The Principle of Least Action in
Paret, R. Nussbaum, Dynamical Systems. Cetraro, Italy Geometry and Dynamics (2004)
2000. Editors: J. W. Macki, P. Zecca (2003) Vol. 1845: Min Ho Lee, Mixed Automorphic Forms, Torus
Vol. 1823: A. M. Anile, W. Allegretto, C. Ringhofer, Bundles, and Jacobi Forms (2004)
Mathematical Problems in Semiconductor Physics. Vol. 1846: H. Ammari, H. Kang, Reconstruction of Small
Cetraro, Italy 1998. Editor: A. M. Anile (2003) Inhomogeneities from Boundary Measurements (2004)
Vol. 1824: J. A. Navarro González, J. B. Sancho de Salas, Vol. 1847: T.R. Bielecki, T. Bjrk, M. Jeanblanc, M.
C ∞ – Differentiable Spaces (2003) Rutkowski, J.A. Scheinkman, W. Xiong, Paris-Princeton
Vol. 1825: J. H. Bramble, A. Cohen, W. Dahmen, Mul- Lectures on Mathematical Finance 2003 (2004)
tiscale Problems and Methods in Numerical Simulations, Vol. 1848: M. Abate, J. E. Fornaess, X. Huang, J. P. Rosay,
Martina Franca, Italy 2001. Editor: C. Canuto (2003) A. Tumanov, Real Methods in Complex and CR Geom-
Vol. 1826: K. Dohmen, Improved Bonferroni Inequal- etry, Martina Franca, Italy 2002. Editors: D. Zaitsev, G.
ities via Abstract Tubes. Inequalities and Identities of Zampieri (2004)
Inclusion-Exclusion Type. VIII, 113 p, 2003. Vol. 1849: Martin L. Brown, Heegner Modules and Ellip-
Vol. 1827: K. M. Pilgrim, Combinations of Complex tic Curves (2004)
Dynamical Systems. IX, 118 p, 2003. Vol. 1850: V. D. Milman, G. Schechtman (Eds.), Ge-
Vol. 1828: D. J. Green, Grbner Bases and the Computation ometric Aspects of Functional Analysis. Israel Seminar
of Group Cohomology. XII, 138 p, 2003. 2002-2003 (2004)
Vol. 1829: E. Altman, B. Gaujal, A. Hordijk, Discrete- Vol. 1851: O. Catoni, Statistical Learning Theory and
Event Control of Stochastic Networks: Multimodularity Stochastic Optimization (2004)
and Regularity. XIV, 313 p, 2003. Vol. 1852: A.S. Kechris, B.D. Miller, Topics in Orbit
Vol. 1830: M. I. Gil’, Operator Functions and Localization Equivalence (2004)
of Spectra. XIV, 256 p, 2003. Vol. 1853: Ch. Favre, M. Jonsson, The Valuative Tree
Vol. 1831: A. Connes, J. Cuntz, E. Guentner, N. Hig- (2004)
son, J. E. Kaminker, Noncommutative Geometry, Martina Vol. 1854: O. Saeki, Topology of Singular Fibers of Dif-
Franca, Italy 2002. Editors: S. Doplicher, L. Longo (2004) ferential Maps (2004)
Vol. 1832: J. Azéma, M. Émery, M. Ledoux, M. Yor Vol. 1855: G. Da Prato, P.C. Kunstmann, I. Lasiecka,
(Eds.), Séminaire de Probabilités XXXVII (2003) A. Lunardi, R. Schnaubelt, L. Weis, Functional Analytic
Vol. 1833: D.-Q. Jiang, M. Qian, M.-P. Qian, Mathe- Methods for Evolution Equations. Editors: M. Iannelli,
matical Theory of Nonequilibrium Steady States. On the R. Nagel, S. Piazzera (2004)
Frontier of Probability and Dynamical Systems. IX, 280 Vol. 1856: K. Back, T.R. Bielecki, C. Hipp, S. Peng,
