Geometry Dynamics and Topology of FoliationsA First Course

GEOMETRY,
DYNAMICS AND
TOPOLOGY OF
FOLIATIONS
A First Course

This page intentionally left blank

GEOMETRY,
DYNAMICS AND
TOPOLOGY OF
FOLIATIONS
A First Course
Bruno Scardua
Federal University of Rio de Janeiro, Brazil
Carlos Arnoldo Morales Rojas

Federal University of Rio de Janeiro, Brazil
World Scientific
NEW JERSEY LONDON SINGAPORE BEIJING SHANGHAI HONG KONG TAIPEI CHENNAI TOKYO
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office 57 Shelton Street, Covent Garden, London WC2H 9HE
Library of Congress Cataloging-in-Publication Data

Names: Scardua, Bruno. | Rojas, Carlos Arnoldo Morales.
Title: Geometry, dynamics, and topology of foliations : a first course / by
Bruno Scardua (University of Rio de Janeiro, Brazil),
Carlos Arnoldo Morales Rojas (Federal University of Rio de Janeiro, Brazil).
Description: New Jersey : World Scientific, 2017. | Includes bibliographical references and index.
Identifiers: LCCN 2016059937 | ISBN 9789813207073 (hardcover : alk. paper)
Subjects: LCSH: Foliations (Mathematics) | Differential topology.
Classification: LCC QA613.62 .S33 2017 | DDC 514/.72--dc23
LC record available at https://lccn.loc.gov/2016059937
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
Copyright 2017 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.
Printed in Singapore
Dedicated to our families


Preface
The Geometrical Theory of Foliations is one of the elds in Mathematics

that gathers several distinctive domains such as; Topology, Dynamical Sys-
tems, Dierential Topology and Geometry, etc. It originated from the works
of C. Ehresmann and G. Reeb ([Ehresmann (1947)]), ([Ehresmann and Reeb
(1944)]). The huge development has allowed a better comprehension of sev-
eral phenomena of mathematical and physical nature. Classical theorems,
like the Reeb stability theorem, Haeigers theorem, and Novikovs compact
leaf theorem, are now searched for holomorphic foliations. Several authors
have began to investigate such phenomena (e.g. C. Camacho, A. Lins Neto,
E. Ghys, M. Brunella, R. Moussu, S. Novikov and others). The study of
such eld presumes knowledge of results, techniques of the real case and
superior familiarity with the classical aspects of Holomorphic Dynamical
Systems.
There is a number of important books dedicated to the study of folia-
tions, specially in the non-singular smooth framework. We shall not list all
of them, but we cannot avoid mentioning the books of Camacho and Lins
Neto ([Camacho and Lins-Neto (1985)]), Candel and Conlon ([Candel and
Conlon (2000)]), C. Godbillon ([Godbillon (1991)]) and Hector and Hirsch
([Hector and Hirsch (1987)]), which are among our favorite. Each of these
has inuenced in our text. From the choice of topics, to the path taken
in some demonstrations. In our viewpoint, the book of Camacho and Lins
Neto is important for its wise choice of topics, and for aiming at the geome-
try of foliations. The book of Hector and Hirsch has inuenced us specially
in the interplay between geometry and dynamics of foliations. The book of
C. Godbillon is very interesting for the wide range of topics that are cov-
ered. The proofs are elegant, usually short, but still precise. Finally, the
book of Candel and Conlon, divided into two volumes, is a very detailed
vii
viii Geometry, Dynamics and Topology of Foliations - a first course
introduction to the general theory of foliations. There, one can nd a very

complete exposition of the important results in the theory of codimension
one foliations. Special attention is given to the theory of minimal sets. This
material has greatly inuenced our exposition here. Their book is therefore
a natural complement to any introduction to the theory of foliations, and
should used as a reference any further information and courses.
Finally, we would like to mention that our personal interest for foliations
has been greatly supported by some amazing works from several authors.
Besides the authors mentioned above or in the text itself we wish to men-
tion P. Schweitzer, D. Calegari, P. Molino, S. Fenley, P. Thondeur and T.
Tsuboi.
These notes are mainly introductory and only cover part of the basic
aspects of the rich theory of foliations. In particular, additional extensive
information in some of the results presented here, may be searched in the
bibliography we give. We have tried to clarify the geometry of some classical
results and provide motivation for further study. Our goal is to highlight
this geometrical viewpoint despite some loss (?) of formalism. We hope that
this text may be useful to those who appreciate Mathematics. Specially,
to the students that are interested in this exquisite and conducive eld of
Mathematics.
This text is divided into two basic parts. The rst part, which cor-
responds to the rst eight chapters, consists of an exposition of classical
results in Geometric Theory of (real) Foliations. Special attention is paid
to the classical Reeb Stability theorems, Haeigers theorem and Novikovs
compact leaf theorem.
Starting at Chapter 9, the second part contains a robust proof of
Plantes Theorem on growth and compact leaves. This is followed by the
basic ingredients of the theory of foliation cycles and currents which is de-
veloped in Chapter 10. Then in Chapter 11 this is applied in D. Sullivans
homological proof of Novikovs compact leaf theorem. It is based in a mix
of topological argumentation and invariant measure theory for foliations.
Chapter A is dedicated to some more specialized results concerning the
structure of codimension one foliations on closed manifolds. We present
the results of Dippolito on the structure of codimension one foliations and
semi-stability. Also we present Catwell-Conlons result on the minimal sets
for such foliations.
We present in Chapter 11, D. Sullivans homological proof of Novikovs
compact leat theorem. This appears to be an applicable procedure for
complex foliations. We invite the reader to think about it.
Preface ix
In the last part of the book we give an exposition of some important

results on the structure of codimension one foliations. We state results of
Dippolito and Cantwell-Conlon on the their structure.
We hope the reader will enjoy reading this book as much as we have
enjoyed writing it.
Carlos Morales and Bruno Scardua


Contents
Preface vii
1. Preliminaries 1
1.1 Denition of foliation . . . . . . . . . . . . . . . . . . . . . 1
1.2 Examples of foliations . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Foliations derived from submersions . . . . . . . . 4
1.2.2 Reeb foliations . . . . . . . . . . . . . . . . . . . . 9
1.2.3 Lie group actions . . . . . . . . . . . . . . . . . . 13
1.2.4 Rn actions . . . . . . . . . . . . . . . . . . . . . . 17
1.2.5 Turbulization . . . . . . . . . . . . . . . . . . . . . 18
1.2.6 Suspensions . . . . . . . . . . . . . . . . . . . . . . 20
1.2.7 Foliations transverse to the bers of a ber bundle 23
1.2.8 Transversely homogeneous foliations . . . . . . . . 27
1.2.9 Fibrations and the theorem of Ehresmann . . . . . 30
1.3 Holomorphic Foliations . . . . . . . . . . . . . . . . . . . . 33
1.3.1 Holomorphic foliations with singularities . . . . . 33
2. Plane elds and foliations 39

2.1 Denition, examples and integrability . . . . . . . . . . . . 39
2.1.1 Frobenius Theorem . . . . . . . . . . . . . . . . . 39
2.2 Orientability . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3 Orientability of singular foliations . . . . . . . . . . . . . . 46
2.4 Orientable double cover . . . . . . . . . . . . . . . . . . . 47
2.5 Foliations and dierentiable forms . . . . . . . . . . . . . 50
xi
xii Geometry, Dynamics and Topology of Foliations - a first course
3. Topology of the leaves 53
3.1 Space of leaves . . . . . . . . . . . . . . . . . . . . . . . . 53

3.2 Minimal sets . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4. Holonomy and stability 59

4.1 Denition and examples . . . . . . . . . . . . . . . . . . . 59
4.2 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Reeb stability theorems . . . . . . . . . . . . . . . . . . . 69
4.4 Thurston stability theorem . . . . . . . . . . . . . . . . . . 74
5. Haeigers theorem 77
5.1 Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Morse theory and foliations . . . . . . . . . . . . . . . . . 78
5.3 Vector elds on the two-disc . . . . . . . . . . . . . . . . . 82
5.4 Proof of Haeigers theorem . . . . . . . . . . . . . . . . . 85
6. Novikovs compact leaf 87

6.1 Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2 Proof of Auxiliary theorem I . . . . . . . . . . . . . . . . . 88
6.3 Proof of Auxiliary theorem II . . . . . . . . . . . . . . . . 90
6.4 Some corollaries of the Novikovs compact leaf theorem . . 100
7. Rank of 3-manifolds 103
8. Tischlers theorem 107

8.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.2 Proof of Tischlers theorem and generalizations . . . . . . 108
9. Plantes compact leaf theorem 121

9.1 Growth of foliations and existence of compact leaves . . . 121
9.1.1 Growth of Riemannian manifolds . . . . . . . . . 121
9.1.2 Growth of leaves . . . . . . . . . . . . . . . . . . . 122
9.1.3 Growth of orbits . . . . . . . . . . . . . . . . . . . 122
9.1.4 Combinatorial growth of leaves . . . . . . . . . . . 123
9.1.5 Growth of groups . . . . . . . . . . . . . . . . . . 124
9.2 Holonomy invariant measures . . . . . . . . . . . . . . . . 125
9.3 Plantes theorem . . . . . . . . . . . . . . . . . . . . . . . 129
Contents xiii
10. Currents, distributions, foliation cycles and transverse measures 135

