Seifert Manifolds as

Circle Manifolds
Joakim Gåsøy
Master’s Thesis, Spring 2020
This master’s thesis is submitted under the master’s programme Mathematics,
with programme option Mathematics, at the Department of Mathematics,
University of Oslo. The scope of the thesis is 60 credits.

The front page depicts a section of the root system of the exceptional
Lie group E8 , projected into the plane. Lie groups were invented by the
Norwegian mathematician Sophus Lie (1842–1899) to express symmetries in
differential equations and today they play a central role in various parts of
mathematics.
 Abstract

Seifert manifolds play a crucial role in the classification of 3-manifolds,

as they occupy six of the eight geometries proposed by Thurston. In
this thesis we determine which Seifert manifolds can be given a smooth
action from the circle group. We show how the stabilizer groups and
orbit space of this circle manifold relate to the Seifert invariants and
base space of the Seifert manifold. In particular, we give a smooth circle
action to Brieskorn manifolds and spherical space forms and calculate
their stabilizer groups. Additionally, we show that the orbit space is an
orbifold.

i
 Acknowledgements

First and foremost, I wish to thank my supervisor Kim Anders Frøyshov for
presenting me with this topic and helping me find valuable sources. Thank you
for your patience and assistance.
I also wish to dedicate a thank you to August and Haakon, who assisted me
in proofreading, and for their general support.
Finally I wish to thank my partner, friends and family for their invaluable
support throughout the years.

iii
 Contents

Abstract i

Acknowledgements iii

Contents iv

1 Introduction 1
1.1 Terminology and notation . . . . . . . . . . . . . . . . . . . . 2
1.2 Thurston’s geometrization conjecture . . . . . . . . . . . . . . 2

2 Seifert manifolds as G-manifolds 5

2.1 Seifert manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 G-manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Circle manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Examples of Seifert manifolds 17

3.1 Brieskorn manifolds . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Spherical space forms . . . . . . . . . . . . . . . . . . . . . . . 21

4 Orbifolds 29

Bibliography 31

iv
 CHAPTER 1

Introduction

Seifert manifolds were introduced and classified by Herbert Seifert in the 1930s.
Interest in these manifolds has had a resurgence due to the work of William
Thurston and his celebrated geometrization conjecture. This was proposed
by Thurston in 1992 and proved later by Grigorij Perelman in 2002. The
geometrization conjecture essentially states that any closed, oriented 3-manifold
can be canonically decomposed into something called geometric 3-manifolds. In
other words the classification of closed, oriented 3-manifolds can be reduced
to the classification of geometric 3-manifolds. Thurston further shows that
there are eight classes of geometric 3-manifolds (or eight geometries), seven of
which are completely classified. The only remaining case is the spaces with a
hyperbolic geometry.
The Seifert manifolds are special in that they occupy precisely six of these
Thurston geometries, and one can identify their geometry by calculating their
Seifert invariants. We give a short introduction to the geometrization conjecture
and its relevance in this first introductory chapter.
Seifert manifolds are a generalization of circle bundles where we allow certain
"singularities". One major part of this thesis to show that due to this circle
"bundle" we can consider most Seifert manifolds as special G-manifolds where
G is the circle group S 1 , with Lie group structure inherited from U (1). A
G-manifold is a manifold with a group action of the Lie group G. We introduce
Seifert manifolds and G-manifolds in general in the second chapter, and show
that most Seifert manifolds are S 1 -manifolds with a certain finiteness condition.
As the Seifert manifolds cover as many as six of the eight geometries, they
naturally include a lot of different examples. In the third chapter we introduce
some classes of manifolds and show that they are Seifert manifolds by identifying
them as certain S 1 -manifolds as explained in chapter 2.
In the fourth and final chapter we introduce the concept of an orbifold, a
generalization of manifolds. We prove that the orbit space of G-manifolds with
a certain finiteness condition are indeed orbifolds. In particular, all Seifert
manifolds that are S 1 -manifolds have this property. Since all the information
of the Seifert manifold is encoded into this orbifold and its projection map,
we conjecture that you can calculate the Seifert invariants by studying these
orbifolds further.
The main sources used for this thesis is [Sco83] and [Mar16], which summarize
the geometrization conjecture and the classification of seven of the geometries.
Further we have used a standard textbook about smooth manifolds ([Tu08]),
lecture notes by professors at the University of Oslo, and some articles for

1
1. Introduction

specific results.

1.1 Terminology and notation

For the most part we will use terminology as in [Tu08] for talking about
differential topology. In particular, we assume that topological manifolds are
second countable and we will use the term manifold to mean smooth manifold
without boundary. Occasionally we will make remarks regarding manifolds with
a boundary.
The symbol Hn is used to represent the hyperbolic n-space, while H is used
in chapter 3 to denote the quaternions. Other than this we try to avoid any
ambiguity, and the symbols Z, Q, R and C denote as usual the integers, rational
numbers, real numbers and complex numbers, respectively.
A couple of definitions that can be found in [Lee12, p. 327, 337] and that
do not belong anywhere will be presented here:
Definition 1.1.1. Let M be a smooth manifold. A Riemannian metric on M
is a smooth symmetric covariant 2-tensor field on M that is positive definite
at each point (i.e. a choice of inner product at each tangent space that vary
smoothly across M .) A Riemannian manifold (M, r) is a smooth manifold M
with a Riemannian metric r.
We often refer to a Riemannian manifold (M, r) simply as M for simplicity.
The Riemannian metric is the extra structure that makes smooth manifolds
into geometric objects. It is possible to define familiar notions such as distance
and angle from the Riemannian metric. Importantly, every smooth manifold
has a Riemannian metric.
One construction that will be important to us is the normal bundle.
Definition 1.1.2. Let (M, r) be an n-dimensional Riemannian manifold and
S ⊆ M is a k-dimensional submanifold. The normal space to S at x is the
subspace Nx S ⊆ Tx M consisting of all vectors that are normal to S at x with
respect to h·, ·ir . The normal bundle of S is the subset N S ⊆ T M consisting
of the union of all the normal spaces Nx S for x ∈ S.
The normal bundle is the total space of the vector bundle N S → S, defined
as the restriction to N S of the tangent bundle T M → M . We also denote the
vector bundle N S → S itself the normal bundle of S.

1.2 Thurston’s geometrization conjecture

Our story starts with the prime decomposition of compact oriented 3-manifolds.
This statement is found and proved in [Mar16, p. 277].
Theorem 1.2.1. Every compact oriented 3-manifold M decomposes into prime
manifolds:
M = M1 # . . . #Mk
This list of prime factors is unique up to permutations and adding/removing
copies of S 3 .
The term prime here means not diffeomorphic to S 3 and with the property
that every separating 2-sphere in the manifold bounds a 3-ball.

2
 1.2. Thurston’s geometrization conjecture

This canonical decomposition of 3-manifolds along spheres is then followed

by another canonical decomposition along tori and Klein bottles. This next
result is a precise statement of the geometrization conjecture due to [Mor14, p.
1].
Theorem 1.2.2. Any closed, orientable, prime 3-manifold M contains a dis-
joint union of embedded incompressible 2-tori and Klein bottles such that each
connected component of the complement admits a complete, locally homogenous
Riemannian metric of finite volume.
The term incompressible means that the fundamental group of the surface
injects into the fundamental group of the 3-manifold. We shall call a manifold
geometrizable if there exists a complete, locally homogenous Riemannian metric
of finite volume.
There is also a relative version of this statement ([Mor14, p. 6]), replacing
closed manifolds by compact manifolds, showing that this is also true for
manifolds with boundary.
What these two results imply is that every compact oriented 3-manifold can
be decomposed into geometrizable manifolds in a canonical way. This allows
us to classify all oriented, compact 3-manifolds by classifying all geometrizable
manifolds.
We say that a Riemannian 3-manifold N has a geometric structure modelled
on M if every point p ∈ N has a neighbourhood isometric to some open set on
M . If M has a complete, homogenous Riemannian metric, then so will N . It
turns out that every geometrizable manifold is modelled on one of the following
complete, homogenous Riemannian manifolds ([Mar16, p. 363]):

S 3 , R3 , H3 , S 2 × R, H2 × R, Nil, Sol, SL
g2 .