p, 2004. W. Schachermayer, Stochastic Methods in Finance, Bres-
Vol. 1834: Yo. Yomdin, G. Comte, Tame Geometry with sanone/Brixen, Italy, 2003. Editors: M. Fritelli, W. Rung-
Application in Smooth Analysis. VIII, 186 p, 2004. galdier (2004)
Vol. 1835: O.T. Izhboldin, B. Kahn, N.A. Karpenko, Vol. 1857: M. Émery, M. Ledoux, M. Yor (Eds.),
A. Vishik, Geometric Methods in the Algebraic Theory Séminaire de Probabilités XXXVIII (2005)
of Quadratic Forms. Summer School, Lens, 2000. Editor: Vol. 1858: A.S. Cherny, H.-J. Engelbert, Singular Stochas-
J.-P. Tignol (2004) tic Differential Equations (2005)
Vol. 1836: C. Nǎstǎsescu, F. Van Oystaeyen, Methods of Vol. 1859: E. Letellier, Fourier Transforms of Invariant
Graded Rings. XIII, 304 p, 2004. Functions on Finite Reductive Lie Algebras (2005)
Vol. 1860: A. Borisyuk, G.B. Ermentrout, A. Friedman, Vol. 1884: N. Hayashi, E.I. Kaikina, P.I. Naumkin,
D. Terman, Tutorials in Mathematical Biosciences I. I.A. Shishmarev, Asymptotics for Dissipative Nonlinear
Mathematical Neurosciences (2005) Equations (2006)
Vol. 1861: G. Benettin, J. Henrard, S. Kuksin, Hamilto- Vol. 1885: A. Telcs, The Art of Random Walks (2006)
nian Dynamics – Theory and Applications, Cetraro, Italy, Vol. 1886: S. Takamura, Splitting Deformations of Dege-
1999. Editor: A. Giorgilli (2005) nerations of Complex Curves (2006)
Vol. 1862: B. Helffer, F. Nier, Hypoelliptic Estimates and Vol. 1887: K. Habermann, L. Habermann, Introduction to
Spectral Theory for Fokker-Planck Operators and Witten Symplectic Dirac Operators (2006)
Laplacians (2005) Vol. 1888: J. van der Hoeven, Transseries and Real Differ-
Vol. 1863: H. Führ, Abstract Harmonic Analysis of Con- ential Algebra (2006)
tinuous Wavelet Transforms (2005) Vol. 1889: G. Osipenko, Dynamical Systems, Graphs, and
Vol. 1864: K. Efstathiou, Metamorphoses of Hamiltonian Algorithms (2006)
Systems with Symmetries (2005) Vol. 1890: M. Bunge, J. Funk, Singular Coverings of
Vol. 1865: D. Applebaum, B.V. R. Bhat, J. Kustermans, Toposes (2006)
J. M. Lindsay, Quantum Independent Increment Processes Vol. 1891: J.B. Friedlander, D.R. Heath-Brown,
I. From Classical Probability to Quantum Stochastic Cal- H. Iwaniec, J. Kaczorowski, Analytic Number Theory,
culus. Editors: M. Schürmann, U. Franz (2005) Cetraro, Italy, 2002. Editors: A. Perelli, C. Viola (2006)
Vol. 1892: A. Baddeley, I. Bárány, R. Schneider, W. Weil,
Vol. 1866: O.E. Barndorff-Nielsen, U. Franz, R. Gohm,
Stochastic Geometry, Martina Franca, Italy, 2004. Editor:
B. Kümmerer, S. Thorbjønsen, Quantum Independent
W. Weil (2007)
Increment Processes II. Structure of Quantum Lévy
Vol. 1893: H. Hanßmann, Local and Semi-Local Bifur-
Processes, Classical Probability, and Physics. Editors: M.
