Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Riemann Orbifolds
Jack Kelly
Abstract
The goal of this note is to establish the Galois correspondence for Rie-
mann orbifolds.
Introduction
We will always assume that Riemann surfaces under consideration are con-
nected. A holomorphic map of Riemann surfaces f : X Y is called a covering
if every point y Y has a connected neighbourhood U such that f restricted to
each connected component of f
1
(U) is a biholomorphism onto U. Such maps
are necessarily local homeomorphisms. One can describe all coverings of Y up
to isomorphism as follows. First, there is a unique (up to isomorphism) covering
p :
Y Y by a simply-connected Riemann surface
Y. It is called the univer-
sal covering. This is already interesting because according to the uniformization
theorem a simply connected Riemann surface is biholomorphic to one of the
three domains: the open unit disk, the complex plane, or the Riemann sphere.
Therefore Riemann surfaces Y split into three classes according to the type of
the uniformization
Y. In order to describe all possible coverings one studies the
group of so called deck transformations of the universal covering. Deck trans-
formations are biholomorphisms g :
Y
Y such that g p = g. In other words,
they act brewise. Deck transformations form a group under composition. In
the view of above mentioned universality this group is often called
1
(Y), the
fundamental group of the Riemann surface Y. But in this note we prefer the
notation Aut(p) since it is useful to keep in mind the covering map p. Next,
for a subgroup Aut(p) one can dene the quotient Riemann surface \
Y
and there is a natural covering \
=
C\ |0|.
Example 1.1. Let n be a positive integer. The map
C
z z
n
is a covering. Every point has n preimages, so one says it is a covering of degree n.
It is a general topological observation that all bers f
1
( y) of a covering will
have the same number of points, either nite or innite. The number of points
in bers is called the degree of the covering.
Example 1.2. The map
C C
z e
2iz
is a covering of innite degree. Since C is simply-connected, this map gives us the
universal covering of C
.
4 [Spring,
Example 1.3. The map
C C
z z
n
is proper, and is therefore a branched covering. It ramies at z = 0 and the rami-
cation index is e(0) = n.
Observe that every covering is a branched covering, just there is no ramica-
tion. And from every branched covering f : X Y one can get a covering by
throwing away branch points and their preimages, i.e. f : X
0
Y
0
in the nota-
tion of Denition 0.2. The degree f is dened to be the degree of the covering
f : X
0
Y
0
. In other words, the number of points in the ber f
1
( y) is still the
same for all y Y
0
and this number is the degree. The degree of the branched
covering in Example 1.3 is n, all points z ,= 0 have exactly n preimages while 0
has only one preimage.
Denition 1.4. Let f : X Y be a holomorphic map. By Aut(X) we denote the
group of biholomorphisms : X X with the operation of composition. The group
of deck transformations Aut( f ) of f is the subgroup of those Aut(X) which act
brewise, i.e.
Aut( f ) = | : X X [ f = f |.
Denition 1.5. A group is said to act properly discontinuously on a Riemann
surface X if for each x, y X there are open neighobrhoods U, V of x, y respectively
such that the set | [.U V ,= | is nite.
We shall be interested in Aut( f ) where f is a branched covering map. In this
case we have the following
Proposition 1.6. Let f : X Y be a branched covering map of Riemann surfaces.
Then the deck group Aut( f ) acts properly discontinuously on X.
In order to prove this statement, we need the following observation.
Proposition 1.7. Let f : X Y, h, g : Z X be holomorphic maps satisfying
f g = f h . Suppose x X is an unramied point of f , and that for some z Z,
h(z) = g(z) = x. Then h = g.
Proof. Let U be a neighborhood of x such that f maps U biholomorphically
onto f (U). Put V = h
1
(U) g
1
(U). This is non-empty since, z V. Now
f h[
V
= f g[
V
and since f [
U
is a bijective, it follows that h[
V
= g[
V
. Since h and
g are holomorphic and equal on an open set, they must be equal everywhere.