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 135
10.2 Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
10.2.1 Examples . . . . . . . . . . . . . . . . . . . . . . . 136
10.3 Invariant measures . . . . . . . . . . . . . . . . . . . . . . 139
10.3.1 Examples . . . . . . . . . . . . . . . . . . . . . . . 142
10.4 Currents and transverse measures . . . . . . . . . . . . . . 146
10.4.1 Examples . . . . . . . . . . . . . . . . . . . . . . . 148
10.5 Cone structures in manifolds . . . . . . . . . . . . . . . . . 151
11. Foliation cycles: A homological proof of Novikovs com-

pact leaf theorem 153
11.0.1 Examples . . . . . . . . . . . . . . . . . . . . . . . 154
11.0.2 Homological proof of Novikovs compact
leaf theorem . . . . . . . . . . . . . . . . . . . . . 156
Appendix A Structure of codimension one foliations: Dip-

politos theory 161
A.1 Semi-proper leaves, Dippolitos semi-stability . . . . . . . 161
A.2 Completion of an invariant open set . . . . . . . . . . . . 163
A.2.1 Completion of an invariant open set - revisited . . 165
A.3 Proof of Dippolitos semi-stability theorem . . . . . . . . . 168
A.4 Guided exercises: Cantwell-Conlons theory . . . . . . . . 168
Bibliography 173
Index 177


Chapter 1
Preliminaries
1.1 Denition of foliation
There are essentially three ways to dene foliations. Let M be a m-

dimensional manifold, m N. Let Dk be the open unit ball of Rk where
k N. Let 0 n m be xed.
Denition 1.1. A C r foliation of codimension m n of M will be a max-

imal atlas F = {(Ui , Xj )}iI of M satisfying the following properties:
(1) Xi (Ui ) = Dn Dmn ;

(2) For all i, j I the map Xj (Xi )1 : Xi (Ui Uj ) Xj (Ui Uj ) is C r
and has the form
Xj (Xi )1 (x, y) = (fi,j (x, y), gi,j (y)).
The number n is called the dimension of F . A plaque of F is a set

= Xi1 ({y = C}) for some C Rmn . The plaques of F dene a relation
in M as follows: If x, y M then x y if and only if there is a nite
collection of plaques 1 , , k such that x 1 , y k and i i+1 =
for all 1 i k 1. Clearly is an equivalence and then we can consider
the equivalence class Fx of containing x M . A leaf of F is dened as
an equivalence class L = Fx of (for some x M ). One can easily prove
that every leaf of F is an immersed submanifold of M . We shall see later
that a leaf may self-accumulate, and so, the leaves of F are not embedded
in general. Under the viewpoint of the equivalence , one can dene F as
a partition of M by immersed submanifolds L such that for all x M there
is a neighborhood U dieomorphic to Dmn Dn such that the leaves of
the partition intersect U in the trivial foliation {Dn y : y Dmn } on
1
2 Geometry, Dynamics and Topology of Foliations - a first course
Dmn Dn . Thus, we have the following equivalent denition of foliation.
Denition 1.2. A C r foliation of codimension m n of M is a partition

F of M consisting of disjoint immersed C r submanifolds F M with the
following property: for each point x M there is a neighborhood U of x,
and a C r dieomorphism X : U Dn Dmn , such that y Dmn
F F satisfying
X 1 (Dn y) F.
The elements of the partition F are called the leaves of F . The element Fx
of F containing x M is called the leaf of F containing x.
Warning: Not every decomposition of M into immersed submanifolds

with the same dimension is a foliation. Indeed, consider the partition of
R2 depicted in Figure 1.1 (note that the condition for foliation fails at the
point x).
Fig. 1.1
The third denition of foliation uses the notion of distinguished appli-

cations. Let F = {(Ui , Xi )} be a foliation of a manifold M in the sense of
Preliminaries 3
Denition 1.1. Then i, j the transition map Xj (Xi )1 has the form
Xj (Xi )1 (x, y) = (fi,j (x, y), gi,j (y)).
The map gi,j is a dieomorphism in its domain of denition. This
follows from the fact that the derivative D(Xj (Xi )1 )(x, y) has non-

zero determinant equals to x fi,j (x, y) gi,j (y). We dene for all i the
map gi = 2 Xi , where 2 is the projection onto the second coordinate
(x, y) Dn Dmn y. One has gj = gi,j gi as (2 Xj ) Xi1 = gi,j
1
and then gi,j gj = Xi 2 gi,j gj = 2 Xi = gi gj = gi,j gi since
2 is the identity in Dmn . Therefore, a C r foliation F of codimension
m n of a manifold M m is equipped with a cover {Ui } of M and C r
submersions gi : Ui Dmn such that for all i, j there is a dieomorphism
gi,j : Dmn Dmn satisfying the cocycle relations
gj = gi,j gi , gi,i = Id .
The gi s are the distinguished maps of F .

Conversely, suppose that M m admits an open cover M = Ui such
iI
that for each i I there is a C r submersion gi : Ui Dmn such that for
all i, j there is a dieomorphism gi,j satisfying the cocycle relations above.
By the Local form of the submersions we can assume that for each i I
there is a C r dieomorphism Xi : Ui Dn Dmn such that
gi = 2 Xi
since
2 Xj (Xi )1 ) = gj (Xi )1 = gi,j gi (Xi )1 = gi,j 2 ,
we have that the atlas
F = {(Ui , Xi )}
denes a foliation of class C r and codimension m n of M . The above
suggests the following equivalent denition of foliation.
Denition 1.3. A C r foliation of codimension m n of M is a cover {Ui :

i I} of M such that i I there is a C r submersion gi : Ui Dmn such
that i, j I there is a dieomorphism gi,j : Dmn Dmn satisfying the
cocycle relations
gj = gi,j gi , gi,i = Id .
The gi s are the distinguished applications of F.
This last denition leads to several interesting denitions. For instance,

a foliation F of M is said to be transversely orientable or transversely
ane depending on whether, for some convenient choice, its distinguished
applications gi,j are orientation preserving or ane maps. An equivalent
denition will be given in Section 2.2. In order to distinguish foliations, we
shall use the following denition.
Denition 1.4. Two foliations F, F dened on M, M respectively are
C r -equivalent if there is a C r -dieomorphism h : M M (h is a homeo-
morphism if r = 0), sending leaves of F into leaves of F . In other words,

h(Fx ) = Fh(x) , x M.
The above relation denes an equivalence in the space of foliations.

As an illustration, observe that the foliations F1 , F2 in the band I R in
Figure 1.2 are not equivalent.
F1 F2
Fig. 1.2 Non equivalent foliations.
1.2 Examples of foliations
1.2.1 Foliations derived from submersions

A submersion between two manifolds is a smooth map whose derivative has
maximal rank everywhere. Submersions provide the very rst examples of
Preliminaries 5
foliations.
Theorem 1.1. Let f : M m N n be a C r submersion. Then the connected

components of the level submanifolds
Lc = f 1 (c), cN
r
are the leaves of a C foliation of codimension n of M .
Proof. By the Local Form of the Submersions [do Carmo (1992)] there are
atlases {(U, X)}, {(V, Y )} of M, N respectively such that
(1) X(U ) = D n Dmn .
(2) Y (V ) = Dmn .
(3) Y f X 1 = 2 (see Figure 1.3).
V
f
U
X Y
-1
Y f X
m-n
D
Dm-n
Dn
Fig. 1.3 Foliation and submersion.
We claim that the collection F = {(U, X)} denes a foliation of M . In

fact, let (U, X), (U , X ) be two elements of the cover. Then,
2 X X 1 = Y f (X )1 X X 1
= Y f X 1 = Y Y 1 Y f X 1 = Y Y 1 2 .
Hence 2 (X X 1 ) = (Y Y 1 ) 2 does not depend on x Dn . This

proves that F is a foliation of class C r and codimension m n of M . It is
clear by denition that the plaques of F are contained in the level sets of
f . This proves that the leaves of F are precisely the level sets of f and the
result follows.
Let us present some examples to illustrate the above result.
Example 1.1. Let M and N be C r manifolds and take f : M N M as

the rst coordinate projection f (x, y) = x. Clearly f is a C r -submersion.
In this case f denes a C r -foliation of M N whose leaves are the vertical
bers in {x} N, x M .
Exercise 1.2.1. Let F be a foliation on M of codimension q. A dierentiable

map f : N M is transverse to F if it is transverse to each leaf L F as
an immersed submanifold in M . Show that in this case there is a naturally
dened foliation f (F ) in N of codimension q such that for each leaf L F
the inverse image f 1 (L) is a union of leaves of f (F ).
Exercise 1.2.2 (Double of a foliation). Let F be a smooth foliation on

M . Suppose that we have a relatively compact domain D M with smooth
boundary D transverse to F . Consider
the manifold with boundary M0 =
M \ D and the restriction F0 = F M0 . Given two copies M1 and M2 of
M0 we can construct a manifold Md by gluing these copies by the common

boundary D and equip it with a smooth foliation Fd such that Fd Mj is
naturally conjugate to F0 .
Example 1.2. Let M = R2 and f (x, y) = y x, where R. The

level curves of f dene a foliation F in M whose leaves are the straight-
lines y = x + c, c !R. Observe that F is invariant by the translations
(x, y) (x + k, y + l), (k, l) Z2 . Indeed, if y = x + c then y + l =
x + c + l = (x + k) + c , where c = c k proving the invariance.
It follows that F projects into a foliation of the 2-torus T 2 = R2 /Z2 still
denoted by F . See Figure 1.4. When is irrational then all the leaves of
the induced foliation are lines, and if is rational then all the leaves are
circles. We call this example as the linear foliation in T 2 .
Example 1.3 (Reeb foliation on S 3 ). Let M = R3 and f (x, y, z) =