cations in Hamiltonian Dynamical Systems, Results and
Schürmann, U. Franz, (2005)
Examples (2007)
Vol. 1867: J. Sneyd (Ed.), Tutorials in Mathematical Bio-
Vol. 1894: C.W. Groetsch, Stable Approximate Evaluation
sciences II. Mathematical Modeling of Calcium Dynamics
of Unbounded Operators (2007)
and Signal Transduction. (2005)
Vol. 1895: L. Molnár, Selected Preserver Problems on
Vol. 1868: J. Jorgenson, S. Lang, Posn (R) and Eisenstein Algebraic Structures of Linear Operators and on Function
Series. (2005) Spaces (2007)
Vol. 1869: A. Dembo, T. Funaki, Lectures on Probabil- Vol. 1896: P. Massart, Concentration Inequalities and
ity Theory and Statistics. Ecole d’Eté de Probabilités de Model Selection, Ecole d’Été de Probabilités de Saint-
Saint-Flour XXXIII-2003. Editor: J. Picard (2005) Flour XXXIII-2003. Editor: J. Picard (2007)
Vol. 1870: V.I. Gurariy, W. Lusky, Geometry of Mntz Vol. 1897: R. Doney, Fluctuation Theory for Lévy Pro-
Spaces and Related Questions. (2005) cesses, Ecole d’Été de Probabilités de Saint-Flour XXXV-
Vol. 1871: P. Constantin, G. Gallavotti, A.V. Kazhikhov, 2005. Editor: J. Picard (2007)
Y. Meyer, S. Ukai, Mathematical Foundation of Turbu- Vol. 1898: H.R. Beyer, Beyond Partial Differential Equa-
lent Viscous Flows, Martina Franca, Italy, 2003. Editors: tions, On linear and Quasi-Linear Abstract Hyperbolic
M. Cannone, T. Miyakawa (2006) Evolution Equations (2007)
Vol. 1872: A. Friedman (Ed.), Tutorials in Mathemati- Vol. 1899: Séminaire de Probabilités XL. Editors:
cal Biosciences III. Cell Cycle, Proliferation, and Cancer C. Donati-Martin, M. Émery, A. Rouault, C. Stricker
(2006) (2007)
Vol. 1873: R. Mansuy, M. Yor, Random Times and En- Vol. 1900: E. Bolthausen, A. Bovier (Eds.), Spin Glasses
largements of Filtrations in a Brownian Setting (2006) (2007)
Vol. 1874: M. Yor, M. Émery (Eds.), In Memoriam Vol. 1901: O. Wittenberg, Intersections de deux
Paul-Andr Meyer - Sminaire de Probabilits XXXIX quadriques et pinceaux de courbes de genre 1, Inter-
(2006) sections of Two Quadrics and Pencils of Curves of Genus
Vol. 1875: J. Pitman, Combinatorial Stochastic Processes. 1 (2007)
Ecole d’Et de Probabilits de Saint-Flour XXXII-2002. Ed- Vol. 1902: A. Isaev, Lectures on the Automorphism
itor: J. Picard (2006) Groups of Kobayashi-Hyperbolic Manifolds (2007)
Vol. 1876: H. Herrlich, Axiom of Choice (2006) Vol. 1903: G. Kresin, V. Maz’ya, Sharp Real-Part Theo-
rems (2007)
Vol. 1877: J. Steuding, Value Distributions of L-Functions
Vol. 1904: P. Giesl, Construction of Global Lyapunov
(2007)
Functions Using Radial Basis Functions (2007)
Vol. 1878: R. Cerf, The Wulff Crystal in Ising and Percol-
Vol. 1905: C. Prévôt, M. Röckner, A Concise Course on
ation Models, Ecole d’Et de Probabilités de Saint-Flour
Stochastic Partial Differential Equations (2007)
XXXIV-2004. Editor: Jean Picard (2006)
Vol. 1906: T. Schuster, The Method of Approximate
Vol. 1879: G. Slade, The Lace Expansion and its Appli- Inverse: Theory and Applications (2007)
cations, Ecole d’Et de Probabilits de Saint-Flour XXXIV- Vol. 1907: M. Rasmussen, Attractivity and Bifurcation for
2004. Editor: Jean Picard (2006) Nonautonomous Dynamical Systems (2007)
Vol. 1880: S. Attal, A. Joye, C.-A. Pillet, Open Quantum Vol. 1908: T.J. Lyons, M. Caruana, T. Lévy, Differential
Systems I, The Hamiltonian Approach (2006) Equations Driven by Rough Paths, Ecole d’Été de Proba-
Vol. 1881: S. Attal, A. Joye, C.-A. Pillet, Open Quantum bilités de Saint-Flour XXXIV-2004 (2007)
Systems II, The Markovian Approach (2006) Vol. 1909: H. Akiyoshi, M. Sakuma, M. Wada,
Vol. 1882: S. Attal, A. Joye, C.-A. Pillet, Open Quantum Y. Yamashita, Punctured Torus Groups and 2-Bridge Knot
Systems III, Recent Developments (2006) Groups (I) (2007)
Vol. 1883: W. Van Assche, F. Marcellàn (Eds.), Orthogo- Vol. 1910: V.D. Milman, G. Schechtman (Eds.), Geo-
nal Polynomials and Special Functions, Computation and metric Aspects of Functional Analysis. Israel Seminar
Application (2006) 2004-2005 (2007)
Vol. 1911: A. Bressan, D. Serre, M. Williams, and Algebraic Surfaces. Cetraro, Italy 2003. Editors:
K. Zumbrun, Hyperbolic Systems of Balance Laws. F. Catanese, G. Tian (2008)
Cetraro, Italy 2003. Editor: P. Marcati (2007) Vol. 1939: D. Boffi, F. Brezzi, L. Demkowicz, R.G. Durán,
Vol. 1912: V. Berinde, Iterative Approximation of Fixed R.S. Falk, M. Fortin, Mixed Finite Elements, Compati-
Points (2007) bility Conditions, and Applications. Cetraro, Italy 2006.