2012] 5
Proof of Proposition 1.6. This proof only makes use of part (3) of the denition of
branched covering map. Let x, y X. If p(x) ,= p( y) then, since Y is Hausdorff
it is trivial to nd neighboruhoods U, V around x, y respectively such that for
any , .U V = . Suppose p(x) = p( y) and let W be a neighbourhood
of p(x) such that each connected component of p
1
(W) maps onto W by a
proper map. Let U be the connected component of p
1
(W) containing x, and
V the connected component containing y. Now suppose for some Aut( f ),
.U V ,= . Since .U is mapped (on)to U by p and it is connected, it is
contained in a connected component. It follows that .U V. Since the ramied
points of p are discrete we may pick some u U which is unramied. Consider
the set S = |.u[ Aut( f ) and .U V|. Now, S is contained in the bre of
p(u) under the restriction p[
V
. This set is compact, since it is the preimage of
a compact set under a proper map. It must also be discrete, since p is a non-
constant holomorphic map. Since discrete, compact sets are nite this bre, and
therefore S, must be nite. So S = |
1
.u, . . . ,
n
.u|. Therefore, if .UV ,= then
.U V, so .u =
i
.u for some i. Since u is unramied, =
i
by Proposition
1.7 . Therefore | [.UV ,= | = |
1
, . . . ,
n
| which is nite as required.
2. Properly Discontinuous Group Actions
We shall now discuss the Riemann surface structure on
X where X is a Rie-
mann surface and a group of automorphisms acting properly discontinuously
on it. Here we make use of Mirandas book [Miranda(1995)], which deals with
the case of a nite group action. However, as is mentioned in [Miranda(1995)],
this can be modied easily to work for properly discontinuous group actions;
and almost every proof in this section is essentially identical to the corresponding
proof in [Miranda(1995)], with nite replaced by properly discontinuous. Usu-
ally, when is a group acting on a set X we shall denote by the quotient map
X
X, [x]
/
the
/
orbit of X and Stab
/
(x) the
/
stabilizer of x X. On
occasion we will be dealing with more than two subgroups, and in this instance
we shall use the notations
, [x]
, Stab
(x).
The following are technical propositions, which will facilitate the construc-
tion of neighbourhoods in
X is Hausdorff.
6 [Spring,
Lemma 2.2. Let be a group of automorphisms acting properly discontinuously on
a Riemann surface X. The points of X with non-trivial stabilizers do not accumulate.
In particular they form a discrete set.
Proof. [Miranda(1995)] Suppose there is a sequence (x
n
)
n=1
converging to x
such that each x
i
has a nontrivial element
i
xing it. Let U be a neighborhood
of x such that S = | : .UU ,= | is nite. Since for sufciently large m, x
m
will
be contained in U, we can assume without loss of generality that the sequence is
in U. For each m,
m
.x
m
= x
m
, so
m
S. Since S is nite it is possible to choose
a subsequence of (x
n
) such that each element of the subsequence is xed by the
same non-trivial . But this subsequence also converges to x, so by continuity
must then x x as well. Since is holomorphic either its set of xed points is
discrete, or it is the identity, a contradiction.
Lemma 2.3. Let be a group of automorphisms acting properly discontinuously
on a Riemann surface X. For each x X there is an open neighborhood U of x such
that
1. U is invariant under Stab(x)
2. U (.U) = for every / Stab(x)
3. no point of U except x is xed by any non-trivial element of
4. the natural map q : Stab(x)
X.
Proof. [Miranda(1995)] Choose a neighborhood V of x such that S = | :
.V V ,= | is nite. Clearly Stab(x) S since for Stab(x), x .V V.
Let S \ Stab(x) = |
1
, . . . ,
n
| be the elements of S which do not x x. Since X
is Hausdorff, we can nd for each i open neighbrohoods V
i
V of x and W
i
of
i
.x with V
i
W
i
= . Clearly,
1
i
W
i
is an open neighborhood of x for each i.
Put R
i
= V
i
(
1
i
.W
i
), R =
n
i=1
R
i
and
U =
Stab(x)
.R
R and therefore U are nite intersections of open sets containing x and are
therefore open neighbourhoods of x. Also, acting on U by Stab(x) will just
2012] 7
permute the sets in the intersection dening U. It follows that for Stab(x),
.U = U proving (1).