(r2 )ez , where r2 = x2 + y 2 and is a C function such that (0) =
1, (1) = 0 and (t) < 0 for all t > 0 (see Figure 1.5).
Preliminaries 7
IR 2
T2
Fig. 1.4 Linear foliation on T 2 .
0 1
Fig. 1.5 Graph of .
The map f is a submersion since

f (x, y, z) = (2 (r2 )xez , 2 (r2 )yez , (r2 )ez ) = (0, 0, 0)
x = y = 0 and (r2 ) = 0 x = y = 0, x2 + y 2 = 1 contradiction.
Hence f (x, y, z) does not vanish and so f is a submersion. It follows from
Theorem 1.1 that the level curves f 1 (c) dene a foliation of class C and
codimension 1 of M . The leaves of this foliation, (i.e., the level curves of

f ) can be described as follows.
f (x, y, z) = c (r2 )ez = c.
If c = 0 then (r2 ) = 0 x2 + y 2 = 1. Hence the level curve corresponding

to c = 0 is the cylinder x2 , y 2 = 1 in M . If c > 0 then
(r2 )ez = c (r2 ) > 0.
Moreover,
z = K ln((r2 )),
(K = ln(c)). When c = 1 we have
z = ln((r2 )).
The graph of the above curve in the plane y = 0 is given by
z = ln((x2 )).
We have
2(x2 ))
z = x = 0 x = 0.
(x2 ))
Hence x = 0 is the sole critical point of z. We have that z as
x x 1+ or 1 . The graph of z is a parabola-like curve. The graph of
the leaves of F is depicted in Figure 1.6.
Example 1.4 (Fibrations). Let E, B, F be smooth manifolds. We say

that E is a ber bundle of class C r over B with ber F if there is a C r onto
submersion : E B satisfying the following properties:
(1) 1 (b) is dieomorphic to F , b B.

(2) For each point b B there exists a neighborhood U B of b and a
dieomorphism : 1 U F such that 1 = , where 1 is the
projection onto the rst coordinate in U F .
Clearly the family { 1 (b) : b B} is a C codimension dim(B) folia-

tion of E since is a C submersion. Note that the leaves of the resulting
foliation are all dieomorphic to a common manifold F .
Preliminaries 9
c>0
c=0 c<0
Fig. 1.6
1.2.2 Reeb foliations
Several are the classes of foliations called Reeb foliations. The very rst ones
are the Reeb foliations in the cylinder and in the Moebius band, constructed
as follows. Dene M = [1, 1]R and let F be the foliation in M dened by
the submersion g(x, y) = (x2 )ey where is decreasing. Let G : M M be
given by G(x, y) = (x, y + 2). The quotient manifold M/G is the cylinder.
Analogously we can replace G by the map F (x, y) = (x, y + 2). In this
case the quotient manifold M/F is the Moebius band. In each case one
can see that F is invariant for G and F . Hence F induces a foliation F in
either M/G or M/F . These are the Reeb foliations in the cylinder and the
Moebius band respectively. These foliations are depicted in Figure 1.7.
Reeb foliation in the cylinder Reeb foliation in the Moebious band
Fig. 1.7 Reeb foliation in the cylinder and Moebius band.
Consider the foliation F constructed in Example 1.3 of Section 1.2

restricted to the solid cylinder {(x, y, z) : x2 + y 2 1}. One can eas-
ily check that the leaves of this foliation are invariant by the transla-
tions (x, y, z) (x, y, z + 1). Note that the quotient manifold solid
cylinder/(x, y, z) (x, y, z + 1) is a solid torus D2 S 1 . The invariance
mentioned above implies that F induces a foliation in D2 S 1 whose leaves
are depicted in Figure 1.8. This foliation is called the Reeb foliation in the
solid torus ST = D2 S 1 . The Reeb foliation in the solid torus is used to
construct a C foliation in the 3-sphere S 3 in the following way:
Let ST1 and ST2 be two solid tori and denote ST1 = T1 and ST2 = T2
the corresponding boundaries. Consider a dieomorphism : T2 T1
sending the meridian curves in T2 into the parallel curves in T1 . For instance
we can choose by rst considering

01 x
(x, y) = .
10 y
Because is linear and det = 1 we have (Z2 ) = Z2 and then denes

the desired map.
In ST1 ST2 we consider the equivalence relation given by y = (x).
In other words we use the identication below.
Preliminaries 11
Fig. 1.8 Reeb foliation in the solid torus.

x, y Int ST2 and x = y, or

x y x Int ST1 , y Int ST1 and x = y, or (1.1)

x T , . . . , y T and (y) = x
1 2
Consider the quotient manifold M = (ST1 ST2 )/ = ST1 ST2 .
Claim 1.1. M = S 3 .
In the proof of this claim we use:

Alexanders trick ([Hansen (1989)]): Let B1 , B2 two closed 3-balls, S1 =
B1 and S2 = B2 be the corresponding 2-sphere boundaries. If : S1
S2 is a dieomorphism then B1 B2 = S 3 .
Now we return to the proof of the claim. Take a region P in between

two meridians of ST1 . Delete P and cap it into the torus hole in ST 2
as explained in Figure 1.9. With this procedure we obtain two 3-balls
whose union along the corresponding boundaries yields S 3 by the Alexander
trick. This proves the claim. The Reeb foliation in S 3 is precisely the one
obtained by the gluing map setting inside each solid torus the Reed
foliation of the solid torus (see Figure 1.10). The above construction leads
to construct foliations with only one compact leaf in any manifold of the
form ST1 ST2 . For instance if were the identity map then the resulting
manifold is S 2 S 1 . We have then constructed a foliation with just one
ST1
ST2
Fig. 1.9
compact leaf in S 2 S 1 . The resulting foliation is clearly dierent to the

one obtained by the trivial bration {S 2 y : y S 1 } of S 2 S 1 .
Exercise 1.2.3. Show that the Reeb foliation in S 3 cannot be obtained from
a submersion.
Exercise 1.2.4 (Novikov). Show that a vector eld transverse to the

Reeb foliation in S 3 has a periodic orbit. Observe that there are non-
singular C vector elds in S 3 without periodic orbits (these vector elds
are precisely the counterexamples for the Seifert Conjecture [Seifert (1950);
Schweitzer (1974)]). Use Novikovs compact leaf theorem (see Chapter 6)
to show that any vector eld transverse to a codimension one C 2 foliation
in a compact manifold with nite fundamental group has a periodic orbit.
Denition 1.5 (Reeb component). A Reeb component of a codimen-

sion one foliation F in M 3 is a solid torus ST M 3 which is union of
Preliminaries 13
leaves of F such that F restricted to ST is equivalent to the Reeb foliation

in the solid torus D2 S 1 . A foliation is said to be Reebless if it has no Reeb
components. Novikovs compact leaf theorem in Chapter 6 states that a
compact 3-manifold supporting a Reebless foliation must have an innite
fundamental group.
Fig. 1.10 Reeb foliation in S 3 .
1.2.3 Lie group actions

A Lie group is a group (G, ) with a dierentiable structure making the
maps
GG G G G
and
x, y x y x x1
dierentiable.
Example 1.5. (Rn , +) is a Lie group, therefore S n \ {point} is a Lie group

via stereographic projection. If C denotes the set of complex numbers
(with the complex product), then C = C {0} with the complex number
product is a Lie group. The circle S 1 C is a Lie group when equipped
with the product induced by C. Actually it is a Lie subgroup of C . The
n-torus T n = S 1 S 1 with the product (z1 , . . . , zm )(z11 , . . . , zm
1
) =

n copies
(z1 z11 , . . . , zn zm
1
), zi , zi1 S 1 is a compact Lie group.1 The linear
group GLn (R) = {A Mnn (R), det A = 0} with the usual matrix product
is a Lie group. The subgroup O(n) GLn (R) of orthonormal matrices is
a compact Lie subgroup of GLn (R).
Denition 1.6. An action of a Lie group G in M is a map : GM M

satisfying the following properties:
(1) (e, x) = x for all x M ;
(2) (g h, x) = (g, (h, x)) for all x M , g, h G.
We write (g, x) = g x. With this notation, the orbit of x M is the set

Ox = {g x : g G}.
The isotropy group of x M is the subgroup Gx G consisting of the
elements g G xing x, namely
Gx = {g G : g x = x}.
Denition 1.7. An action : G M M is locally free if the isotropy

group Gx is discrete, x M . This is equivalent to say that the map
x : g G g x,
is an immersion for all x M xed.
The action G M M is locally free if, and only if, the orbit Ox of
x is an immersed submanifold of M with constant dimension dim(Ox ) =
dim(G). Since Gx < G is a closed subgroup it is itself a Lie group (Cartans
Theorem) and also the quotient G/Gx has the structure of a dierentiable
manifold. Actually we have Gx = Gy x, y belonging to some orbit of p
and we may introduce the isotropy subgroup of an orbit as well.
Given any x M we have a natural (dieomorphism) identication
G/Gx = Ox . This gives an immersed submanifold structure Ox M .
Theorem 1.2. The orbits of a C r locally free action of a Lie group G on a

manifold M m are the leaves of a C r foliation of codimension m dim(G)
of M .
Proof. Let : G M m M m be a locally free C r action. Fix a point

z0 M and set n = dim(G). By assumption dim(Oz0 ) = n.
1 The product of Lie groups is a Lie group with a natural product structure ([Hirsch
(1971)]).
Preliminaries 15
v
Oz
z 0
0
e
m-n
0
vm-n
m-n
vn
D
n
D
Fig. 1.11 Locally free action.
Let mn
0 be a m n dimensional submanifold transverse to Oz0 at z0 .
Let (P, U ) be a local chart at the unity e G such that P (U ) = Dn . Let
B : (Dmn , 0) (mn
O , z0 ) be a parametrization of mn
0 . Recalling the
notation (g, x) = g x we dene the map A : D Dn mn
M by
A(x, y) = P 1 (x) B(y).
The derivative DA(x, y) is given by the expression below:
g (P 1 (x), B(y)) DP 1 (x) + z (P 1 (x), B(y)) DB(y).
Replacing by (x, y) = (0, 0) one has
DA(0, 0) = Dz0 (e) DP 1 (0) + z (e, z0 ) B(0)
= Dz0 (e) DP 1 (0) + DB(0).