Vol. 1913: J.E. Marsden, G. Misiołek, J.-P. Ortega, Editors: D. Boffi, L. Gastaldi (2008)
M. Perlmutter, T.S. Ratiu, Hamiltonian Reduction by Vol. 1940: J. Banasiak, V. Capasso, M.A.J. Chaplain,
Stages (2007) M. Lachowicz, J. Miȩkisz, Multiscale Problems in the Life
Vol. 1914: G. Kutyniok, Affine Density in Wavelet Sciences. From Microscopic to Macroscopic. Bȩdlewo,
Analysis (2007) Poland 2006. Editors: V. Capasso, M. Lachowicz (2008)
Vol. 1915: T. Bıyıkoǧlu, J. Leydold, P.F. Stadler, Laplacian Vol. 1941: S.M.J. Haran, Arithmetical Investigations.
Eigenvectors of Graphs. Perron-Frobenius and Faber- Representation Theory, Orthogonal Polynomials, and
Krahn Type Theorems (2007) Quantum Interpolations (2008)
Vol. 1916: C. Villani, F. Rezakhanlou, Entropy Methods Vol. 1942: S. Albeverio, F. Flandoli, Y.G. Sinai, SPDE in
for the Boltzmann Equation. Editors: F. Golse, S. Olla Hydrodynamic. Recent Progress and Prospects. Cetraro,
(2008) Italy 2005. Editors: G. Da Prato, M. Rckner (2008)
Vol. 1917: I. Veselić, Existence and Regularity Properties Vol. 1943: L.L. Bonilla (Ed.), Inverse Problems and Imag-
of the Integrated Density of States of Random Schrdinger ing. Martina Franca, Italy 2002 (2008)
(2008) Vol. 1944: A. Di Bartolo, G. Falcone, P. Plaumann,
Vol. 1918: B. Roberts, R. Schmidt, Local Newforms for K. Strambach, Algebraic Groups and Lie Groups with
GSp(4) (2007) Few Factors (2008)
Vol. 1919: R.A. Carmona, I. Ekeland, A. Kohatsu- Vol. 1945: F. Brauer, P. van den Driessche, J. Wu (Eds.),
Higa, J.-M. Lasry, P.-L. Lions, H. Pham, E. Taflin, Mathematical Epidemiology (2008)
Paris-Princeton Lectures on Mathematical Finance 2004. Vol. 1946: G. Allaire, A. Arnold, P. Degond, T.Y. Hou,
Editors: R.A. Carmona, E. inlar, I. Ekeland, E. Jouini, J.A. Quantum Transport. Modelling, Analysis and Asymp-
Scheinkman, N. Touzi (2007) totics. Cetraro, Italy 2006. Editors: N.B. Abdallah,
Vol. 1920: S.N. Evans, Probability and Real Trees. Ecole G. Frosali (2008)
d’Été de Probabilités de Saint-Flour XXXV-2005 (2008) Vol. 1947: D. Abramovich, M. Mariño, M. Thaddeus,
Vol. 1921: J.P. Tian, Evolution Algebras and their Appli- R. Vakil, Enumerative Invariants in Algebraic Geo-
cations (2008) metry and String Theory. Cetraro, Italy 2005. Editors:
K. Behrend, M. Manetti (2008)
Vol. 1922: A. Friedman (Ed.), Tutorials in Mathematical
Vol. 1948: F. Cao, J-L. Lisani, J-M. Morel, P. Mus, F. Sur,
BioSciences IV. Evolution and Ecology (2008)
A Theory of Shape Identification (2008)
Vol. 1923: J.P.N. Bishwal, Parameter Estimation in
Vol. 1949: H.G. Feichtinger, B. Helffer, M.P. Lamoureux,
Stochastic Differential Equations (2008)
N. Lerner, J. Toft, Pseudo-Differential Operators. Quan-
Vol. 1924: M. Wilson, Littlewood-Paley Theory and
tization and Signals. Cetraro, Italy 2006. Editors: L.