For part (2) observe that if , S then since U V, by assumption .U U =
. For
i
S \ Stab(x), note that R
i
(
i
.R
i
) V
i
W
i
= so R (
i
.R) = and
U
i
(U) = for each i.
For part (3), since the set of points with non-trivial stabilizer has no accu-
mulation points, by Lemma 2.2 we may choose V at the beginning of this proof
such that no point of V \ |x| has non-trivial stabilizer. Then clearly, no point of
U \ |x| has non-trivial stabilizer.
For part (4), let denote the quotient map X
X and
Stab(x)
the quotient
map U Stab(x)
(x) of at a point x
is equal to [Stab(x)[.
Proof. [Miranda(1995)] Let [x]
X and = z [
1
U
is a chart centred at [x].
Now suppose n 2. Dene
h : U , y
Stab(x)
(z ( y))
Note that h has ramication index n at x, and that h is Stab(x) invariant. There-
fore it lifts to a well-dened continuous map h on Stab(x)
U satisfying h q = h
where q : Stab(x)
2
: U
2
in the n 2 case. For i = 1, 2 let U
i
be the neighbourhood in
X from which U
i
is constructed. Then U
i
is and ideal neighbourhood of some
point x
i
X, and there are corresponding local coordinates z
i
centred at x
i
with
domain U
i
. The chart on
1
is just z
1
[
1
U
1
. The chart
2
is h q
1
where h is
the lift of
Stab(x
2
)
(z
2
( y)) to Stab(x
2
)
U
2
. and q : Stab(x
2
)
U
2
X is the
2012] 9
natural map, which maps homeomorphically onto U
2
. We then have
2
1
1
(x)
= h q
1
[
U
1
z
1
1
(x)
= h q
1
(z
1
1
(x)])
= h([z
1
1
(x)]
Stab(x
2
)
)
= h([z
1
1
(x))
=
Stab(x
2
)
(z
2
(z
1
1
(x)))
Therefore
2
1
1
=
Stab(x
2
)
z
2
z
1
1
Each is a holomorphic function
on X, so z
2
z
1
1
and thereofre
Stab(x
2
)
z
2
z
1
1
is holomorphic. Its
inverse is holomorphic by the inverse function theorem. The charts are therefore
compatible.
That the quotient map is holomorphic is clear. Moreover, its ramication in-
dex at a point x is the same as that of h which is [Stab(x)[. If there is some
other Riemann surface structure on
/
of , the natural map
/
X, [x]
/
[x]
is a branched covering map.
Proof. First note that by Theorem 2.5 both
X and
/
1
.u
1
=
2
.u
2
. But then
1
2
/
1
.u
1
U
1
2
/
2
.U. Since U is ideal,
1
2
/
2
Stab(x). Hence
1
2
/
1
.U = U, and
/
(
2
.U) =
/
(
/
1
.U) =
/
(
1
.U). We
have therefore shown that two sets
/
(
1
.U) and
/
(
2
.U) are either disjoint
or equal. It follows that indeed, each component of q
1
()(U) is of the form
/
(.U).
Suppose [x] is not a branch point of q. Then, no point of the bre q
1
([x]) =
|[.x]
/
: | is a ramied point of q. In particular for each , Stab
/
(.x) =
Stab(.x) = Stab(x)
1
= Stab
/
(x)
1
. Suppose u
1
, u
2
U and q([.u
1
]
/
) =
q([.u
2
]
/
), so [u
1
] = [u
2
] i.e. u
1
= u
2
for some . But since U is an
ideal neighbourhood, Stab(x) = Stab
/
(x). Now
1
Stab
/
(x)
1
=
Stab
/
(.x), so
1
= for some Stab
/
(.x). Therefore .u
1
= .u
2
=
u
2
and [.u
1
]
/
= [.u
2
]
/
= [.u
2
]
/
. The restriction of q to each component
is therefore injective, so it is indeed a homeomorphism, as required.
Finally we must check the 3rd condition for q to be a branched covering. Let
K be an ideal compact neighbourhood of x. By shrinking, we may assume K
is contained in an ideal open nieghbourhood. Then by a similar argument as
before, the connected components of q
1
((K)) will again be given by (.K).