We call these Thurston’s eight geometries, or simply the eight geometries.

Furthermore, if N is modelled on one of these eight geometries, that geometry
is unique ([Sco83, p. 476]). Seifert manifolds occupy precisely six of these
geometries ([Mar16, p. 364]):
Theorem 1.2.3. A closed orientable 3-manifold has a geometric structure mod-
elled on one of the following six geometries:

S 3 , R3 , S 2 × R, H2 × R, Nil, SL
g2

if and only it is a Seifert manifold.

It is worth noting that the Seifert manifolds are classified and placed accord-
ing to their geometries, as are the manifolds with the Sol geometry. Classifying
the manifolds modelled on H3 i.e. the hyperbolic manifolds, is still an open
problem ([Mor14, p. 3]).

3
 CHAPTER 2

Seifert manifolds as G-manifolds

In this chapter we introduce a way to define Seifert manifolds, namely as Seifert

fiber spaces. This point of view will allow us to show that Seifert manifolds
under certain conditions are S 1 -manifolds. We also introduce the basic theory
of G-manifolds that will be required to show this.

2.1 Seifert manifolds

This section will vaguely follow [Sco83, p.428-430], unless otherwise stated. The
aim is to introduce Seifert manifolds and their Thurston geometries.
Definition 2.1.1. A trivial fibered solid torus is the space S 1 × D2 where the
circles Fx = S 1 × {x} for x ∈ D2 are called fibers. A fibered solid torus is
the space S 1 × D2 finitely covered by a trivial fibered solid torus. If π is the
covering map, the fibers of a fibered solid torus are Fπ(x) := π(Fx ).
A fibered solid torus can be constructed from a trivial fibered solid torus
by cutting it open along {z} × D2 for any z ∈ S 1 , rotating one of the resulting
discs q/p times of a full turn and gluing the discs back together. Therefore we
use the notation T(p,q) for a solid torus, where p and q are coprime integers.
The fibers of T(p,q) can be parameterized as
q
Fx (t) = (e2πit , e2πit p x)

for x ∈ D2 .
We say that two fibered solid tori T(p,q) and T(p0 ,q0 ) are isomorphic if there
exists a diffemorphism between them sending fibers to fibers. We must have
p = p0 for two tori to be isomorphic, since they are p-fold and p0 -fold covered
by a trivial fibered solid torus, respectively. However it is sufficient to have
q = ±q 0 (mod p) for them to be isomorphic. We can verify this by looking at
how the diffeomorphisms

φ : T(p,q) → T(p0 ,q0 )

(z, x) 7→ (z, e2πint x)

for n ∈ Z and

ψ : T(p,q) → T(p0 ,q0 )

(z, x) 7→ (z, −x)

5
2. Seifert manifolds as G-manifolds

act on the fiber Fx ⊆ T(p,q) for x ∈ D2 . We get

q
φ(Fx (t)) = φ(e2πit , e2πit p x)
q
= (e2πit , e2πit e2πit p x)
q+pn
= (e2πit , e2πit p x)
= Fx ⊆ T(p,q+pn)
and
q
ψ(Fx (t)) = ψ(e2πit , e2πit p x)
q
= (e2πit , −e2πit p x)
−q
= (e2πit , e2πit p x)
= Fx ⊆ T(p,−q) .
This shows that we can choose the pair (p, q) such that p ∈ Z \ {0} and
0 ≤ q < p where p and q are coprime. We call (p, q) under these restrictions the
orbit invariants of T(p,q) . For q = 0 and any p, the fibered solid torus T(p,q) is
trivial, so we say that (1, 0) are the orbit variants of the trivial fibered solid
torus and denote it simply by T0 .
We can use these fibered solid tori to define a Seifert fiber space.
Definition 2.1.2. A Seifert fiber space is a 3-manifold M with a decomposition
of M into disjoint circles, called fibers, such that each circle has a neighbourhood
in M which is a union of fibers and is isomorphic to a fibered solid torus.
We note that it is possible to extend the definition to include neighbourhoods
which are diffeomorphic to fibered Klein bottles as well, but we will not need
that.
Definition 2.1.3. Let M be a Seifert manifold and Fx be the fiber at the point
x ∈ M . We say that Fx is singular if it is the central fiber (i.e. the fiber F0 (t)
in T(p,q) ) of a non-trivial fibered solid torus, and regular otherwise.
We note that there are only finitely many singular fibers, and they are all
separated.
The reason for the term fiber is that we can consider a Seifert fiber space
as a kind of circle bundle. If E is a Seifert fiber space we define B to be the
quotient of E obtained by identifying all fibers to a point. In particular, every
circle bundle is a Seifert fiber space with no singular fibers. In general B is not
a manifold, but it is something called an orbifold, as we will discuss in chapter
4. This means that E → B is not in general a circle bundle of manifolds, but it
is possible to make sense of it as a circle bundle of orbifolds. We will not need
this, but the space B is called the base space of E and makes sense topologically,
and it will be relevant for classifying the geometries of Seifert manifolds.
It is possible to define Seifert manifolds in terms of Dehn surgery. In [Mar16,
p. 310] it is shown that these are equivalent to Seifert fiber spaces. The
Seifert fiber space definition is more useful for our purposes, but because of this
equivalence we will refer to Seifert fiber spaces as Seifert manifolds from now
on.
From [Mar16, p. 311] we know that all Seifert manifolds M are characterized
by its base space and the orbit invariants of its singular fibers. So we write M =

6
 2.1. Seifert manifolds

(S, (p1 , q1 ), . . . , (pk , qk )) where S is the base space of M and (p1 , q1 ), . . . , (pk , qk )
are the orbit invariants of the singular fibers.
Much like compact 2-manifolds can be classified into geometries based on
their Euler characteristic, there is a similar result for Seifert manifolds. We now
state the definition of the relevant invariants ([Mar16, p. 37, p. 159, p. 312]).
Definition 2.1.4. The Euler characteristic of an n-complex X is given by the
integer
X n
χ(X) = (−1)i Ci
i=1

where Ci is the number of i-cells in X.

Definition 2.1.5. The Euler characteristic of a Seifert manifold

M = (S, (p1 , q1 ), . . . , (pk , qk ))

is given by
k  
X 1
χ(M ) = χ(S) − 1−
i=1
pi

Definition 2.1.6. The Euler number e of a Seifert manifold

M = (S, (p1 , q1 ), . . . , (pk , qk ))

is given by
k
X qi
e(M ) =
p
i=1 i

This is only defined up to modulo Z if M has a boundary.

While the notation M = (S, (p1 , q1 ), . . . , (pk , qk )) is not unique, the Euler
characteristic and Euler number are invariant under different notations. From
[Mar16, p. 364] these invariants are sufficient to classify the geometry of Seifert
manifolds, as shown in the following table:

χ>0 χ=0 χ<0

e=0 S2 × R R3 H2 × R
e 6= 0 S3 Nil SL
g2

7
2. Seifert manifolds as G-manifolds

2.2 G-manifolds
In this section we introduce basic definitions of G-manifolds and the equivariant
tubular neighbourhood theorem from [Jän68, p. 1-4].
Definition 2.2.1. Let G be a compact Lie group and M a smooth manifold. By
an action of G on M we will mean a smooth map G × M → M , where the
image of (g, x) is written gx, with the following properties:
(i) g1 (g2 x) = (g1 g2 )x (Compatibility)
(ii) 1x = x (Identity)
A smooth manifold M together with a smooth action by G on M is called
a G-manifold, simply denoted by M when the G-action is understood.
For each g ∈ G we can define the (left-)action of g by

lg : M → M
x 7→ gx

By the previous definition this map is smooth and has a smooth inverse lg−1 ,
hence being a diffeomorphism.
Some examples of G-manifolds include any Lie group acting on itself, and
this also induces an H-manifold structure on G for any subgroup H ⊆ G.
We now define the concepts we need to know about G-manifolds.
Definition 2.2.2. Let M be a G-manifold and let x ∈ M . Then the set Gx =
{gx | g ∈ G} is the orbit of the point x, and Gx = {g | gx = x} is the stabilizer
group of the point x. Further we denote by M/G the orbit space, that is the set
of all orbits {Gx | x ∈ M }, equipped with the quotient topology, that is: A set
of orbits is open in M/G whenever their union is open in M .
Let M be a G-manifold and x ∈ M . The Lie group G is a G-manifold by
definition, and hence also a Gx -manifold, since Gx is a subgroup of G. The
orbit space G/Gx is by definition

G/Gx = {(Gx )g | g ∈ G} = {gGx | g ∈ G}

which yields an isomorphism

G/Gx → Gx
gGx 7→ gx

showing the connection between the orbits and stabilizer groups in a different
way.
Definition 2.2.3. Let M and N be two G-manifolds. A map φ : M → N is said
to be equivariant if for every g ∈ G and x ∈ M we have

φ(gx) = gφ(x)

Definition 2.2.4. Let M be a G-manifold. A vector bundle over M together

with a G-action on its total space E is called a G-vector bundle over M , when
the fiber Ex is mapped isomorphically to Egx for g ∈ G.