Exponential-Square Integrability (2008)
Rodino, M.W. Wong (2008)
Vol. 1925: M. du Sautoy, L. Woodward, Zeta Functions of Vol. 1950: M. Bramson, Stability of Queueing Networks,
Groups and Rings (2008) Ecole d’Eté de Probabilits de Saint-Flour XXXVI-2006
Vol. 1926: L. Barreira, V. Claudia, Stability of Nonauto- (2008)
nomous Differential Equations (2008) Vol. 1951: A. Moltó, J. Orihuela, S. Troyanski,
Vol. 1929: Y. Mishura, Stochastic Calculus for Fractional M. Valdivia, A Non Linear Transfer Technique for
Brownian Motion and Related Processes (2008) Renorming (2009)
Vol. 1930: J.M. Urbano, The Method of Intrinsic Scaling. Vol. 1952: R. Mikhailov, I.B.S. Passi, Lower Central and
A Systematic Approach to Regularity for Degenerate and Dimension Series of Groups (2009)
Singular PDEs (2008) Vol. 1953: K. Arwini, C.T.J. Dodson, Information Geo-
Vol. 1931: M. Cowling, E. Frenkel, M. Kashiwara, metry (2008)
A. Valette, D.A. Vogan, Jr., N.R. Wallach, Representation Vol. 1954: P. Biane, L. Bouten, F. Cipriani, N. Konno,
Theory and Complex Analysis. Venice, Italy 2004. N. Privault, Q. Xu, Quantum Potential Theory. Editors:
Editors: E.C. Tarabusi, A. D’Agnolo, M. Picardello U. Franz, M. Schuermann (2008)
(2008) Vol. 1955: M. Bernot, V. Caselles, J.-M. Morel, Optimal
Vol. 1932: A.A. Agrachev, A.S. Morse, E.D. Sontag, Transportation Networks (2008)
H.J. Sussmann, V.I. Utkin, Nonlinear and Optimal Control Vol. 1956: C.H. Chu, Matrix Convolution Operators on
Theory. Cetraro, Italy 2004. Editors: P. Nistri, G. Stefani Groups (2008)
(2008) Vol. 1957: A. Guionnet, On Random Matrices: Macro-
Vol. 1933: M. Petkovic, Point Estimation of Root Finding scopic Asymptotics, Ecole d’Eté de Probabilits de Saint-
Methods (2008) Flour XXXVI-2006 (2009)
Vol. 1934: C. Donati-Martin, M. Émery, A. Rouault, Vol. 1958: M.C. Olsson, Compactifying Moduli Spaces
C. Stricker (Eds.), Séminaire de Probabilités XLI (2008) for Abelian Varieties (2008)
Vol. 1935: A. Unterberger, Alternative Pseudodifferential Vol. 1959: Y. Nakkajima, A. Shiho, Weight Filtrations
Analysis (2008) on Log Crystalline Cohomologies of Families of Open
Vol. 1936: P. Magal, S. Ruan (Eds.), Structured Population Smooth Varieties (2008)
Models in Biology and Epidemiology (2008) Vol. 1960: J. Lipman, M. Hashimoto, Foundations of
Vol. 1937: G. Capriz, P. Giovine, P.M. Mariano (Eds.), Grothendieck Duality for Diagrams of Schemes (2009)
Mathematical Models of Granular Matter (2008) Vol. 1961: G. Buttazzo, A. Pratelli, S. Solimini,
Vol. 1938: D. Auroux, F. Catanese, M. Manetti, P. Seidel, E. Stepanov, Optimal Urban Networks via Mass Trans-
B. Siebert, I. Smith, G. Tian, Symplectic 4-Manifolds portation (2009)
Vol. 1962: R. Dalang, D. Khoshnevisan, C. Mueller, Vol. 1991: F. Gazzola, H.-C. Grunau, G. Sweers, Polyhar-
D. Nualart, Y. Xiao, A Minicourse on Stochastic Partial monic Boundary Value Problems (2010)
Differential Equations (2009) Vol. 1992: A. Parmeggiani, Spectral Theory of Non-
Vol. 1963: W. Siegert, Local Lyapunov Exponents (2009) Commutative Harmonic Oscillators: An Introduction
Vol. 1964: W. Roth, Operator-valued Measures and Inte- (2010)