Such sets are clearly compact since they are the images of compact sets. Since
any continuous map from a compact set to a Hausdorff set is proper, the restric-
2012] 11
tion of q to
/
(.K) is proper.
In particular, the quotient map in Proposition 3.1 is a branched covering map.
Let us now compute the deck group.
Proof of Proposition 3.1. Clearly Aut(). Now, suppose Aut(). Let x
X be such that [Stab(x)[ = 1, so x is an unramied point of . Now ((x)) =
(x) so (x) = (x) for some . But since x is unramied, the stabilizer of
x is trivial by Proposition 1.7, so = .
4. Orbifolds and Orbifold Coverings
In this section we shall prove some technical details about Riemann orbifolds,
which allow us to extend some of our results about branched covering maps
to results about orbifold coverings. If (X, ), (Y, ) are Riemann orbifolds, and
f : X Y is holomorphic and non-constant, we say that g is compatible with the
orbifold structures (X, ), (Y, ) if e
f
(z)(z) = ( f (z)), or that it is a map of Rie-
mann orbifolds. A homomorphism of orbifold coverings is a homomorphism of
the underlying branched coverings which is compatible with the orbifold struc-
tures, and a deck transformation of an orbifold covering f : (X, ) (Y, ) is an
orbifold automorphism of (X, ) such that f = f . However the following
proposition implies that if f : (X, ) (Y, ), f
/
: (X
/
,
/
) (Y, ) are orbifold
coverings, and : X X
/
is a homomorphism of the underlying branched cover-
ings, then is a homomorphism of the orbifold coverings. In particular the deck
group of an orbifold covering is the same as the deck group of the underlying
branched covering map.
Proposition 4.1. Let (Y, ), (S
/
,
/
), (S
//
,
//
) be Riemann orbifolds. Suppose g :
S
/
S
//
, f : S S
/
, and g f : S S
//
are holomorphic.
1. If f : (Y, ) (S
/
,
/
), g f : (Y, ) (S
//
,
//
) are comptaible, then g is
compatible.
2. If g : (S
/
, ) (S
//
,
/
), g f : (Y, ) (S
//
,
//
) are compatible, then f
compatible.
Proof. We have
1.
//
(g( f (x))) = e
gf
(x)(x) = e
g
( f (x))e
f
(x)(x) = e
g
( f (x))
/
( f (x)x)
2.
//
(g( f (x))) = e
gf
(x)(x) = e
g
( f (x))e
f
(x)(x) =
/
( f (x))
(x)
e
f
(x)(x) =
e
f
(x)
/
( f (x))
12 [Spring,
The deck group of a branched covering map acts properly discontinuously,
so the deck group of an orbifold cover will act properly discontinuously on the
underlying Riemann surface. The question that naturally arises now, is, for a
group of orbifold automorphisms of a Riemann orbifold (X, ) acting properly
discontinuously on X, does there exist an orbifold structure on
X such that
the quotient map is an orbifold covering? The answer is yes. We already know
that there is a Riemann surface structure on
(x). So (
X, ) where
:
X N, [x] (x)[Stab(x)[
Theorem 4.3. Let (X, ) be a Riemann orbifold and let be a group acting prop-
erly discontinuously on X. The quotient orbifold (
X such
that is holomorphic. is clearly the unique map such that : (X, )
(
X, ) is an orbifold covering.
Henceforth, when (X, ) is a Riemann orbifold, and a group of automor-
phisms acting properly discontinuously on it,
Y, ) (Y, ).
A covering of orbifolds f : (
Y, ) (Y, )
is said to be universal if
Y is simply connected, 1 and f is regular. Ac-
cording to the uniformization theorem for Riemann orbifolds (see Theorem A.2
in [Milnor(2006)]), the only cases when an orbifold (Y, ) does not have a uni-
versal covering are when Y is a Riemann sphere with one ramication point or
two ramication points with different ramication indices. Henceforth we shall
assume all orbifolds admit a universal covering. Note that the composition of
an orbifold isomorphism with a universal covering is a universal covering. In
particular, if an orbifold admits a universal covering then every orbifold in its
isomorphism class does as well. The terminology universal covering is due to
the following property (Lemma E.2 in [Milnor(2006)]):
Proposition 5.1. f : (X, ) (Y, ) is a covering map between orbifolds if and
only if it lifts to a biholomorphism
f :
X
Y.