8
 2.2. G-manifolds

The tangent bundle T M of a G-manifold M is a G-vector bundle in a

natural way, the map from Tx M to Tgx M being induced by the differential of
lg : M → M .
We want to show that if S is an equivariantly embedded submanifold of a
G-manifold M , then the normal bundle N S is a G-vector bundle over M .
First let G/H be a homogenous space, that is G is a Lie group and H ⊆ G
is a subgroup, and let E → G/H be a G-vector bundle. Denote the fiber at the
point 1H ∈ G/H by V . Then V is an H-module, because every h ∈ H fixes
the point 1H ∈ G/H, so it sends V to V .
Now let us consider the fiber bundle G ×H V → G/H, associated with the
principal H-fiber bundle G → G/H. Here, G ×H V is defined to be (G × V )/H
where the equivalence classes are [g, v] = [gh, h−1 v]. Then G ×H V → G/H is in
particular a vector bundle, and when we for g ∈ G declare that g[g, v] = [gg, v],
then G ×H V becomes a G-vector bundle over G/H. From the map [g, v] 7→ gv
we obtain a map

G ×H V → E
which is indeed a G-vector bundle isomorphism. We conclude that any
G-vector bundle over G/H is determined by its H-module at the point 1H.
Now let M be a G-manifold and x ∈ M . We denote by Nx = Tx M/Tx Gx
the normal space of the orbit Gx at the point x. The identification of Gx
with G/Gx as we showed earlier and the identification of G-vector bundles over
homogenous spaces shows that we can idnetify the normal bundle of Gx in M
with the bundle G ×Gx Nx → G/Gx .
We can now formulate the equivariant tubular neighbourhood theorem:
Theorem 2.2.5. Let M be a G-manifold. There exists an equivariant diffeo-
morphism between a G-invariant open neighbourhood of the zero section in
G ×Gx Nx and a G-invariant open neighbourhood of Gx in M , which sends the
zero section G/Gx to the orbit Gx in a canonical way.
This is actually just a version of the tubular neighbourhood theorem found
in [Mil74, p.115-117] for G-manifolds:
Theorem 2.2.6. Let M be a Riemannian manifold and S be a smoothly embedded
submanifold of M . There exists an open neighbourhood of S in M which is
diffeomorphic to the total space of the normal bundle which maps each point
x ∈ S to the zero normal vector at x.
We will not reproduce all the details of the proofs, but we will sketch the
proofs.
For the tubular neighbourhood theorem one denotes the total space of the
normal bundle as E and defines the set

E(ε) := {(x, v) ∈ E | kvk < ε}

where x ∈ S and v is a normal vector at x. It is then shown that this set maps
diffeomorphically to a neighbourhood of S in M by the exponential map

Exp : E(ε) → M
(x, v) 7→ γ(1)

9
2. Seifert manifolds as G-manifolds

where γ : [0, 1] → M is a geodesic arc with γ(0) = x and velocity vector equal
to v at 0, which exists for sufficiently small ε. It is then shown that E(ε) is
indeed diffeomorphic to E, proving the result.
For the equivariant version it is sufficient to show that we can find an
equivariant Riemannian metric on M for the tubular neighbourhood to be
equivariant. Here it is shown that for any Riemannian metric p on M one
can construct an equivariant Riemannian metric p on M . Note that the
diffeomorphism between E(ε) and E is not generally equivariant, so the result
does not extend to all of E unless M is complete.

10
 2.3. Circle manifolds

2.3 Circle manifolds

As mentioned in [Sco83, p. 430], a Seifert manifold M has the structure of a
S 1 -manifold (where S 1 has the Lie group structure of U (1)) if and only if we
can coherently orient the fibers. In this section we make sense of what this
means and fill out the details.
Firstly, not all S 1 -manifolds are Seifert manifolds. We need to restrict our-
selves to only look at those S 1 -manifolds with the following finiteness property:
Definition 2.3.1. A G-manifold is said to be finitely stabilized if every stabilizer
group is finite and only finitely many stabilizer groups are non-trivial.
As will be evident, the requirement that we can only have finitely many
stabilizer groups corresponds to the fact that Seifert manifolds have only finitely
many singular fibers. The requirement that all stabilizer groups have to be
finite corresponds to the fact that T(p,q) is finitely covered by T0 , i.e. that q/p
is rational.
Secondly, not every Seifert manifold is a S 1 -manifold. We need to have a
definition of what it means for fibers to be coherently oriented.
Definition 2.3.2. Let M be a Seifert manifold and let Fp be the fiber of p. A
point-wise fiber orientation µ of M is a choice of orientation of the tangent space
Tp (Fp ) for every p ∈ M . A point-wise fiber orientation is said to be continuous
if for every p ∈ M there exists a neighbourhood U 3 p and a continuous vector
field X on U such that Xp gives Tp (Fp ) the same orientation as µ (where Tp (Fp )
is identified with Tp (M )). A Seifert manifold M is said to be fiber orientable if
there exists a continuous point-wise fiber orientation of M .
The special case where π : M → B is a circle bundle where M is orientable
and B is not orientable, M will not be fiber orientable despite being a Seifert
manifold. Assuming the opposite, if X is a continuous vector field on U ⊆ M
corresponding to the fiber orientation of M and [X, Y, Z] is a continuous frame
on U corresponding to the orientation on M , then [Y |B , Z|B ] is a continuous
frame on π(U ) ⊆ B. Since this is true for all U , it would imply that B is
orientable, which is a contradiction.
We are now ready to prove the following:
Proposition 2.3.3. Let M be a compact 3-dimensional S 1 -manifold. If M is
finitely stabilized, then the orbits the S 1 -action give a Seifert structure on M .

Proof. We start by noting that all orbits of M are circles. The orbit of S 1 is
homeomorphic to S 1 /Sx1 . Since the stabilizer Sx1 is finite by assumption, the orbit
S 1 x is homeomorphic to a finite quotient of S 1 , which is again homeomorphic
to S 1 .
By the equivariant tubular neighbourhood theorem there exists a S 1 -
invariant neighbourhood U of S 1 x that is diffeomorphic to E(ε) ⊆ E. Here
E is the total space of the normal bundle of S 1 x in M and E(ε) is the set of
normal vectors of length less than ε for some ε > 0. Since we only have finitely
many non-trivial stabilizer groups, we can choose ε > 0 small enough such that
U contains no orbits with non-trivial stabilizer group other than possibly S 1 x
itself.
The identification of S 1 ×Sx1 Nx with E from the equivariant tubular neigh-
bourhood theorem shows that U is indeed a fibered solid torus. Since S 1 is

11
2. Seifert manifolds as G-manifolds

1-dimensional and M is 3-dimensional, Nx is 2-dimensional. Thus Nx is diffeo-

morphic to D2 , showing that E is a solid torus under the quotient by the finite
stabilizer group Sx1 . The S 1 -vector bundle structure on E(ε) inherited from E
reveals that every orbit in U winds p = |Sx1 | times around S 1 x, showing that U
has the structure of T(p,q) for some 0 ≤ q < p. Hence M is a Seifert manifold.