grals for Cone-valued Functions and Integrals for Cone- Vol. 1993: P. Dodos, Banach Spaces and Descriptive Set
valued Functions (2009) Theory: Selected Topics (2010)
Vol. 1965: C. Chidume, Geometric Properties of Banach Vol. 1994: A. Baricz, Generalized Bessel Functions of the
Spaces and Nonlinear Iterations (2009) First Kind (2010)
Vol. 1966: D. Deng, Y. Han, Harmonic Analysis on Spaces Vol. 1995: A.Y. Khapalov, Controllability of Partial
of Homogeneous Type (2009) Differential Equations Governed by Multiplicative
Vol. 1967: B. Fresse, Modules over Operads and Functors Controls (2010)
(2009) Vol. 1996: T. Lorenz, Mutational Analysis. A Joint Frame-
Vol. 1968: R. Weissauer, Endoscopy for GSP(4) and the work for Cauchy Problems In and Beyond Vector Spaces
Cohomology of Siegel Modular Threefolds (2009) (2010)
Vol. 1969: B. Roynette, M. Yor, Penalising Brownian Vol. 1997: M. Banagl, Intersection Spaces, Spatial
Paths (2009) Homology Truncation, and String Theory (2010)
Vol. 1970: M. Biskup, A. Bovier, F. den Hollander, D. Vol. 1998: M. Abate, E. Bedford, M. Brunella, T.-C. Dinh,
Ioffe, F. Martinelli, K. Netočný, F. Toninelli, Methods of D. Schleicher, N. Sibony, Holomorphic Dynamical Sys-
Contemporary Mathematical Statistical Physics. Editor: tems. Cetraro, Italy 2008. Editors: G. Gentili, J. Guenot,
R. Kotecký (2009) G. Patrizio (2010)
Vol. 1971: L. Saint-Raymond, Hydrodynamic Limits of Vol. 1999: H. Schoutens, The Use of Ultraproducts in
the Boltzmann Equation (2009) Commutative Algebra (2010)
Vol. 1972: T. Mochizuki, Donaldson Type Invariants for Vol. 2000: H. Yserentant, Regularity and Approximability
Algebraic Surfaces (2009) of Electronic Wave Functions (2010)
Vol. 1973: M.A. Berger, L.H. Kauffmann, B. Khesin, H.K. Vol. 2001: O.E. Barndorff-Nielson, J. Bertoin, J. Jacod,
Moffatt, R.L. Ricca, De W. Sumners, Lectures on Topo- C. Kl”uppelberg (Eds.), Lévy Matters I (2010)
logical Fluid Mechanics. Cetraro, Italy 2001. Editor: R.L. Vol. 2002: C. Pötzsche, Geometric Theory of Discrete
Ricca (2009) Nonautonomous Dynamical Systems (2010)
Vol. 1974: F. den Hollander, Random Polymers: École Vol. 2003: A. Cousin, S. Crépey, O. Guéant, D. Hobson,
d’Été de Probabilités de Saint-Flour XXXVII – 2007 M. Jeanblanc, J.-M. Lasry, J.-P. Laurent, P.-L. Lions,
(2009) P. Tankov, Paris-Princeton Lectures on Mathematical
Vol. 1975: J.C. Rohde, Cyclic Coverings, Calabi-Yau Finance 2010. Editors: R.A. Carmona, E. Cinlar,
Manifolds and Complex Multiplication (2009) I. Ekeland, E. Jouini, J.A. Scheinkman, N. Touzi (2010)
Vol. 1976: N. Ginoux, The Dirac Spectrum (2009) Vol. 2004: K. Diethelm, The Analysis of Fractional
Vol. 1977: M.J. Gursky, E. Lanconelli, A. Malchiodi, Differential Equations (2010)
G. Tarantello, X.-J. Wang, P.C. Yang, Geometric Analysis Vol. 2005: W. Yuan, W. Sickel, D. Yang, Morrey and
and PDEs. Cetraro, Italy 2001. Editors: A. Ambrosetti, S.- Campanato Meet Besov, Lizorkin and Triebel (2011)
Y.A. Chang, A. Malchiodi (2009) Vol. 2006: A. Rouault, A. Lejay, C. Donati-Martin (Eds.),
Vol. 1978: M. Qian, J.-S. Xie, S. Zhu, Smooth Ergodic Séminaire de Probabilités XLIII (2011)
Theory for Endomorphisms (2009) Vol. 2007: E. Bujalance, F.J. Cirre, J.M. Gamboa, G.