If p :
Y (Y, ) is a universal covering, then q
/
= p
f
1
:
Y (X, )
is a universal covering of (X, ). This observation is a part of the following
statement.
Lemma 5.2. Assume the universal covering p :
Y (Y, ) is xed.
1. For every covering f : (X, ) (Y, ) it is possible to choose a universal
covering q :
Y (X, ) so that f q = p. It follows that Aut(q) Aut(p) is
a subgroup. Moreover, if q
/
:
Y (X, ) is another universal covering such
that f q
/
= p, then Aut(q) and Aut(q
/
) are conjugate inside of Aut(p).
2. Suppose f : (X, ) (Y, ) and f
/
: (X
/
,
/
) (Y, ) are two coverings, and
h : (
Y, ) (X, ), h
/
: (
Y, ) (X
/
,
/
) are universal coverings satisfying
f h = f
/
h
/
= p. Then f and f
/
are isomorphic coverings if and only if the
corresponding subgroups Aut(h), Aut(h
/
) are conjugate inside of Aut(p), i.e.
there is Aut(p) such that Aut(h) = Aut(h
/
)
1
.
For a particularly important example of a universal covering, consider a sim-
ply connected Riemann surface
Y and a group of automorphisms acting prop-
erly discontinuously on it. Then the quotient map :
Y (
Y, ) where 1,
is a universal orbifold covering.
Lemma 5.3. Let (Y, ) be a Riemann orbifold with a universal orbifold covering
p : (
Y Y
such that p
= p. Moreover p
Aut(p)
: Aut(p)
Y Y is an isomorphism of
orbifolds.
14 [Spring,
Proof. According to Proposition 1.6, Aut(p) and hence also act properly dis-
continuously on
Y. Therefore we can consider the quotient orbifold (\
Y, )
with ([y]
) = [St ab
Y Y, [x]
p(x)
If [x]
= [y]
(x) = p
([x]
) =
p(x). Therefore it is surjective. Moreover by Proposition 2.4 , p
is holomorphic
and by Lemma 4.1 it is compatible with the orbifold structures.
We will now verify that p
Aut(p)
is an isomorphism such that p
Aut(p)
q = p
,
(where q :
Y Aut(p)
Y, [y]
[y]
Aut(p)
) and then use this to show that p
Y
Aut(p)
we have p
Aut(p)
q([x]
) = p
Aut(p)
([x]
Aut(p)
) =
p(x) = p
([x]
). Therefore p
Aut(p)
q = p
is as well.
Theorem 5.4 (The Galois Correspondence). Let (Y, ) be a Riemann orbifold
with universal cover p :
Y Y. There is a natural 1 1 correspondence be-
tween isomorphism classes of orbifold coverings of (Y, ) and conjugacy classes of
subgroups in Aut(p).
Proof. After Lemma 5.2 we clearly have a well-dened map sending the iso-
morphism class of f : (X, ) (Y, ) into the conjugacy class of the subgroup
= Aut(h) of Aut(p) where h : (
/ are
universal coverings of
Y and
/
= p
/ = p.
By Proposition 3.1 the deck groups of these two universal coverings are ,
/
respectively. Then by virtue of Lemma 5.2 (2) there is an isomorphism of the
coverings p
and p
/
(explicitly, this isomorphism is given by [x]
[.x]
/ ).
Therefore, our map is well-dened.
2012] 15
We now have two well-dened maps. Let us check that they are mutually
inverse. Let f : (X, ) (Y, ) be a covering, let h : (
Y, ) (X, ) be a
universal covering such that f h = p, and put = Aut(h). By Lemma 5.3 the
map h
we have f h
([x]
) = f (h(x)) = p(x) = p
([x]
is a universal covering of
Y such
that p
= p. Moreover, Aut(
z
for some z . Moreover,
two elliptic curves E
= C
z
, E
/
= C
z
/ are isomorphic as algebraic varieties if
and only if z
/
=
az+b
cz+d
for some a, b, c, d Z such that ad bc = 1.