Proving the converse requires a few lemmas which we will now state and
prove.
Lemma 2.3.4. Let X and Y be π-related vector fields on the manifolds M and
N , respectively, where π : M → N is a smooth map. If γ is an integral curve
of X then π ◦ γ is an integral curve of Y .

Proof. By the chain rule we have

(π ◦ γ)∗,t = π∗,γ(t) ◦ γ∗,t

and by the definition of an integral curve we have

π∗,γ(t) ◦ γ∗,t = π∗,γ(t) ◦ Xγ(t) .

Finally by π-relatedness we have

π∗,γ(t) ◦ Xγ(t) = Y(π◦γ)(t)

showing that π ◦ γ is indeed an integral curve of Y . 

Lemma 2.3.5. There exists a smooth vector field X on the fibered solid torus
T(p,q) such that each integral curve is a fiber.

Proof. We use the fact that R is a Lie group to define the left-invariant vector
d
field Y generated by dt 0
∈ Te (R). This vector field is smooth by [Tu08, p.181].
We can use this to define a smooth vector field Xe on the infinite solid cylinder
2
R × D as the following composition:
π Y i
R × D2 − → T R ⊕ T D2 ' T (R × D2 ).
→ R −→ T R ,−

The projection π is smooth because projections are always smooth, the

map Y is smooth because Y is a smooth vector field, and the final inclusion
i is smooth because inclusions are always smooth. Hence the vector field
Xe = i ◦ Y ◦ π on R × D2 is smooth. Furthermore since Y is left-invariant, X
e is
invariant under the R-action

r(t, x) = (rt, x)

on R × D2 for r, t ∈ R and x ∈ D2 . The integral curves of X

e are the fibers
R × {x}.
The map

φ : R × D2 → S 1 × D2 ' T(p,q)
q
(t, x) 7→ (e2πit , e2πit p x)

12
 2.3. Circle manifolds

is a covering map sending each fiber of the cylinder to a fiber of the torus. We
will now show that the pushforward X := φ(X) e is a well-defined smooth vector
field on T(p,q) . It will then have the fibers of T(p,q) as integral curves, since φ
sends integral curves to integral curves by Lemma 2.3.4.
Since φ is a covering map and R × D2 is simply connected, the infinite
cylinder is the universal cover of T(p,q) . The deck transformation group is
homomorphic to the fundamental group Z of T(p,q) . If n ∈ Z then this group
acts on R × D2 as n · (t, x) 7→ (t + pn, x). Since X e is invariant under the
previously defined R-action, it is also invariant under this Z-action. Hence if V
is any connected component of φ−1 (U ) where U ⊆ Tp,q is evenly covered and
ψ : U → V is the inverse of π|V , then the composition X e ◦ ψ is independent on
the choice of V . Thus the composition

ψ X π∗
→ R × D2 −→ T (R × D2 ) −→
T(p,q) − T (T(p,q) )
e

is well-defined.
q
−2πit p
The map ψ can be expressed explicitly as ψ(z, x) = ( log(z)
2πi , e x), where
the value of log(z) lies in V . This is clearly smooth as it is the composition of
smooth elementary functions. The map X e is smooth because it is a smooth
vector field, and the differential π∗ is smooth because all differentials are smooth.
This means that X = π∗ ◦ X e ◦ ψ is a smooth vector field on T(p,q) .


Note that the vector field X constructed in the above proof makes T(p,q)
fiber oriented. This is because it is smooth, hence continuous, and restricts
to a basis for each fiber on each fiber. Furthermore the vector field −X has
the opposite orientation of X, so we can always find a smooth vector field
corresponding to either orientation.
Proposition 2.3.6. Every Seifert manifold M with oriented fibers is a finitely
stabilized U (1)-manifold.

Proof. Let {Uj } be a cover of the Seifert manifold M where each Uj is a fibered
solid torus, and let {Xj } be a collection of smooth vector fields on {Uj } with
fibers as integral curves, as constructed in 2.3.5. We choose our Xj ’s such that
they all have the same fiber orientation. We can now find a partition of unity
subordinate to Ui to construct a smooth vector field X on M where all the
integral curves are fibers. We need to modify this vector field in such a way
that the period of all the integral curves of non-singular fibers is the same.
For any fibered solid torus Uj ' T(p,q) in our cover we can pull back the
vector field X|Uj to a vector field X ej on R × D2 . Since we know that the images
of the integral curves of this vector field are the fibers R × {x}, this vector
field must be given by (X ej )(t,x) = f ∂
∂t (t,x) for some smooth, strictly positive
function f . Now let γx : R → R be the integral curve of the fiber R × {x}
(evaluated only in the first coordinate) such that γx (0) = 0. By the definition
of integral curves we have γx0 (t) = (X ej )γ (t) . In terms of calculus this leads to
x

13
2. Seifert manifolds as G-manifolds

the differential equation

γ̇x (t) = f (γx (t))
γ̇x (t)
= 1.
f (γx (t))
Now let F be any function such that Ḟ = f1 . Since f is smooth, so is F . We
then get
Ḟ (γx (t)) = 1
F (γx (t)) = t + a
γx (t) = F −1 (t + a).
Note that we can choose F such that a = 0 to avoid the extra constant.
Let tx be the value such that γx (tx ) = p. If Uj is a non-trivial fibred solid
torus and the fiber of x is non-singular, the period of the projection of γx in Uj
is the tx . Otherwise the period of the projection of γx in Uj is tx /p. We want
all non-singular fibers to have period 1, so let us consider a function c(x) that is
constant on each fiber and the vector field cX ej . This vector field will be smooth
if c(x) is smooth, and the images of the integral curves will be the same as for
Xej . If ωx : R → R is the integral curve of the fiber R × {x} (evaluated only in
the first coordinate) such that ωx (0) = 0, then we have ωx0 (t) = (cX ej )ω (t) . In
x
terms of calculus this leads to the differential equation

ω̇x (t) = c(x)f (ωx (t))

ω̇x (t)
= c(x)
f (ωx (t))
Ḟ (ωx (t)) = c(x)
F (ωx (t)) = c(x)t + b
ωx (t) = F −1 (c(x)t + b)
Since F was chosen such that a = 0, we get F (γx (t)) = t which implies
F (0) = 0. Since we defined ωx (0) = 0 this shows that b = 0 as well. By
substitution we now have ωx (t) = γx (c(x)t). Thus if tx is the period of γx , then
c(x)tx is the period of ωx . So to have a period equal to 1 we need to have
c(x) = 1/tx . This is smooth, because tx = γx−1 (p) where γx as a function of x is
the smooth flow of X, and because its differential is never zero γ −1 is smooth.
Because the period of an integral curve is independent of the choice of Uj ,
c(x) can be extended globally to x ∈ M . Hence cX is a smooth vector field on
M.
To define an S 1 -action on M we set
e2πit x := γx (t)
where γx : R → M is the integral curve of cX at x ∈ M such that γx (0) = x.
This action is smooth because cX is smooth. It is compatible:
e2πit (e2πis x) = e2πit γx (s)
= γγx (s) (t)
= γx (t + s)

14
 2.3. Circle manifolds

and the identity acts as the identity:

e2πi0 x = γx (0)
= x.

Every non-singular fiber has trivial stabilizer group, because they have period
1. Every singular fiber has finite stabilizer group of order p, because they have
period 1/p which means that

e2πit x = γx (t) = γx (t + q/p) = e2πitq/p x

for 0 ≤ q < p. This shows that M is a finitely stabilized S 1 -manifold.



From 2.3.3 and 2.3.6 we see that a Seifert manifold admits a finitely stabilzied
S 1 -structure if and only if it is fiber orientable. From the proofs we see that
we can calculate the invariants p1 , . . . , pk of a fiber oriented Seifert manifold
M = (S, (p1 , q1 ), . . . , (pk , qk )) by finding the orders of the non-trivial stabilizer
groups. Moreover the base space of a Seifert manifold is the orbit space M/S 1 ,
since the points of the orbit space are the fibers of M by definition. We will
show that this is an orbifold in chapter 4.