Let us write v = vn vmn for a tangent vector v of the product Dn Dmn

at (0, 0). Hence
DA(0, 0) v = Dz0 (e) DP 1 (0) vn + DB(0) vmn .
Hence DA(0, 0) = 0 if v = 0 for the vectors vn = Dz0 (e) DP 1 (0) and

vmn = DB(0) vmn are linearly independent in Tz0 M . We conclude that
A is a local dieomorphism. By the Inverse Function theorem the inverse
X = A1 of A is well dened in a neighborhood U of z0 . This denes an
atlas
F = {(X, U )}
of M . Note that for all xed y0 Dmn one has
A({(x, y0 ) : x Dn }) = {P 1 (x) B(y0 ) : x Dn }.
Hence A({(x, y0 ) : x Dn }) is contained in the orbit OB(y0 ) . This proves

that the plaques of F are contained in the orbits of the action. Since the
orbits are pairwise disjoint we conclude that F is a foliation of M .
In a similar way one can prove the following.
Theorem 1.3. The orbits Ox of a C r action on a manifold M are the leaves

of a C r foliation if and only if the map M x dim(Ox ) is constant. In
this case the action is called a foliated action on M .
Example 1.6. If G is a Lie group and H is a Lie subgroup of G then there

is a natural action H G G by left multiplication. This action is foliated
since the left translation is a dieomorphism and so the orbits of the action
dene a foliation of G. Note that the action is locally free if, and only if,
H is discrete.
Example 1.7. A C r ow on a manifold M is an action of the additive

Lie group R in M . Denote by X the corresponding tangent vector eld

X(x) := t t=0
(t, x). Then X is non-singular if and only if is locally
free. The orbits of are either circles or lines. In the rst case the orbit
is called periodic and the ow is periodic if all its orbits are periodic. A
manifold is called Seifert if it admits periodic ows.
Exercise 1.2.5. Let G < R2 be a discrete subgroup. Show that G is isomor-

phic to Z or to Z2 . Conclude that the orbits of a locally free action of the
ane group A(R) are either planes or cylinders.
Preliminaries 17
Theorem 1.4 (Hopf s theorem, [Hopf (1931)]). The three-sphere S 3

is a Seifert manifold.
Indeed, note that S 3 = {(z1 , z2 ) R2 R2 : ||z1 ||2 + ||z2 ||2 = 1}. Dene
Q : S 1 S 3 S 3 by Q(x, (z1 , z2 )) = (x z1 , x z2 ). One sees that Q is
an action of S 1 in S 3 . The orbits of Q dene a foliation by circles in S 3
proving that S 3 is Seifert. The resulting ow is called Hopf bration of S 3
(see [Lyons (2003)] for more details). Seifert manifolds are important in
3-manifold topology since the ones with innite fundamental group can be
described by their fundamental groups. This fact was discovered by Scott
(see [Scott (1983)] for a very complete text on the subject). The fundamen-
tal group classication of Seifert manifolds with nite fundamental group
is false because there are homotopy equivalent lens spaces which are not
homeomorphic.
1.2.4 Rn actions
We shall describe the actions of the additive group Rn on a manifold M .
Let Q : Rn M M an action and {e1 , . . . , en } be a basis of Rn . Fix
i = 1, . . . , n and consider a map
Xi
R M M
(t, x) Q(tei , x)
This map denes an action of R on M . In fact X i (t, x) = Q(t ei , x) is

a ow in M . We still denote by X i the vector eld induced by X i , namely

d d
X i (Xti (x)) = (Xti (x)) and X i (x) = Q(t ei , x).
dt dt t=0
In this way we have n vector elds X 1 , . . . , X n in M such that Xti (x) =
Q(t ei , x), t R.

n
Now, let v = ti ei Rn and x M be xed. Then
i=1

n n1

Q(v, x) = Q ti e i , x = Q ti ei , Q(tn em , x)
n=1 i=1
n1

=Q ti ei , Xtnn (x) = Xt11 Xt22 Xtnn (x), x M.
i=1
Hence, for all action Q : Rn M M , there are vector elds

X , . . . , X m on M such that Q(v, x) = Xt11 (v) Xtnn (v) (x), x M ,
1
v Rn .
Note that X 1 , . . . , X n pairwise commute, namely Xti Xsj = Xxj Xti . In
fact, we have (Xti Xsj )(x) = Q(tei , Xsj (x)) = Q(t ei , Q(s ej , x)) = Q(t ei +
s ej , x) = Q(s ej + t ei , x) = Xsj Xti (x).
It is known that X, Y commute [X, Y ] = 0.
Conversely, if X 1 , . . . , X n are pairwise commuting vector elds, then
they dene an action Q given by
Q(v, x) := Xt1 (v) (x) Xtn (v) (x).
Denition 1.8. The rank of a closed manifold M is the maximal number

of linearly independent pairwise commuting vector elds dened in M .
Clearly M has rank 1. Equivalently, the rank of M is the maximal n

such that Rn acts freely in M .
Theorem 1.5 (The rank theorem, [Lima (1965)]). The 3-sphere S 3

and the product S 2 S 1 are rank 1 manifolds. The rank of the 3-torus
T 3 is 3.
We shall return to this result in Chapter 7 (see Theorem 7.1).
1.2.5 Turbulization
Let F a codimension one foliation on a 3-manifold and be a closed curve
such that is transverse to F . we assume that is orientable, (i.e., it has
a solid torus tubular neighborhood). We modify F along as follows. Pick
a neighborhood U of and suppose that U is dieomorphic to a solid torus
S 1 D2 . Since is transverse to F we can assume that F intersects the
solid torus in the trivial foliation by discs D2 , S 1 .
We consider the Reeb foliation FR in S 1 D2 . We replace (by surgery)
the foliation F U by FR in U to obtain a foliation F as in Figure 1.12.
The resulting foliation F is said to be obtained by turbulization of F
along . Note that F is C r if F is, 0 r , r = w. The number
of Reeb components of the new foliation F is greater than or equal to the
number of components of F .
Example 1.8. Consider the foliation by discs D2 in S 1 D2 , and let

= S 1 {0} be the curve in the middle of S 1 D2 . We have that F ,
Preliminaries 19
Initial foliation F Modified foliation F
Fig. 1.12 Turbulization.
and so, we can modify F by turbulization along . The compact leaves of

the resulting foliation are dieomorphic to the torus T 2 . The non-compact
ones are all either planes R2 (inside of the Reeb component) or punctured
discs D2 \{point} (outside of the Reeb component).
Example 1.9. Consider F as before and as in Figure 1.13.
Fig. 1.13
Note that F and so we can modify F by turbulization along .

In this case, F is a foliation of S 1 D2 whose leaves are T 2 , R2 and
D2 {2 points}. Analogously, it is easy to construct a foliation of S 1 D2
whose leaves are T 2 , R2 and D2 {n points}.
Exercise 1.2.6. Show that the for the Reeb foliation F in S 3 and for every
curve S 3 transverse to F we have the same number of Reeb components
for F and F .
Turbulization can be used to prove the following (see [Thurston (1975)]):
Theorem 1.6. All closed 3-manifolds support C codimension one folia-

tions. On the contrary, the only closed surfaces admitting codimension one
foliations are the torus T 2 and the Klein bottle K 2 .
1.2.6 Suspensions
A representation of a group G in a group H is a homomorphism
Q : G H.
We shall be interested in the case G = 1 (B) and H = Di r (F ) where B, F
are manifolds and Di r (F ) is the group of class C r dieomorphisms in F
endowed with the composition operation.
Suppose that Q : 1 (B) Di r (F ) is a representation of 1 (B) in
Di r (F ). Let B

B be the universal covering of B. Recall that 1 (B)
acts in B by deck transformations: 1 (B), b B, b = (b), we
With
have the lift of . Dene b = (1) as the action of 1 (B) in B.
this action one has B/ 1 (B) B. 1 (B) also acts in B F via Q in the
following way: dene A : 1 (B) (B F) B F by setting
A(, (b, x)) = ( b, Q()(x)).

F )/A B Q F is a manifold.
B Q F = (B
Denition 1.9. The orbit space of A,

F )/A
B Q F = (B
is called the suspension of Q.
Example 1.10. Suppose that Q(g) = IdF (the identity in F ) for all g
1 (B). Then B Q F is precisely the cartesian product B F .
Example 1.11. Suppose B = S 1 . In this case B = R and 1 (B) = Z.

Let Q : 1 (B) Di r (F ) be a representation. Then Q(n) = (f 1 )n =
f 1 f 1 f 1 , with f : F F being a dieomorphism of F . By
denition A(n, (b, x)) = (b + n, f n (x)). Note that A identies (0, x) with
Preliminaries 21
(0 + 1, f 1 (x)) = (1, f 1 (x)) by replacing x by f (x), we have that A iden-

ties (0, f (x)) with (1, x).
Note that S 1 Q F exhibits a ow given by projecting the constant ow

onto R F . See Figure 1.14.
t
The suspension B Q F is equipped with two foliations FQ , FQ de-
ned as follows: The action A leaves invariant the horizontal and vertical
foliations in B F given by F = {B f : f F }, F = {b F : b B}.