Vol. 1979: C. Donati-Martin, M. Émery, A. Rouault, Gromadzki, Symmetries of Compact Riemann Surfaces
C. Stricker (Eds.), Śeminaire de Probablitiés XLII (2009) (2010)
Vol. 1980: P. Graczyk, A. Stos (Eds.), Potential Analysis
of Stable Processes and its Extensions (2009)
Vol. 1981: M. Chlouveraki, Blocks and Families for Recent Reprints and New Editions
Cyclotomic Hecke Algebras (2009)
Vol. 1982: N. Privault, Stochastic Analysis in Discrete and Vol. 1702: J. Ma, J. Yong, Forward-Backward Stochas-
Continuous Settings. With Normal Martingales (2009) tic Differential Equations and their Applications. 1999 –
Vol. 1983: H. Ammari (Ed.), Mathematical Modeling in Corr. 3rd printing (2007)
Biomedical Imaging I. Electrical and Ultrasound Tomo- Vol. 830: J.A. Green, Polynomial Representations of
graphies, Anomaly Detection, and Brain Imaging (2009) GLn , with an Appendix on Schensted Correspondence
Vol. 1984: V. Caselles, P. Monasse, Geometric Description and Littelmann Paths by K. Erdmann, J.A. Green and
of Images as Topographic Maps (2010) M. Schoker 1980 – 2nd corr. and augmented edition
Vol. 1985: T. Linß, Layer-Adapted Meshes for Reaction- (2007)
Convection-Diffusion Problems (2010) Vol. 1693: S. Simons, From Hahn-Banach to Monotonic-
Vol. 1986: J.-P. Antoine, C. Trapani, Partial Inner Product ity (Minimax and Monotonicity 1998) – 2nd exp. edition
Spaces. Theory and Applications (2009) (2008)
Vol. 1987: J.-P. Brasselet, J. Seade, T. Suwa, Vector Fields Vol. 470: R.E. Bowen, Equilibrium States and the Ergodic
on Singular Varieties (2010) Theory of Anosov Diffeomorphisms. With a preface by
Vol. 1988: M. Broué, Introduction to Complex Reflection D. Ruelle. Edited by J.-R. Chazottes. 1975 – 2nd rev.
Groups and Their Braid Groups (2010) edition (2008)
Vol. 1989: I.M. Bomze, V. Demyanov, Nonlinear Vol. 523: S.A. Albeverio, R.J. Høegh-Krohn, S. Maz-
Optimization. Cetraro, Italy 2007. Editors: G. di Pillo, zucchi, Mathematical Theory of Feynman Path Integral.
F. Schoen (2010) 1976 – 2nd corr. and enlarged edition (2008)
Vol. 1990: S. Bouc, Biset Functors for Finite Groups Vol. 1764: A. Cannas da Silva, Lectures on Symplectic
(2010) Geometry 2001 – Corr. 2nd printing (2008)
LECTURE NOTES IN MATHEMATICS 123
Edited by J.-M. Morel, F. Takens, B. Teissier, P.K. Maini
1. Lecture Notes aim to report new developments in all areas of mathematics and their
applications - quickly, informally and at a high level. Mathematical texts analysing new
developments in modelling and numerical simulation are welcome.