15
 CHAPTER 3

Examples of Seifert manifolds

3.1 Brieskorn manifolds

A 3-dimensional Brieskorn manifold Σ(p, q, r) for p, q, r ≥ 2 is the intersection
between the hyperplane in C3 given by
z1p + z2q + z3r = 0
and the odd sphere
S 5 = {(z1 , z2 , z3 ) ∈ C3 | |z1 |2 + |z2 |2 + |z3 |2 = 1}.
These are known for being 3-dimensional homology spheres when p, q and r
are relatively prime (see [Mil75, p.176]). This means that they have the same
homology groups as the n-sphere despite not being diffeomorphic to S n . We
are interested in these manifolds because they turn out to be Seifert manifolds.
We will prove this by giving them a finitely stabilized circle action.
Proposition 3.1.1. A Brieskorn manifold Σ(p, q, r) gets a Seifert structure from
the circle action m m m
u(z1 , z2 , z3 ) := (u p z1 , u q z2 , u r z3 )
where u ∈ S 1 and m is the least common multiple of p, q and r. If Σ(p, q, r)
is a homology sphere it has exactly three singular fibres with cyclic stabilizer
groups of order p, q and r.
Proof. We start by checking that the given circle action is indeed a group
action by checking that it satisfies the definition. It is clearly smooth since it
is a composition of smooth functions. Let z = (z1 , z2 , z2 ) ∈ Σ(p, q, r) for the
remainder of the proof. The the other two axioms are also satisfied:
m m m
(uv)(z) = ((uv) p z1 , (uv) q z2 , (uv) r z3 )
m m m m m m
= ((u) p (v) p z1 , (u) q (v) q z2 , (u) r (v) r z3 )
m m m
= u(v p z1 , v q z2 , v r z3 )
= u(vz)
and
m m m
1z = (1 p z1 , 1 q z2 , 1 r z3 )
= (z1 , z2 , z3 )
=z

17
3. Examples of Seifert manifolds

Next we need to check that the circle action is well-defined by checking that uz
is contained in Σ(p, q, r).
m m m
(u p z1 )p + (u q z2 )q + (u r z3 )r ) = um z1p + um z1q + um z1r
= um (z1p + z2q + z3r )
=0
because um 6= 0.
m m m 2m 2m 2m
|u p z1 |2 + |u q z2 |2 + |u r z3 |2 = |u| p |z1 |2 + |u| q |z2 |2 + |u| r |z3 |2
= |z1 |2 + |z2 |2 + |z3 |2
=1
because |u| = 1.
To calculate the stabilizer groups of the circle action we consider two cases.
The first case we consider is when z1 , z2 , z3 6= 0. Then u is in the stabilizer
group of z if we have
uz = z
m m m
(u z1 , u z2 , u z3 ) = (z1 , z2 , z3 )
p q r

m m m
up =uq =ur =1
The last line shows that the order of u must divide m m m
p , q and r . From
elementary number theory we can find k1 , k2 and k3 such that
 
m m m m m m
k1 + k2 + k3 = gcd , , .
p q r p q r
 
Let d = gcd m , m m
, m m m
p q r . Then dp , dq and dr are integers, hence p, q and r
divide m
d . But since m is defined to be the least common multiple of p, q and r
this means that d = 1. It now follows that
m m m
u1 = uk1 p +k2 q +k3 r
 m k1  m k2 m
k
3
u= up uq ur
u=1
Hence the stabilizer group of z is trivial.
The remaining case is to check when either of z1 , z2 or z3 is equal to zero.
Note that we must have at least two non-zero coefficients in order to satisfy the
requirements to be a point on the Brieskorn manifold. We start by checking for
z3 = 0. Similar to before we get
uz = z
m m m
(u z1 , u z2 , u 0) = (z1 , z2 , 0)
p q r

m m
u p = u q = 1.
In this case the order of u must divide m m
p and q . We can find k1 and k2 such
that  
m m m m
k1 + k2 = gcd , .
p q p q

18
 3.1. Brieskorn manifolds

 
m m
To simplify notation we set d = gcd p, q . This leads to
m m
ud = uk1 p +k2 q
 m k1  m k2
= up uq
=1
 
m m
Hence the order of u divides gcd p, q , which means that we have

2πin
u=e d .

with n ∈ Z. This shows that the stabilizer group of z is the cyclic group of
order d. In the particular case where p, q and r are relatively prime we get that
 
m m
d = gcd ,
p q
 
pqr pqr
= gcd ,
p q
= gcd (qr, pr)
= r.

 findthat the stabilizer group of z for z2 = 0 is the cyclic

By symmetry we also
group of order gcd m m
p, r (which is equal to q when p, q and r are relatively
prime,)
 and  the stabilizer group of z when z1 = 0 is the cyclic group of order
m m
gcd q , r (which is equal to p when p, q and r are relatively prime).
It now remains to show that we only have finitely many orbits with non-trivial
stabilizer group. Since we just showed that the only points with non-trivial
stabilizer group are the ones with either z1 , z2 or z3 equal to zero, all orbits
with non-trivial stabilizer groups are contained in the sets

Vk = {z ∈ Σ(p, q, r) | zk = 0}

for k = 1, 2, 3. We need to show that each set Vk splits into finitely many orbits.
Assume that z3 = 0. We then get the following relationship:

z1p + z2q = 0
z1p = −z2q
|z1 |p = |z2 |q
q
|z1 | = |z2 | p
p q
1 − |z2 |2 = |z2 | p
2q
1 − |z2 |2 − |z2 | p = 0

The final equation has a strictly decreasing function of |z2 | on the left-hand side.
This means that the equation has a unique solution |z2 |, and since the function
yields 1 for |z2 | = 0 and −1 for |z2 | = 1, the solution lies between 0 < |z2 | < 1.
This shows that the length of z2 (and hence z1 ) is only dependent on p and q.

19
3. Examples of Seifert manifolds

Now every orbit of (z1 , z2 , 0) ∈ V3 contains

m
a point on the form (|z1 |, z2 , 0),
since we can find u ∈ S 1 such that u p z1 = |z1 |. We can now reduce our
problem to looking for the orbits of the points (|z1 |, z2 , 0). Let
W = {(|z1 |, z2 , 0) ∈ V3 }
be the set of such points. These points have the property that
|z1 |p + z2q = 0
|z1 |p + (|z2 |eθi )q = 0
|z2 |q θi q
1+ (e ) = 0
|z1 |p
(eθi )q = −1
which has q solution for θ, hence |W | = q.
The number of orbits of V3 under the S 1 -action is the same as the the
1
number of orbits in W under the action of the subset of Sm that leaves W fixed
in V3 . This subgroup contains the elements u such that u p = 1, hence it is the
cyclic subgroup C mp of order mp . The stabilizer group of the C p -action on W
m

contains the elements u such that

u(|z1 |, z2 , 0) = (|z1 |, z2 , 0)
m
u q z2 = z2
m
uq =1
Since the order of u also divides m p , this means that the stabilizer group is
cyclic of order d = gcd( m ,
p q
m
). Since we are working with a finite group, the
order of the orbit of (|z1 |, z2 , 0) is the order of C mp divided by the order of the
stabilizer of (|z1 |, z2 , 0), that is
m
p m
= .
d pd
The number of orbits is therefore the order of W divided by the order of the
orbit, that is
q pqd
m =
pd m
pq gcd( m m
p, q )
=
m
gcd(mp, mq)
=
m
= gcd(p, q)
This means that V3 splits into gcd(p, q) orbits.
By symmetry we also get that V2 splits into gcd(p, r) orbits and that V1
splits into gcd(q, r) orbits. This shows that are finitely many orbits with finite
stabilizer group. In particular we have only three singular orbits when p, q and
r are relatively prime, since each Vk represents a single orbit.


20
 3.2. Spherical space forms

3.2 Spherical space forms

In general a spherical space form is the quotient of S n with a finite subgroup of
its isometry group. For n = 3 and Γ ⊆ SU (2) finite, the spherical space forms
S 3 /Γ are Seifert manifolds. We will show this by giving them an appropriate
circle action.
There are multiple ways to think of S 3 , and in this case it turns out to be
useful to think of S 3 as the unit quaternions which are also diffeomorphic to
the Lie Group SU (2)

SU (2) ' {a + bi + cj + dk | a, b, c, d ∈ R, a2 + b2 + c2 + d2 = 1}
= {α + βj | α, β ∈ C, |α|2 + |β|2 = 1}
⊆ H.