Hence the action A induces a pair of foliations FQ , FQ in B Q F whose
leaves L, L satisfy
L = (leaf of F), L = (leaf of F ),

where : B F B Q F is the quotient map. Note that the foliations

F Q , FQ in B F are transverse. We shall discuss more properties of these
foliations later on.
~
FQ
~
Bx F

F~
Q FQ
F
Q
Bx QF
Fig. 1.14
Example 1.12. Let B be the bitorus, i.e., the genus two orientable closed
surface. The fundamental group 1 (B) has the following presentation:
1 (B) =< a, b, c, d : aba1 b1 cdc1 d1 = 1 > .
Fix f, g Di r (S 1 ) and dene the presentation Q : 1 (B) Di r (S 1 ) by

setting
Q(a) = f, Q(c) = g, Q(b) = Q(d) = Id
and extending linearly. The projection Q is well dened since
Q(aba1 b1 cdc1 d1 ) = 1.
Let us describe the suspension B Q S 1 of Q. On one hand, consider the
subgroup G of 1 (B) generated by b and d, i.e., G =< b, d >. On the
other hand observe that the universal covering of B, B, is the Poincare
disc. Let AG : G B S 1 B S 1 the action A restricted to G, namely
AG (x, ) = (g x, Q(g)(). Clearly Q = Id in G and so
(B S 1 )/AG = B/G S 1 .
Consider S 1 as the unit interval [0, 1] with 0 1. Figure 1.15 describes the
orbit space (B S 1 )/AG of AG .
~ x1
B/ G
~
Bx 1
b -1
a a
-1
-1 b
d
~ ~ x0
B x0 B/ G
c -1 c
Fig. 1.15
The internal surface B/G 0 in the gure is identied with the external
one B/G 1. To obtain B Q S 1 we identify the intermediate curves
a g(), a1 , c f (), c1 according to Figure 1.16.
The leaves of the resulting foliation FQ spiral around the suspended

manifold according to the maps f, g. The other foliation FQ yields a folia-
tion by circles of B Q S , and so, B Q S is Seifert. We shall be back to
1 1
this example later on.

Preliminaries 23
c x f ( ) ax g( )
c-1 x a-1 x
Fig. 1.16
1.2.7 Foliations transverse to the fibers of a fiber bundle
In this section we discuss an important class of foliations given by suspen-

sions, the class of foliations transverse to the bers of a ber bundle. Let
us rst recall some basic denitions:
Example 1.13 (Fibre bundle). A (dierentiable) ber bundle over a

manifold M is given by a dierentiable map : E M from a manifold
E, called total space, which is (the map) a submersion having the follow-
ing local triviality property: for any p M there exist a neighborhood
p U M and a dieomorphism U : 1 (U ) E U F , where F

is xed manifold called typical ber of the bundle, such that the following
diagram commutes
U
1 (U ) U F
1
U
where 1 : U F U is the rst coordinate projection 1 (x, f ) = x. In

other words U is of the form U (x) = ((x), . . . ). Such a dieomorphism
U is called a local trivialization of the bundle and U is a distinguished
neighborhood of p M . Given p M the ber over p is 1 (p) E and by
the local trivialization each ber is an embedded submanifold dieomorphic
to F .
According to the theorem of Ehresmann (Theorem 1.10) any C 2 proper

submersion denes a ber bundle as above. Let us motivate our next de-
nition with an example.
Example 1.14 (Suspension of a foliation by a group of diffeomor-

phisms). A well known way of constructing transversely homogeneous
foliations on bered spaces, having a prescribed holonomy group (see Chap-
ter 4 for the denition of holonomy) is the suspension of a foliation by a
group of dieomorphisms. This construction is briey described below: Let
G be a group of C r dieomorphisms of a dierentiable manifold N . We
can regard G as the image of a representation h : 1 (M ) Di r (N ) of the
fundamental group of a complex (connected) manifold M . Considering the
dierentiable universal covering of M , : M M we have a natural free

action 1 : 1 (M ) M M , i.e., 1 (M ) Di r (M ) in a natural way.

Using this we dene an action H : 1 (M ) M N M N in the natural
N
way: H = (1 , h). The quotient manifold MH = Mh is called the suspen-
sion manifold of the representation h. The group G appears as the global
holonomy of a natural foliation Fh on Mh (see [Camacho and Lins-Neto
(1985)]). We shall explain this construction in more details. Let M and N
be dierentiable manifolds of class C r . Denote by Di r (N ) the group of C r
dieomorphisms of N . Given a representation of he fundamental group of
M in Di r (N ), say h : 1 (M ) Di r (N ), we will construct a dierentiable
ber bundle Mh , with base M , ber N , and projection P : Mh M , and
a C r foliation Fh on Mh , such that the leaves of F are transverse to the
bers of P and if L is a leaf of F then P |L : L M is a covering map. We
will use the notation G = h(1 (M )) Aut(N ).
M be the C r universal covering of M . A covering auto-
Let : M
morphism of M is a dieomorphisms f of M that satises f = . If
we consider the natural representation g : 1 (M ) Aut(M ) then we know
that:
(a) g is injective. In particular g(1 (M )) is isomorphic to 1 (M ).
(b) g is properly discontinuous.
We can therefore dene an action H : 1 (M ) M N M N in a
natural way:
If 1 (M ), m M e n N then
H(, m, n) = (g()(m), h()(n)).

Using (b) it is not dicult to see that H is properly discontinuous.
N , whose
Thus, the orbits of H dene an equivalence relation in M
Preliminaries 25
corresponding quotient space is a dierentiable manifold of class C r .

N
M
Denition 1.10. The manifold H = Mh is called the suspension man-
ifold of the representation h.
Notice that Mh is a C r ber bundle with base M and ber N , whose
projection P : Mh M is dened by
P (O(m, n)) = (m)

where O(m, n) denotes the orbit of (m, n) by H.
Let us see how to construct the foliation Fh . Consider the product
foliation F of M
N whose leaves are of the form M {n}, n N . It is

not dicult to see that F is H-invariant and therefore it induces a foliation
of class C r and codimension q = dim(N ), Fh on Mh , whose leaves are of
where L
the form P (L), is a leaf of F.

Denition 1.11. Fh is called the suspension foliation of F by h.

The most remarkable properties of this construction are summarized
in the proposition below (see [Godbillon (1991)], [Camacho and Lins-Neto
(1985)]):
Proposition 1.1. Let Fh be the suspension foliation of a representation
h : 1 (M ) Di r (N ). Then:
(i) Fh is transverse to bers of P : Mh M . Moreover, each ber of P

cuts all the leaves of Fh .
(ii) The leaves of Fh correspond to the orbits of h in N in a 1-to-1 corre-
spondence.
(iii) 2 If L is a leaf of Fh corresponding to the orbit of a point p N , then
P |L : L M is a covering map (here L is equipped with its natural
intrinsic structure).
This implies that xed a point p M and its ber Np = P 1 (p), we
obtain by lifting of paths in 1 (M, p), to the leaves of Fh , a group
Gp Di r (Np ), which is conjugate to G.
(iv) There exists a collection {yi : Ui N }iI of submersions dened in
open subsets Ui of Mh such that

(a) Mh = Ui
iI
(b) Fh Ui is given by yi : Ui N .
2 Due to (iii) we call G the global holonomy of the suspension foliation Fh .
(c) if Ui Uj = then yi = fij yj for some fij G.

(d) if L is the leaf of Fh through the point q Np , then the holonomy group
of L is conjugate to the subgroup of germs at q of elements of the group
G = h(1 (M, p)) that x the point q.
Conditions (i) and (ii) above motivate the following denition:

F
Denition 1.12. Let : = ( : E B) be a ber bundle. A foliation F
F
on E is said to be transverse to the bration : E B if:
(1) F is transverse to each ber of
(2) dim F + dim F = dim E
(3) For each leaf L F the restriction |L : L B is a covering map.
In this case F is conjugate to the suspension of the global holonomy repre-
sentation : 1 (B) Aut(F ) of F . According to Ehresmann ([Camacho
and Lins-Neto (1985)]) conditions (1) and (2) imply (3) when the ber F
is compact.
Using the holonomy lifting paths given by condition (3) above we can
easily prove:
Theorem 1.7. Let F be a foliation of class C r transverse to the bers of a

F
ber bundle : = ( : E B). Then F is conjugate to the suspension of
a representation : 1 (B) Aut(F ), indeed the global holonomy of F is
naturally conjugate to the image (1 (B)). Conversely if F is the suspen-
sion of a representation : 1 (B) Aut(F ) for some base manifold B and
F
some ber manifold F then there is a ber bundle space : = ( : E B)
such that F is transverse to the bers of and the global holonomy of F is
conjugate to the image (1 (B)) < Aut(F ).
Recall that a discrete nitely generated group is always conjugate to the

fundamental group of a manifold. Thus, suspensions of group presentations
and foliations transverse to ber bundles are in natural bijection. As a
natural complement to the above results we have:
Theorem 1.8. Two representations : 1 (B) Aut(F ) and : 1 (B)

Aut(F ) are conjugate if, and only if, there is a bered dieomorphism
: E E (i.e, is the lift of a dieomorphism : B B such that
= for the projections : E B and : E B), with the
property that is a conjugacy between the suspension foliations in E and
E of and respectively.
Preliminaries 27
1.2.8 Transversely homogeneous foliations