The group
= |
a b
c d
: a, b, c, d Z : ad bc = 1|
|I , I |
is called the modular group and the map z
az+b
cz+d
denes an action of this group
on . Now let j : C be the map which assigns to z the j-invariant of the
elliptic curve isomorphic to C
z
. By our previous remarks this map is invariant
under the action of . It also turns out that it is holomorphic on and satises
lim
z
j(z) = . These three properties mean that j is what we call a modular
function for the group .
Denition 6.3. Let
/
be a subgroup of nite index. A meromorphic function
t on is said to be a modular function for the group
/
if for all
/
/
t
/
= t
and for all , lim
z
t(.z) exists in C.
A family of elliptic curves is simply an elliptic curve over C(t), the eld of
rational functions over C in one variable. This curve then has a j-invariant
which is a rational function, which we denote by J(t). We say that a family with
j-invariant J admits a modular parametrization, if there is a subgroup
/
of
of nite index and a modular function t for
/
such that J(t(z)) = j(z). As an
example, consider the family
X
3
+Y
3
+ Z
3
+ t XY Z = 0, O = (0 : 1 : 1)
The rational j-invariant of this family is J(t) =
t
3
(t
3
216)
3
(27+t
3
)
3
. A modular
parametrization of this family is given by
(
z
3
)
3
(3z)
3
3 where is the Dedekind
eta function given by
(z) = e
iz
12
n1
(1 e
2iz
)
with corresponding modular group
(3) = |i
a b
c d
: a d 1, b c 0 mod 3|
2012] 17
We want to know when it is possible to nd a modular parametrization for
a family of elliptic curves. This problem can be solved by applying the Galois
Correspondence, using the fact that acts properly discontinuously on and
that the function j denes a universal orbifold covering. We state without proof
the solution.
Theorem 6.4. Let E
t
be a family of elliptic curves with rational j-invariant J(t).
This family admits a modular parametrization if and only if J is surjective, e
J
(x)[3
for all x J
1
(0), e
J
(x)[2 for all x J
1
(1728) and e
J
(x) = 1 for all x
J
1
(C\ |0, 1728|)
Write R(t) =
P(t)
Q(t)
where P, Q are coprime polynomials. If R(t) = w then
the ramication index of R at t is simply the mutliplicty of t as a root of the
polynomial P(t) wQ(t). Therefore, the theorem can be stated as:
Theorem 6.5. Let E
t
be a family of elliptic curves with rational j-invariant J(t) =
P(t)
Q(t)
where P, Q are coprime polynomials. This family admits a modular parametriza-
tion if and only if J is surjective and P(t) only has roots of multiplicty 1 or
3, P(t) 1728Q(t) only has roots of multiplicty 1 or 2 and for all other w,
P(t) wQ(t) only has roots of multiplicity 1
References
[McMullen()] Curtis T. McMullen. Complex Dynamics and Renormaliza-
tion. URL http://www.math.harvard.edu/~ctm/papers/home/
text/papers/real/book.pdf.
[Milne(a)] J.S. Milne. Elliptic Curves. a. URL http://www.jmilne.org/
math/Books/ectext0.pdf.
[Milne(b)] J.S. Milne. Modular Functions and Modular Forms. b. URL http:
//www.jmilne.org/math/CourseNotes/MF110.pdf.
[Milnor(2006)] J. Milnor. Dynamics in One Complex Variable. (AM-160): Third
Edition. (AM-160). Princeton University Press, 2006. ISBN 9780691124889.
URL http://books.google.ie/books?id=DsthOelUMlkC.
[Miranda(1995)] R. Miranda. Algebraic Curves and Riemann Surfaces. American
Mathematical Society, 1995. ISBN 9780821802687. URL http://books.
google.ie/books?id=qjg6GOQaHNEC.
[Thurston(1997)] William P. Thurston. The Geometry and Topology of
Three-Manifolds. 1997. URL http://pi.unl.edu/~mbrittenham2/
classwk/990s08/public/thurston.notes.pdf/6a.pdf.