Now S 3 inherits the Lie group structure of SU (2), and by letting it act
on itself it is therefore an SU (2)-manifold. Taking this further we know that
U (1) ⊆ SU (2) and that S 1 can be identified with U (1), so we also inherit an
S 1 -action on S 3 .
From [Mar16, p. 365] we have a group homomorphism

φ : SU (2) → SO(3)

with kernel {1, −1}. The preimage under φ of a subgroup of SO(3) is called
a binary group, and Martelli shows that all finite subgroups Γ ∈ SU (2) up
to conjugation are either cyclic or binary. The finite subgroups of SO(3) are
the cyclic groups Cn , the dihedral groups D2m , the tetrahedral group T12 , the
octahedral group O24 and the icosahedral group I60 . Thus the finite subgroups
of SU (2) up to conjugation are known to be the cyclic groups Cn , the binary
∗ ∗
dihedral groups D4n , the tetrahedral group T24 , the binary octahedral group
∗ ∗
O48 and the binary icosahedral group I120 . Thought of in terms of quaternions
they can be expressed as follows:
2πin
Cm = {e m | 0 ≤ n < m} (Cyclic groups)
πin πin
∗
D4m = {e , e j | 0 ≤ n < 2m}
m m (Binary dihedral groups)
 
∗ ±1 ± i ± j ± k
T24 = D8 ∪ (Binary tetrahedral group)
2
 
∗ ∗ a+b
O48 = T24 ∪ √ | a, b ∈ D8 , a 6= b (Binary octahedral group)
2
 √ √ 
∗ ∗ a ( 5 − 1)b ( 5 + 1)c
I120 = T24 ∪ ± ± ± | (a, b, c, d) = σ(1, i, j, k)
2 4 4
(Binary icosahedral group)

for all even permutations σ and where D8 = Q8 = {±1, ±i, ±j, ±k} is the set
of Lipschitz units.
The key to define a circle action is to let Γ ⊆ SU (2) act on S 3 from the
right and let S 1 act on S 3 from the left to obtain a S 1 -action on S 3 /Γ. We do
this by defining
g[x] := [gx]

21
3. Examples of Seifert manifolds

for g ∈ S 1 and [x] ∈ S 3 /Γ. This is well-defined because

g(xh) = (gx)h ∈ [gh].
This S -action does not always give rise to a Seifert structure on S 3 /Γ, however.
1

In the case where −1 ∈ Γ (that is, Γ having even order) all the stabilizer groups
would contain −1, so in that case we use the fact that the quotient group
S 1 /{±1} is diffeomorphic to S 1 by the map
φ : S 1 /{±1} → S 1
[eθi ] 7→ e2θi
and define a circle action on S 3 /Γ by using the quotient group instead. We do
this by setting
g[x] := [gx]
for [g] ∈ S 1 /{±1}. Note that this definition does not depend on the repre-
sentative g, since we either have g = eiθ or g = −eiθ , and [gx] = [−gx] when
1
the order of Γ is even. It is easy to see that if S[x] is the stabilizer group of
3 1 1 1
[x] ∈ S /Γ under the S -action, then (S /{±1})[x] = (S[x] )/{±1}. Thus we
can calculate the stabilizer groups of the S /{±1}-action from the S 1 -action.
1
1
In particular if S[x] has finite order n, then (S 1 /{±1})[x] has order n/2.
For the remainder of this section let G denote S 1 for Γ of odd order and
denote S 1 /{±1} for Γ of even order. We will now prove that this G-action yields
a Seifert structure on S 3 /Γ. We start by proving the following proposition.
Proposition 3.2.1. Let Γ ∈ SU (2) be a finite subgroup. The number of orbits
of the set  
Im(h)
Γ0 = | h ∈ Γ, Im(h) 6= 0
|Im(h)|
under the action of Γ-conjugation is the same as the number of orbits of S 3 /Γ
with non-trivial stabilizer group under the previously defined G-action.
Proof. The stabilizer group of [x] ∈ S 3 /Γ by the G-action is given by
G[x] = {g ∈ G | [gx] = [x]}
= {g ∈ G | ∃h ∈ Γ, gx = xh}
= {g ∈ G | ∃h ∈ Γ, g = xhx}.
In other words, g is in the stabilizer group of [x] if and only if there exists h ∈ Γ
such that g = xhx. To find the elements [x] with non-trivial stabilizer group we
need to solve the equation g = xhx for g with non-zero imaginary part, that is
g = cos(ψ) + i sin(ψ) for sin(ψ) 6= 0. We note that any quaternion h ∈ H can be
written as h = cos(θ) + q sin(θ) where q ∈ H0 is the normalized imaginary part
of h (see for instance [Frø19]). The set Γ0 is the set of normalized imaginary
parts of elements h ∈ Γ, so we can write h = cos(θ) + q sin(θ) for q ∈ Γ0 . We
then get that
g = xhx
cos(ψ) + i sin(ψ) = x(cos(θ) + sin(θ)q)x
cos(ψ) + i sin(ψ) = x cos(θ)x + x sin(θ)qx
cos(ψ) + i sin(ψ) = cos(θ) + sin(θ)xqx

22
 3.2. Spherical space forms

cos(ψ) = cos(θ) sin(ψ)i = sin(θ)xqx

The bottom two equations imply that sin(θ) = ± sin(ψ) which yields
xqx = ±i.
Thus [x] ∈ S 3 /Γ has non-trivial stabilizer group if and only if there exists
q ∈ Γ0 such that xqx = ±1. We can improve this by noting that if q ∈ Γ0 is
the normalized imaginary part of h ∈ Γ then the normalized imaginary part of
h ∈ Γ is −q ∈ Γ0 , so if xqx = −i then x(−q)x = i. With this improvement we
see that [x] ∈ S 3 /Γ has non-trivial stabilizer group if and only if there exists
q ∈ Γ0 such that xqx = i.
Now we need to check that conjugation by Γ is a group action on Γ0 . Clearly
conjugation is compatible because h2 h1 = h1 h2 for h1 , h2 ∈ Γ, and h = 1 acts
as the identity. So we just need to check that conjugation by Γ is a closed
action. We know that conjugation by h ∈ Γ (for Γ of even order) is the rotation
of φ(h) ∈ SO(3), and the normalized imaginary part q ∈ Γ0 of h is a point on
the axis of rotation of φ(h). Since SO(3) is a group of symmetries, it preserves
the axes of rotation, hence conjugation by Γ leaves Γ0 fixed. (In the case where
the order of Γ is odd we have Γ = Cn for n odd, which has the same properties
as Cn for even n in terms of conjugation.) The orbit of q ∈ Γ0 is the set of
conjugations hqh for h ∈ Γ.
Let x, xh ∈ S 3 for h ∈ Γ be two representatives for [x] ∈ S 3 /Γ and let [x]
be on a singular G-orbit. Then there exists q, q 0 ∈ Γ0 such that
xqx = (xh)q 0 xh = i
which shows that q = hq 0 h. Thus q and q 0 are Γ-conjugate i.e. on the same
orbit in Γ0 . If g[x] ∈ S 3 /Γ for g ∈ G is another point on the same orbit as [x]
then there exists u ∈ G and h ∈ Γ such that
g(ux) = (ux)h
(ugu)x = xh
gx = xh
showing there exists q ∈ Γ0 such that
xqx = (gx)q(gx) = i.
We have now shown that for every G-orbit in S 3 /Γ there exists an orbit in Γ0 .
Conversely, let q, q 0 ∈ Γ0 be Γ-conjugate, that is q 0 = hqh for h ∈ Γ. There
exist [x], [y] ∈ S 3 /Γ on singular orbits such that
xqx = yq 0 y = i.
This leads to
xqx = (yh)q(yh) = i.
Due to the Lie group structure on S 3 ' SU (2) we can find an element g ∈ SU (2)
such that x = g(yh). We then get
(yh)q(yh) = i
g(xqx)g = i
gig = i

23
3. Examples of Seifert manifolds

showing that conjugating by g leaves i fixed. The only non-identity elements

that leave i fixed under conjugation are the elements that rotate around the
axis through i, so we get

g = cos(θ) + i sin(θ) ∈ G.