Now we introduce an important class of foliations. Let G be a Lie group
and denote by G the Lie algebra of G. The Maurer-Cartan form over G is
the unique 1-form w : T G G satisfying:
(i) w(X) = X, X G
(ii) Lg w = w, g G; where Lg : G G is the left-translation x G
gx G, g G xed.
The 1-form w satises the Maurer-Cartan formula dw + 12 [w, w] = 0. In

fact, given X, Y G we have
dw(X, Y ) = X.w(Y ) Y.w(X) w([X, Y ]) = [X, Y ].
But
[w, w](X, Y ) = [w(X), w(Y )] [w(Y ), w(X)] = 2[X, Y ]
because X and Y belong to G and w(X) = X, X G.
Thus we have dw(X, Y ) + 12 [w, w](X, Y ) = 0, X, Y G which proves the
Maurer-Cartan formula.
k
Let now {X1 , . . . , Xn } be a basis of G. We have [Xi , Xj ] = cij Xk for
k
some constants ckij C, skew-symmetric in (i, j). The ckij s are the structure
constants of G in the basis {X1 , . . . , Xn }.
Let now {w1 , . . . , wn } be the dual basis to {X1 , . . . , Xn }, with wj left-
k
invariant. We have dwk = 12 cij wi wj and then it is easy to see that
i,j
w = wk Xk is the Maurer-Cartan form of G.
k
We recall the following theorem of Darboux and Lie:
Theorem 1.9 ([Godbillon (1991)] pag. 230). Let be a dierentiable

1-form on a manifold M taking values on the Lie algebra G of G. Suppose
satises the Maurer-Cartan formula d + 12 [, ] = 0. Then is locally the
pull-back of the Maurer-Cartan form of G by a dierentiable map. Moreover
the pull-back is globally dened if M is simply-connected; and two such local
maps coincide up to a left translation of G.
As an immediate corollary we have:
Corollary 1.1. Let 1 , . . . , n be linearly independent dierentiable 1-

k
forms on a manifold M . Assume that we have dk = 21 cij i j
i,j
where the ckij s are the structure constants of a Lie group G in the basis
{X1 , . . . , Xn }. Then, locally, there exist dierentiable maps : U M

G such that j = wj , j where {w1 , . . . , wn } is the dual (left-invariant)
basis of {X1 , . . . , Xn }. Moreover if M is simply-connected then we can take
U = M and if : U G, : U G are two such maps with U U =
and connected then we have = Lg for some left-translation Lg of G.
This way we may construct foliated actions of Lie groups on manifolds

by dening suitable integrable systems of 1-forms on the manifold. This
gives rise to the notion of transversely homogeneous foliations which is a
very important notion in the theory.
Denition 1.13 (Transversely homogeneous foliation). A foliation

F has a homogeneous transverse structure if there are a complex Lie group
G, a connected closed subgroup H < G such that F admits an atlas of
submersions yj : Uj M G/H satisfying yi = gij yj for some locally
constant map gij : Ui Uj G for each Ui Uj = . In other words, a
suitable atlas of submersions for F has transition maps given by left transla-
tions on G and submersions taking values on the homogeneous space G/H.
We shall say that F is transversely homogeneous of model G/H.
Example 1.14. Let F = G/H be a homogeneous space of a complex Lie

group G (H G is a closed Lie subgroup). Any homomorphism representa-
F )/ = M
tion : 1 (N ) Di(F ) gives rise to a foliation F on (N
which is transversely homogeneous of model G/H.
Example 1.15. G = A(Rn ) = GLn (R) Rn acts on Rn by (A, B), X

AX + B and the isotropy subgroup of 0 Rn is GLn (R) 0 = H so
that G/H
= Rn and then the transversely homogeneous foliations of type
n
A(R )/ GLn (R) are the transversely ane foliations.
Example 1.16. The real projective unimodular group G = PSL (2, R) acts
on RP (1) by

xu xz + u
, z
yv yz + v
and the isotropy subgroup of 0 R is naturally identied with H = A(R),

so that G/H = RP (1) and then the transversely homogeneous foliations of
type PSL(2, R)/ A(R) are the transversely projective foliations.
Preliminaries 29
Now we introduce the concept of development of a transversely homo-

geneous foliation which is a basic tool in the study of these foliations:
Proposition 1.2. Let F be a transversely homogeneous foliation of model

G/H on M . Then there exist a homomorphism h : 1 (M ) G, a transitive
covering space p : P M corresponding to the kernel H = Ker(h) 1 (M )
and a submersion : P G/H satisfying:
(i) is h-equivariant which means that ( x) = h(x) , x P ,
1 (M ).
(ii) The foliation p F coincides with the foliation dened by the sub-
mersion .
Such a construction is called a development of the foliation F (see [Godbil-

lon (1991)] page 209 for a detailed denition).
We will give an idea of the proof of the above Proposition 1.2 according to
[Godbillon (1991)]:
Let {yi : Ui G}iI be a homogeneous transverse structure for F in M .
Denote by fij the transformation fij : G/H G/H such that yi = fij yj
in Ui Uj = .
We can identify fij in a natural way as an element of G. Now let E be the
space obtained as the sum of the Ui G, i I. Denote by G1 the subgroup of
G generated by the fij s. Consider in E the equivalence relation identifying
(x, y) Ui G, where x Ui Uj , with (x, fij g) Uj G.
Denote by P the quotient space E/ . Then P is a principal ber bundle
p : P M having a discrete structural group G1 G, P being dened
by the cocycle (Ui , fij ). The transitive covering space p : P M has
G1 as group of automorphisms so that there is a natural homomorphism
h : 1 (M ) G1 G.
Now in each Ui G we can construct a holomorphic submersion i : Ui
G G/H by i (x, g) = g(yi (x)). The submersion : P G/H is con-
structed by gluing the submersions i . Finally we remark that if P is not
connected we can replace this space by one of its connected components.
Corollary 1.2. Let F be a non-singular transversely homogeneous folia-

tion on a simply-connected manifold M . Then F is given by a smooth
submersion f : M G/H.
Proof. This corollary is a straightforward consequence of the Darboux-

Lie theorem above but can also be proved by the use of Proposition 1.2:
In fact, if M is simply connected in Proposition 1.2 then we have H =

Ker(h) 1 (M ) = 0 so that H = 0 and then P = M . Thus Corollary 1.2
follows from ii) of this same proposition.
Remark 1.1.
(i) 1 (M ) acts over P in the following way: Given x P we dene x
as the end-point of the lifting x of the path x based at the point p(x).
(ii) Conditions (i) and (ii) in the statement of Proposition 1.2 (equivari-
ance conditions) are essential in the theory of transversely homogeneous
foliations.
Exercise 1.2.7. Give a demonstration of Darboux-Lie Theorem (Theo-

rem 1.9) according to the following suggestion:
Given 1-forms forming a basis {1 , ..., n } of the Lie Algebra of the
Lie group G and given 1-forms {1 , ..., n } a rank-n system of 1-forms
k
in a manifold M such that dk = cij i j , where the {ckij } are the
i,j
structure constants of the Lie Algebra relatively to the given basis, we
can dene 1-forms j = j j , j = 1, ..., n; in a natural way in the
product manifold M G. The system {1 , ..., n } is integrable and by
Frobenius Theorem denes a foliation F of the product manifold. Given a
leaf L F we have that j and j coincide over L. Using then the natural
projections M G M and M G G we can obtain local submersions
: U M G such that j = j , j. In order to conclude one has to
prove that if a dieomorphism of G preserves j for all j then is a left
translation in G.
1.2.9 Fibrations and the theorem of Ehresmann

The bers of the bundle are the leaves of a foliation on E. Such a foliation
is also called a bration. This situation is quite usual as shown in the
following result:
Theorem 1.10 (Ehresmann). Let f : M N be a C 2 submersion which

is a proper map, (i.e., f 1 (K) M is compact K N compact). Then
f denes a ber bundle over N .
Proof. The proof is based in the construction of suitable compactly sup-

ported vector elds. Let q M be given and let F := 1 (q) E. Then
F is a compact submanifold of E. Choose local coordinates (t1 , . . . , tm ) in
Preliminaries 31
a neighborhood U of q in M , with tj (q) = 0, j = 1, . . . , m. We take U

small enough so that we have:
(i) 1 (U ) is relatively compact (recall that is proper) in E.
(ii) There exist smooth vector elds X1 , . . . , Xm in 1 (U ) such that

(Xj ) =
tj
Claim 1.2. We have ((y, p)) = y (y, p) V1 F .
Proof. Given y V1 and p F denote by (z) the solution of the ordinary

dierential equation (z) = Zy ((z)) with initial condition (0) = p which
is dened for all z D(2). Then (y, p) = (1) by denition. We have

m
(z) = Zy ((z)) = tj (y) Xj ((z)).
j=1
Therefore

m

( (z)) = tj (y) , that is,
j=1
tj
d m

(( )(z)) = tj (y) in Rm .
dz j=1
tj
Therefore, ( )(z) = ( )(0) = z (t1 (y), . . . , tm (y)) and then

( )(1) = (p) + (t1 (y), . . . , tm (y)) (since (p) corresponds to the
origin and (t1 (y), . . . , tm (y)) to y in the local chart (t1 , . . . , tm ))
((1)) = y and therefore quad ((y, p)) = y.
It remains to prove that (V1 F ) = 1 (V1 ) for suciently small

V1 q. Since = 2 we have ((V1 F )) V1 so that (V1 F )
1 (V1 ). If we do not have equality for suciently small V1 then we obtain
a sequence qn U with qn q and such that 1 (qn ) contains some point
pn which does not belong to the image of and in fact {pn } avoids some
neighborhood W of F in E. Therefore, since is proper, {pn } has some
convergent subsequence say, pnj p. But this implies (pnj ) (p) .
j j
For any point y U we consider the vector eld Zy := t1 (y).X1 + +
tm (y).Xm , dened in 1 (U ). In particular Zq = 0 and its ow is complete
(dened for all real time). Since Zy depends dierentiably on y U we
have the following:
Lemma 1.1. There exists a neighborhood q V U such that:

(i) for each y V , the ow of Zy is dened in
D(2) 1 (V ) (where D(2) = {z C ; |z| < 2}),
giving a smooth map
y : D(2) 1 (V ) 1 (U )
(where t is the real time)
(t, p) y (t, p)
with y (0, p) = p, p 1 (V ),

y
(t, p) = Zy (y (t0 , p).
t (t=0)
(ii) For some neighborhood q V1 V we have y (t, p) V , p V1 ,

t with |t| 1.
Now we may consider the time one ow map
: V1 F E, (y, p) := y (1, p) U.
Then is holomorphic and we have an inverse for , which is given by
1 : (V1 F ) V1 F, 1 (p) := y (1, p).
This inverse is well-dened because of (i) and (ii) above so that qnj (p)
and (p) = q. Thus p F what is not possible for pn E\W , n.
This contradiction show that we must have (V1 F ) = 1 (V1 ) for every
suciently small neighborhood V1 of q in M .
This is the case for instance if M is compact. One very important result
concerned with this framework is due to Tischler (see Chapter 8).
Theorem 1.11 (Tischler, [Tischler (1970)]). A compact (connected)

manifold M bers over the circle S 1 if, and only if, M supports a closed
non-singular 1-form.
This is the case if M admits a codimension one foliation F which is

invariant by the ow of some non-singular transverse vector eld X on M
as we will see in Chapter 8.
Preliminaries 33
1.3 Holomorphic Foliations
A (real) manifold M 2n is a complex manifold if it admits a dierentiable

atlas {j : Uj M R2n }jJ whose corresponding change of coordinates
are holomorphic maps j 1 i : 1 (Ui Uj ) R
2n
Cn (Ui Uj )
R C . Such an atlas is called holomorphic.
2n n
In this case all the basic concepts of dierentiable manifolds (as tangent
space, tangent bundle, etc...) can be introduced in this complex setting.
This is the case of the concept of foliation:
Denition 1.14. A holomorphic foliation F of (complex) dimension k an a

complex manifold M is given by a holomorphic atlas {j : Uj M Vj
Cn }jJ with the compatibility property: Given any intersection Ui Uj =
the change of coordinates j 1i preserves the horizontal bration on
Cn Ck Cnk .
Examples of such foliations are, like in the real case, given by non-
singular holomorphic vector elds, holomorphic submersions, holomorphic
brations and locally free holomorphic complex Lie group actions on com-
plex manifolds.
Remark 1.2. (i) As in the real case, the study of holomorphic foliations
may be very useful in the classication Theory of complex manifolds.
(ii) In a certain sense, the holomorphic case is closer to the algebraic
case than the case of real foliations.
1.3.1 Holomorphic foliations with singularities

One of the most common compactications of the complex ane space Cn
is the complex projective space CP (n). It is well-known that any foliation
(holomorphic) of codimension k 1 on CP (n) must have some singularity
(in other words, CP (n), for n 2, exhibits no holomorphic foliation in the
sense we have considered up to now.) Thus one may consider such objects:
singular holomorphic foliations as part of the zoology. Thus one may have
consider singular foliations when dealing with complex settings.
Example 1.17 (Polynomial vector elds on C2 ). Let X be a polyno-

mial vector eld on C2 , say X = P (x, y)(/x) + Q(x, y)(/y) = (P, Q).
We have an ordinary dierential equations:

x = P (x, y)
y = Q(x, y)
We have local solutions given by the theorem of existence and uniqueness

of Picard for ordinary dierential equations ([Hirsch and Smale (1974)]):
(z) = (x(z), y(z))
d
= (z) = X((z))
dz
Gluing the images of these unique local solutions, we can introduce the
orbits of X on C2 . The orbits are immersed Riemann surfaces on C2 ,
which are locally given by the solutions of X.
Now we may be interested in what occurs these orbits in a neigh-
borhood of the innity L . We may for instance compactify C2 as the
projective plane CP (2) = C2 L , L = CP (1). Some natural questions
are then:
What happens to X in a neighborhood of L ?

Is it still possible to study the orbits of X in a neighborhood of L ?
We may rewrite X as the coordinate system (u, v) : X(u, v) =

1
um Y (u, v), m N {0} where Y is a polynomial vector eld. The ex-
terior product of X and Y is zero in common domain U : X Y = 0. So,
orbits of Y (or X) are orbits of X (or Y ), respectively in U . Then the orbits
of X extend to the (u, v)-plane as the corresponding orbits of Y along L .
This same way, we may consider in the (r, s) coordinate system. These ex-
tensions are called leaves of a foliation induced by X on CP (2). We obtain
this way: A decomposition of CP (2) into immersed complex curves which
are locally arrayed, as the orbits (solutions) of a complex vector eld. This
is a holomorphic foliation F with singularities of dimension one on CP (2).
Denition 1.15. Let M be a complex manifold. A singular holomorphic

foliation of codimension one F on M is given by an open cover M = jJ Uj
1
and holomorphic integrable 1-forms j (Uj ) such that if Uj Uj = ,
then i = gij j in Ui Uj , for some gij O (Ui Uj ). We put sing(F )Uj =
{p Uj ; j (p) = 0} to obtain sing(F ) M , a well-dened analytic subset
of M , called singular set of F . M \ sing(F ) is foliated by a holomorphic
codimension one (regular) foliation F 1 .
Remark 1.3. We may always assume that sing(F ) M has codimension

2. If (fj = 0) is an equation of codimension one component of sing(F)
Uj , then we get j = fjn j where j is a holomorphic 1-form and sing(j )
does not contain (fj = 0).
Preliminaries 35
Using this we may also reformulate the denition above as follows:
Denition 1.16. A singular holomorphic foliation F of codimension one

on M is given by a pair F = (F , sing(F )) where:
(1) sing(F ) M is an analytic subset of codimension 2.
(2) F is a regular holomorphic foliation of codimension one on M = M \
sing(F ).
Remark 1.4. Assume that we have a holomorphic regular foliation F 1

on U \0, 0 C2 , U sing(F ) = \0. Choose local coordinates (x, y)
centered at 0 and dene a meromorphic function f : U \0 C, p
U \0, as f (p) = inclination of the tangent to the leaf Lp of F 1 . By
Hartogs Extension Theorem [Siu (1974)],[Gunning I (1990)] f extends to
a meromorphic function f : U C. We may write f (x, y) = a(x,y)
b(x,y) , a, b
O(U ) and dene
dy b(x, y)
= f (x, y) = ,
dx a(x, y)
that is,

x = a(x, y)
.
y = b(x, y)
Therefore, F is dened by a holomorphic 1-form = a(x, y) dy b(x, y) dx
in U .
Example 1.18. Let f : M C be a meromorphic function on the complex

manifold M . Then = df denes a holomorphic foliation of codimension
one with singularities on M . The leaves are the connected components
of the levels {f = const.}. The singular points of the foliation are divided
into two classes: (1) points where f or 1/f is well-dened, but has a critical
point. (2) indeterminacy points of f .
Example 1.19. Let G be a complex Lie group and : G M M a

holomorphic action of G on M . The action is foliated if all its orbits have
a same xed dimension. In this case there exists a holomorphic regular
foliation F on M , whose leaves are orbits of . However, usually, actions are
not foliated, though they may dene singular holomorphic foliations. For
instance, an action of G = (C, +) on M , : CM M is a holomorphic
ows. We have a holomorphic complete vector eld X = t |t=0 on M . The

singular set of X may be assumed to be of codimension 2 and we obtain
a holomorphic singular foliation of dimension one F on M whose leaves are
orbits of X, or equivalently, of .
One very general problem is the study and classication of actions of

complex Lie groups G on a given compact complex M . One possible ap-
proach is to consider the stand-point of singular holomorphic foliations
theory.
Example 1.20 (Darboux foliations). Let M be a complex manifold

and let fj : M C be meromorphic functions and j C complex num-

r
r
bers, j = 1, . . . , r. The meromorphic integrable 1-form = fj i df
fi
i
j=1 i=1
denes a Darboux foliation F = F (w) on M . The foliation F has
r

f= fj j as a logarithmic rst integral.
j=1
Example 1.21 (Riccati foliations). A Riccati Foliation on C C is

given in some ane chart (x, y) C C by a polynomial 1-form =
p(x)dy (y 2 c(x) yb(x) a(x))dx. Such a foliation is transverse to the
bration C C C , (x, y) x, except for a nite number of invari-
ant bers given in the ane part by p(x) = 0. This transversality allows
to dene a global holonomy of the horizontal projective line 0 = (y = 0)
which gives us a group of Moebius transformations G SL(2; C) of a non-
invariant vertical ber. If a 0 then 0 is an invariant divisor and G is
the usual holonomy of the leaf 0 \ sing F as dened above. In this case
z
the elements of G are of the form f (z) = 1+z . Thus the holonomy of the
leaf 0 \ sing F is solvable. In fact, the elements of G are ane maps after
the change of coordinates Z = z1 on C. Using this remark it is easy to see
that the foliation is transversely ane outside the invariant set S given by
the union of 0 and the invariant vertical bers given by the zeros of p(x).