Monograph manuscripts should be reasonably self-contained and rounded off. Thus
they may, and often will, present not only results of the author but also related work
by other people. They may be based on specialised lecture courses. Furthermore, the
manuscripts should provide sufficient motivation, examples and applications. This clearly
distinguishes Lecture Notes from journal articles or technical reports which normally are
very concise. Articles intended for a journal but too long to be accepted by most journals,
usually do not have this “lecture notes” character. For similar reasons it is unusual for
doctoral theses to be accepted for the Lecture Notes series, though habilitation theses may
be appropriate.
2. Manuscripts should be submitted either to Springer’s mathematics editorial in Heidelberg,
or to one of the series editors. In general, manuscripts will be sent out to 2 external referees
for evaluation. If a decision cannot yet be reached on the basis of the first 2 reports, further
referees may be contacted: The author will be informed of this. A final decision to publish
can be made only on the basis of the complete manuscript, however a refereeing process
leading to a preliminary decision can be based on a pre-final or incomplete manuscript.
The strict minimum amount of material that will be considered should include a detailed
outline describing the planned contents of each chapter, a bibliography and several sample
chapters.
Authors should be aware that incomplete or insufficiently close to final manuscripts
almost always result in longer refereeing times and nevertheless unclear referees’ recom-
mendations, making further refereeing of a final draft necessary.
Authors should also be aware that parallel submission of their manuscript to another
publisher while under consideration for LNM will in general lead to immediate rejection.
3. Manuscripts should in general be submitted in English. Final manuscripts should contain
at least 100 pages of mathematical text and should always include
– a table of contents;
– an informative introduction, with adequate motivation and perhaps some historical re-
marks: it should be accessible to a reader not intimately familiar with the topic treated;
– a subject index: as a rule this is genuinely helpful for the reader.
For evaluation purposes, manuscripts may be submitted in print or electronic form, in
the latter case preferably as pdf- or zipped ps-files. Lecture Notes volumes are, as a rule,
printed digitally from the authors’ files. To ensure best results, authors are asked to use
the LaTeX2e style files available from Springer’s web-server at:
ftp://ftp.springer.de/pub/tex/latex/svmonot1/ (for monographs).
ftp://ftp.springer.de/pub/tex/latex/svmultt1/ (for summer schools/tutorials).
Additional technical instructions, if necessary, are available on request from:
lnm@springer.com.
4. Careful preparation of the manuscripts will help keep production time short besides en-
suring satisfactory appearance of the finished book in print and online. After acceptance
of the manuscript authors will be asked to prepare the final LaTeX source files (and also
the corresponding dvi-, pdf- or zipped ps-file) together with the final printout made from
these files. The LaTeX source files are essential for producing the full-text online version
of the book (see www.springerlink.com/content/110312 for the existing online volumes
of LNM).
The actual production of a Lecture Notes volume takes approximately 12 weeks.
5. Authors receive a total of 50 free copies of their volume, but no royalties. They are entitled
to a discount of 33.3% on the price of Springer books purchased for their personal use, if
ordering directly from Springer.
6. Commitment to publish is made by letter of intent rather than by signing a formal contract.
Springer-Verlag secures the copyright for each volume. Authors are free to reuse material
contained in their LNM volumes in later publications: a brief written (or e-mail) request
for formal permission is sufficient.
Addresses:
Professor J.-M. Morel, CMLA,
École Normale Supérieure de Cachan,
61 Avenue du Président Wilson, 94235 Cachan Cedex, France
E-mail: Jean-Michel.Morel@cmla.ens-cachan.fr
Professor F. Takens, Mathematisch Instituut,
Rijksuniversiteit Groningen, Postbus 800,
9700 AV Groningen, The Netherlands
E-mail: F.Takens@math.rug.nl
Professor B. Teissier, Institut Mathématique de Jussieu,
UMR 7586 du CNRS, Équipe “Géométrie et Dynamique”,
175 rue du Chevaleret
75013 Paris, France
E-mail: teissier@math.jussieu.fr