Thus we have x = gyh for g ∈ G and h ∈ Γ, showing that [x] and [y] are on the
same G-orbit in S 3 /Γ. Therefore for every orbit in Γ0 there exists a G-orbit in
S 3 /Γ. 

From this proposition it readily follows that S 3 /Γ is a Seifert manifold.

Corollary 3.2.2. Let Γ ⊆ SU (2) be finite. Then S 3 /Γ is a Seifert manifold.

Proof. From the previous proposition we know that the number of singular
G-orbits of S 3 /Γ is the same as the number of orbits of Γ0 under Γ-conjugation.
Since Γ0 is finite, hence has finitely many orbits, there are only finitely many
singular orbits of S 3 /Γ.
Now recall that g ∈ G is in the stabilizer group of [x] if and only if there
exists h ∈ Γ such that g = xhx. Since there are only finitely many h ∈ Γ, there
are only finitely many g ∈ G[x] , hence all stabilizer groups are finite.
Since the G-action on S 3 /Γ is finitely stabilized, S 3 /Γ is a Seifert manifold.


The next thing to do is to calculate the stabilizer groups of the spherical

space forms. We denote by Cl(q) for q ∈ Γ0 the orbit of q under conjugation by
Γ. By Cl(q)∗ for q ∈ Γ0 we shall mean the set

Cl(q)∗ := {h = cos(θ) + q 0 sin(θ) ∈ Γ | q 0 ∈ Cl(q)}.

Now if [x] ∈ S 3 /Γ is a point with non-trivial stabilizer group corresponding

to Cl(q) and g ∈ G has non-real imaginary part, all h ∈ Γ that solve the
equation
gx = xh
are precisely the h ∈ Cl(q)∗ . Due to the Lie group structure on S 3 we have
exactly one solution g ∈ S 1 for each solution h ∈ Γ, so the number of non-real
1
elements in the stabilizer group S[x] is |Cl(q)∗ |. Thus G[x] has order |Cl(q)∗ | + 1
∗
if Γ has odd order, and order |Cl(q) |/2 + 1 if Γ has even order (since we have to
add the identity element, which is real). We now note that all finite subgroups
of S 1 are cyclic, hence uniquely decided by its order.
We now have all we need to find the stabilizer groups of S 3 /Γ for all cases
of Γ.

Cyclic groups
Let Γ = Cn . Then Γ0 = {±i}. Since Γ consists of rotations about the axis
through i and −i, both i and −i are left fixed, so the orbits are Cl(i) and
Cl(−i).
Every non-real element of Cn has normalized imaginary part ±i, hence all
non-real elements are sent to both conjugacy classes. Therefore if n is odd we
have |Cl(±i)∗ | = n−1, so both the orbit of Cl(i) and Cl(−i) have stabilizer group

24
 3.2. Spherical space forms

of order n − 1 + 1 = n. If instead n is even we have |Cl(±i)∗ | = n − 2, so both

the orbit of Cl(i) and Cl(−i) have stabilizer group of order (n − 2)/2 + 1 = n/2.
In summary, S 3 /Cn has two singular orbits when n ≥ 3, with stabilizer
groups Cn for n odd, and with stabilizer groups Cn/2 for n even.

Binary dihedral groups

πin
∗
Let Γ = D4m . Then Γ0 = {±i, e m j | 0 ≤ n < 2m}. Now Γ consists of
nπ
m -rotations about the axis through i and π-rotations about the axis through
πin nπ
e j. The m -rotations about the axis through i leave i and −i fixed while
m
πin1 πin2
taking any e m j to any other e m j. The π-rotations about the axis through
πin πin πin
e m j take i to −i and e m j to −e m j. Hence we get the two orbits Cl(i) and
Cl(j).
πin
Every non-real element e m has normalized imaginary part ±i and every
πin
element e m j is itself normalized and imaginary. Therefore we get |Cl(i)∗ | =
2m − 2 and |Cl(j)∗ | = 2m. So the orbit of Cl(i) has stabilizer group of order
(2m − 2)/2 + 1 = m, and the orbit of Cl(j) has stabilizer group of order
(2m)/2 + 1 = m + 1.
∗
In summary, S 3 /D4m has two singular orbits when m ≥ 2, with stabilizer
groups Cm and Cm+1 . For m = 1 it has one singular orbit with stabilizer group
C2 .

Binary groups of platonic solids

The final three cases are when Γ is the binary group of a platonic solid. We
know that Γ0 consists of points on the rotational axes of the platonic solid
that Γ corresponds to. We know that all these axes pass through either a
vertex, the midpoint of an edge or the centroid of a face on the given platonic
solid. Hence the set Γ0 actually consist of the normalized vertices, normalized
midpoints of edges and normalized centroids of faces. Conjugating Γ0 by Γ
is the same as rotating the platonic solid corresponding to Γ such that every
vertex is sent to another vertex, every edge is sent to another edge and every
face is sent to another face. It is therefore clear that Γ0 has precisely three
orbits under Γ-conjugation, one consisting of normalized vertices, one consisting
of normalized midpoints of edges and one consisting of normalized centroids of
faces. Identifying which orbit the normalized part of h ∈ Γ can simply be done
by considering the real part of h. The real part of h represents the rotation
when conjugating by h, and different types of rotational axes have different
rotations (see for instance [CR17, p. 20]). For the tetrahedron we have

i. Rotating 2π 4π
3 and 3 about the axis through a vertex and the centroid of
the opposing face.

ii. Rotating π about the axis through the midpoints of opposing edges.

For the cube we have

π 3π
i. Rotating 2, π and 2 about the axis through the centroid of opposing
faces.

ii. Rotating π about the axis through the midpoints of opposing edges.

25
3. Examples of Seifert manifolds

2π 4π
iii. Rotating 3 and 3 about the axis through opposing vertices.

For the dodecahedron we have

i. Rotating 2π 4π
5 , 5 ,
6π
5 and 8π
5 about the axis through the centroid of
opposing faces.

ii. Rotating π about the axis through the midpoints of opposing edges.

2π 4π
iii. Rotating 3 and 3 about the axis through opposing vertices.

The only thing we have to be careful about is the vertex-face rotation of

the tetrahedron, as in every other case we know whether Cl(q) represents the
vertices, edges or faces from the real part of h = cos(θ) + q sin(θ) ∈ Γ. In the
case of the tetrahedron, because every vertex is opposed by a face, we have that
Cl(q) and Cl(−q) are different orbits unless q is a normalized edge. In this case
every element h ∈ Γ belongs to Cl(q)∗ if and only if it also belongs to Cl(−q)∗ .
∗
Now let Γ = T24 . From the above discussion we see that the three orbits
are the normalized edges Cl(i), and the normalized vertices and faces Cl( i+j+k
√
3
)
and Cl( −i−j−k
√
3
) for the tetrahedron. We further see that

Cl(i)∗ = {±i, ±j, ±k}

and
 ∗ ∗ 
 
i+j+k −i − j − k ±1 ± i ± j ± k
Cl √ √= Cl = .
3 3 2
 ∗  ∗
−i−j−k
Thus we have |Cl(i)∗ | = 6 and |Cl i+j+k
√
3
| = |Cl √
3
| = 16. So the
orbit of Cl(i)∗ has stabilizer group of order 6/2+1 = 4, and the orbits Cl( i+j+k
√
3
)
and Cl( −i−j−k
√
3
) have stabilizer groups of order 16/2 + 1 = 9.
∗
Next let Γ = O48 . We now have the orbits of normalized faces Cl(i),
normalized vertices Cl( i+j+k
√
3
) and of normalized edges Cl( i+j
√ ) of the cube,
2
with

|Cl(i)∗ | = 18
 ∗
i+j+k
Cl √ = 16
3
 ∗
i+j
Cl √ = 12
2

So the orbit of Cl(i) has stabilizer group of order 18/2 + 1 = 10, the orbit of
Cl( i+j+k
√
3
) has stabilizer group of order 16/2 + 1 = 9, and the orbit of Cl( i+j
√ )
2
has stabilizer group of order 12/2 + 1 = 7.
∗
Finally let Γ = I120 . We now have the orbits of normalized
q √edges q Cl(i), nor-
√

i+j+k 5+ 5 5− 5
malized vertices Cl( 3 ) and of normalized faces Cl 10 i + 10 j
√

26
 3.2. Spherical space forms

of the dodecahedron, with

|Cl(i)∗ | = 30
 ∗
i+j+k
Cl √ = 40
3
∗
√ √
s s
5 + 5 5 − 5
Cl  i+ j  = 48
10 10

So the orbit of Cl(i) has stabilizer group of order 30/2 + 1 = 16, the orbit
of Cl( i+j+k
√
3
) has stabilizer group of order 40/2 + 1 = 21, and the orbit of
q √ q √ 
5+ 5 5− 5
Cl 10 i + 10 j has stabilizer group of order 48/2 + 1 = 25.
∗
In summary, S 3 /T24 has three singular orbits, one with stabilizer group C4
∗
and two with stabilizer group C9 . S 3 /O48 has three singular orbits with stabilizer
∗
groups C10 , C9 and C7 . S 3 /I120 has three singular orbits with stabilizer groups
C16 , C21 and C25 .

27
 CHAPTER 4

Orbifolds

In this chapter we introduce the concept of an orbifold and show in particular

that the base space of a fiber oriented Seifert manifold is an orbifold.
An orbifold is a generalization a manifold where we allow charts to be finite
quotients of open sets in Rn . The following definition is due to [Thu, ch.13, p.6]
Definition 4.0.1. An orbifold O is a Hausdorff space XO with a covering of open
sets {Ui } closed under finite intersections. To each Ui there is an associated
finite group Γi acting on an open subset U ei of Rn and a homeomorphism
φi : Ui ' Ui /Γi . Additionaly, for every inclusion Ui ⊆ Uj there should be an
e
injective homomorphism fij : Γi ,→ Γj and a Γi -equivariant smooth embedding
ψij : Ui ,→ Uj such that φj ◦ ψij = φj .
The triple (Ui , Γi , φi ) is called an orbifold chart. We will add the additional
restriction that Γi ⊆ O(n), like in [Mar16, p.101].
The following result gives us a way of constructing orbifolds as orbit spaces
of finitely stabilized G-manifolds.
Proposition 4.0.2. Let G be a compact Lie-group and let P be a finitely stabi-
lized G-manifold. Then M = P/G is an orbifold.

Proof. Let P and G be n- and m-dimensional, respectively, and let N Gx be the

normal bundle of Gx ⊂ P . Since Gx ' G/Gx and Gx is finite, Gx must also
be m-dimensional. Hence each fiber of N Gx is k-dimensional for k = n − m.
Call all points with trivial stabilizer group regular, and points with non-trivial
stabilizer group singular.
By the equivariant tubular neighbourhood theorem we can choose a G-
invariant metric on P such that the subset E(ε) ⊆ E of the total space E of
N Gx is equivariantly diffeomorphic to some neighbourhood Gx(ε) of Gx for
sufficiently small ε > 0 with respect to this metric.
Now let U (ε) be the image of the restriction E(ε)|Nx by the exponential
map, where Nx is the normal space at x. This is a k-submanifold of P , since
E(ε)|Nx is an open ε-ball of the normal space Nx which is a k-vector space,
and it is embedded in P by the exponential map. Thus we have a chart
φ : U (ε) ← U ⊂ Rk .
Choose ε > 0 small enough such that Gx(ε) (and hence also U (ε)) contains
only regular points, except possibly the points of Gx. This is possible since we
only have finitely many non-trivial stabilizer groups.

29
4. Orbifolds

Now let V be the image of U (ε) under the projection π. From the projection

π : U (ε) → V ⊂ M

we note that V is open, since π −1 (V ) = Gx(ε) is open.

Recall that U (ε) is Gx -invariant, because if h ∈ Gx and v ∈ Nx then
hv ∈ Nhx = Nx . Thus the quotient map θ : U (ε) → U (ε)/Gx is well-defined.
Since the induced G-action from P (and hence the induced Gx -action) acts
trivially on M (and hence on V ) we can factor π through θ:
θ ψ
U (ε) U (ε)/Gx V

π=ψ◦θ
The map ψ sends each equivalent class [y] = {gy | y ∈ U (ε), g ∈ Gx } to
the orbit Gy. This is clearly injective, because if Gy 6= Gz then [y] 6= [z] since
[y] ⊂ Gy and [z] ⊂ Gz. ψ is also surjective because π is a projection. The
continuity of ψ and its inverse is clear from the topologies of U (ε)/Gx and
V : A subset {[y] ∈ U (ε)/Gx | y ∈ W } of U (ε)/Gx is open if and only if W
is open in U (ε). A subset {Gy ∈ V | y ∈ W } is open if and only if the set
{gy ∈ P | g ∈ G, y ∈ W } is open in P . However, W is open in U (ε) if and only
if it is the restriction of an open set in P , showing that a subset of M is open
if and only if it is the image of a subset of U (ε)/Gx under ψ. Hence ψ is a
homeomorphism.
The triple (V, Gx , φ ◦ ψ) an oribfold chart, because U (ε) is homeomorphic
to U ⊂ Rk , which means that U (ε)/Gx is homeomorphic to U/Gx . Since Gx is
distance preserving it is orthogonal, and since it is also finite it is homeomorphic
to a finite subgroup of O(n).
By our construction all the transition maps between charts are trivially
compatible, since none of the charts containing singular points intersect. We
complete the orbifold atlas by adding every finite intersection of every chart.
Finally, from [Lee12, p.543] we have that M is Hausdorff and second count-
able since G is compact and acts smoothly on P .


As mentioned earlier, the orbit space M/S 1 of a fiber oriented Seifert

manifold M is the base space of M . We have now shown that this space has an
orbifold structure.

Final remarks
We have shown that we can find the invariants (p1 , . . . , pk ) and base space S
of a fiber oriented Seifert manifold M by considering it as a finitely stabilized
S 1 -manifold. This allows us to calculate the Euler characteristic χ of M .
To calculate the Euler number e of M we need to also know the invariants
q1 , . . . , qk . These invariants are decided by the particular way S 1 acts on M ,
which means that it should be able to recover them from the projection map to
M/S 1 somehow.
By developing a theory of smooth orbifolds, similar to smooth manifolds, we
might be able to calculate the Euler number e of M directly by looking at the
projection M → M/S 1 . We leave others to investigate this perspective further.

30
 Bibliography

[CR17] Christophersen, J. A. and Ranestad, K. Geometri - MAT 2500. https:

/ / www. uio. no / studier / emner / matnat / math / MAT2500 / ressurser /
geometri2500h17.pdf. University of Oslo, 2017.
[Frø19] Frøyshov, K. A. Matrix groups. http://www.uio.no/studier/emner/
matnat/math/MAT4500/h19/beskjeder/matrix.pdf. University of Oslo,
2019.
[Jän68] Jänich, K. Differenzierbare G-Mannigfaltigkeiten. Berlin, 1968.
[Lee12] Lee, J. M. Introduction to Smooth Manifolds. eng. Vol. 218. Graduate
Texts in Mathematics. New York, NY: Springer New York, 2012.
[Mar16] Martelli, B. “An Introduction to Geometric Topology”. In: (2016).
[Mil74] Milnor, J. W. Characteristic classes. eng. Princeton, N.J, 1974.
[Mil75] Milnor, J. “On the 3-dimensional Brieskorn manifolds M(p, q, r)”.
eng. In: Knots, Groups, and 3-Manifolds 84 (1975), pp. 175–225.
[Mor14] Morgan, J. The geometrization conjecture. eng. Providence, 2014.
[Sco83] Scott, P. “The Geometries of 3-Manifolds”. eng. In: Bulletin of the
London Mathematical Society 15.5 (1983), pp. 401–487.
[Thu] Thurston, W. P. The geometry and topology of 3-manifolds : 6. eng.
[Tu08] Tu, L. An Introduction to Manifolds. eng. Universitext. New York,
NY: Springer New York, 2008.

31