If a 0 then F () is transversely projective outside S = {x} C,
p(x)=0
which is also invariant. We may induce a foliation on CP (2) with similar
properties.
Exercise 1.3.1 (Implicit ordinary dierential equations). An alge-

braic implicit ordinary dierential equation in n 2 complex variables
is given by expressions:
Preliminaries 37
() fj (x1 , ..., xn , xj ) = 0

where fj (x1 , ..., xn , y) C[x1 , ..., xn , y] are polynomials and the
(x1 , ..., xn ) Cn are ane coordinates. Clearly, any polynomial vector eld
X on Cn denes such an equation. In general () denes a one-dimensional
singular foliation in some algebraic variety of dimension n. In order to
see it we begin by dening Fj (x1 , ..., xn , y2 , ..., yn ) := fj (x1 , ..., xn , yj )
C[x1 , ..., xn , y2 , ..., yn ] polynomials in n + (n 1) = 2n 1 variables.
Put also Sj := {(x, y) Cn Cn1 ; Fj (x, y) = o} {(x1 , ..., xn , yj )
Cnx Cyj ; fj (x1 , ..., xn , yj ) = 0} Cn2 =: j Cn2
(y2 ,...,yj ,...,yn ) .
We consider the projectivizations Sj CP (2n 1) and the complete
intersection subvariety S := S2 ... Sn CP (2n 1). Given by the
dierential forms j := yj dx1 dxj (j = 2, ..., n) on Cn Cn1 . Prove
that {j = 0, j = 2, ..., n} denes an integrable system on S. We say that
the implicit dierential equation () is normal if S admits a normalization
(desingularization) by blow-ups : S S. In particular we obtain in
general a singular foliation F () of dimension one on the algebraic n-
dimensional subvariety S CP (2n 1). Denote by f1 : S Cn C1 the
projection in the rst coordinate f1 (x1 , ..., xn , y2 , ..., yn ) = x1 , and extend
it to a holomorphic proper mapping f1 : S CP (1). Assume now that S
admits a normalization : S S. Show that the foliation F () lifts to a
foliation by curves F () on S and f1 = f1 denes a holomorphic proper
mapping from S over CP (1). Finally, using Stein Factorization Theorem
f
nd a splitting f1 : S B CP (1) where : B CP (1) is a nite
ramied covering and f: S B is an extended holomorphic bration over
the compact Riemann surface B such that the following diagram therefore
commutes

S S
f f1

B CP (1)
for a map f1 : S CP (1).

Exercise 1.3.2. Let X, = x x + y y be a complex vector eld dened
in a neighborhood of the origin 0 C2 . Show that X, is transverse to the
3-spheres S 3 (0, R) for R > 0 small enough, if and only if, / C\R . Let
now X be a polynomial vector eld in C2 and assume that the singularities
of the corresponding foliation F on CP (2) are of local form X, with

/ / R. Choose small balls B(pj ) around the singularities pj sing(F )
in CP (2). Show that there is a foliation Fd in a manifold Md with the
following properties: This is a C regular codimension-two real foliation
Fd on a compact real 4-manifold Md , which contains two copies of the

r
foliated pair (CP (2) \ B(pj ), F
r ). By Schwarz Reection
j=1 CP (2)\ B(pj )
j=1
Principle the leaves of Fd have also natural structures of Riemann surfaces.
Any Riemannian metric g in CP (2) induces a C Riemannian metric gd
in Md , that can be chosen to be hermitian along the leaves of Fd . Show
that the leaves of the non-singular foliation F CP (2)\sing(F ) have the same
growth type than the corresponding leaves of Fd .
Chapter 2
Plane fields and foliations
2.1 Denition, examples and integrability
A k-plane eld on a manifold M m , 1 k m, is a map x M P (x),

such that P (x) is a k-dimensional subspace of Tx M . When k = 1, P
is called line eld. A k-plane eld P is of class C r if each point x M
has a neighborhood U x where there are dened k linearly independent
Cr
vector elds X 1 , . . . , X k : U T U generating P in U , namely P (x) =
Span(X 1 (x), . . . , X k (x)). In this case we say that X 1 , . . . , X k generate P
in U .
Example 2.1. A C r foliation F of dimension k denes the plane eld

T F (x) of class C r1 given by
T F (x) = Tx Fx .
The plane eld N F given by
N F (x) = Tx M/TxFx
is called the normal plane eld of F .
Question 2.1. Is any plane eld P of the form P = T F for some foliation
F ? Locally the answer is yes but in general the answer is no. This question
suggests the following denition.
Denition 2.1. A k-plane eld P of class C r is integrable if P = T F for

some C r+1 foliation F .
2.1.1 Frobenius Theorem

Let X, Y be two vector elds in a manifold M and p M be xed. Denote
by Xt the ow of X and similarly Yt . X, Y C r , r 2. We dene
39
Xt (Y )(p) = DXt (Xt (p)) Y (Xt (p)). Note that Xt (X)(p) = X(p), t.
Denition 2.2. The Lie bracket of X, Y is the vector eld [X, Y ] dened
by
d
LX (Y )(p) = [X, Y ](p) = t=0 (Xt (Y )(p)) X, Y C r , r 2.
dt
In local coordinates, [X, Y ] has the following form: Writing

X= ai , Y = bi
i
xi i
xi
one has
bj aj

[X, Y ] = ai bi .
i,j
xi xi x j
When X and Y are dened in an open set of Rm , the formula above

yields
[X, Y ] = DY (p) X(p) DX(p) Y (p).
A vector eld X is tangent to a plane eld P (denoted by X P ) if
X(x) P (x) for all x M .
Denition 2.3. A plane eld P is involutive if X, Y P [X, Y ] P .
Lemma 2.1. If F is a foliation, then its associated plane eld T F is in-

volutive.
Proof. Let X, Y be two vector elds tangent to T F . By using local

coordinates dening F one can assume that X, Y are dened in an open
set of Rm and if dim F = k, then
X(x, y) = (f (x, y), 0), Y (x, y) = (f (x, y), 0)

x g y g f x f y f g
[X, Y ] =
0 0 0 0 0 0
= (f x f f x f, 0).
Hence [X, Y ] T F and the proof follows.
Theorem 2.1 (Frobenius Theorem). Involutive plane elds are

integrable.
Plane fields and foliations 41
The converse holds by the previous lemma. Hence the following assertions
are equivalent:
(1) P is integrable.
(2) P = T F .
(3) P is involutive, (i.e., X, Y P [X, Y ] P ).
Since all line elds are involutive one has
Corollary 2.1.
All line elds are integrable.
Example 2.2. Dene M = R3 and let P the map given by P (x, y, z) =

Span(X, Y ), where X, Y are the vector elds dened by X(x, y, z) = (1 +
y, y, z) and Y (x, y, z) = (y1 , 1 + y, 0). As X and Y are orthogonals and
non-zero everywhere one has that P is a plane eld of class C . Let us use
the Frobenius theorem to show that P is not integrable. Easy computations
yield

0 1 0 010
DY = 0 1 0 , DX = 0 1 0 .
0 0 0 001
So,
DY X = (y, y, 0), DX Y = (1y , 1 + y, 0).
Hence
[X, Y ](x, y, z) = DY (x, y, z) X(x, y, z) DX(x, y, z) Y (x, y, z)
= (1 2y, 1, 0).
So [X, Y ] P [X, Y ] = X + Y , for some , R. But
[X, Y ] = X + Y (1 1y, 1, 0)
= (1 + y, y, z) + (y, 1 + y, 0)

1 2y = (1 + y) y

1 = y + (1 + y)

0 = z
Replacing by (x, y, z) = (1, 0, 1) one has = 0, = 1, 0 = 1, a

contradiction. We conclude that [X, Y ](1, 0, 1)
/ P (1, 0, 1) and then P is
not integrable by Frobeniuss.
Regarding (m 1)-plane elds and foliations we mention the following

result:
Theorem 2.2 (Thurstons homotopy theorem, [Thurston (1976)]).

Every (m1)-plane eld in a m-manifold M m is homotopic to an integrable
plane eld T F , where F is a C codimension one foliation of M .
Example 2.3 (integrable systems of dierential forms). Let

1 , ..., r be dierential 1-forms of class C r on a manifold M and assume
that they are linearly independent at each point p M n . We may consider
the distribution of (n r)-dimensional planes dened by (p) Tp M is
(p) = {v Tp M, j (p) v = 0, j = 1, ..., r}.
This distribution is called integrable if it is tangent to a r dimensional

foliation F on M . According to Frobenius Integrability theorem (see also
[Camacho and Lins-Neto (1985)]) this occurs if and only if the system of
1-forms is integrable what means that we have dj 1 ... r = 0 for
all j = 1, ..., r. This occurs for instance if we have a closed 1-form with
(p) = 0, p M . In this case we have a codimension one foliation F on
M which is dened by the Paan equation = 0. The leaves of F are
locally given by f = cte, where f is a local primitive for .
2.2 Orientability
Recall that a k-form w in M is a map
w : M k (T M )
p w(p) : Tp M Tp M R .
Remark 2.1 (Criterium for orientability of manifolds). A manifold

M m is orientable if, and only if, it admits a volume form w, (i.e., a m-form
w such that w(p) = 0, p M ).
Denition 2.4. A k-plane eld P in M m , 1 k m is orientable if

there is a covering {Ui } of M and k continuous linearly independent vector
elds X 1,i , . . . , X k,i : Ui T Ui so that
=1,i span(X k,i(x), .. . , X (x)), x Ui
1,i k,i
1) P (x)
X (x) . . . X (x)
2) Det > 0, x Ui Uj .
X 1,j (x) . . . X k,j (x)
We say that P is transversely orientable if there is an orientable plane

eld P on M such that T M = P P .
Proposition 2.1. A line eld P in M is orientable if, and only if, P (x) =
Span(X(x)) for some continuous everywhere non-vanishing vector eld X
on M .
Proof. The only if part is obvious by taking the trivial covering {Ui } =
{M } of M and X 1,i = X. Now, if P orientable then there exists {Ui }iI
open covering of M and k vector elds X i = X 1,i : Ui T Ui such that
1) P (x)
= Span(X

i
(x)), x Ui .
i
X (x)
2) det > 0, x Ui Uj , i.e., X i (x) = aij (x), X i (x),
X j (x)
x Ui Uj , aij (x) > 0.
Dene X(x) = X i (x) ||X i (x)||, x Ui . Then X is well dened
since
X i (x) X j (x)
= , x U i Uj
||X (x)||
i ||X j (x)||
X i (x) aij (x)X j (x) X j (x)
= =
||X (x)||
i ||aij (x)|| ||X (x)||
j ||X j (x)||
as ai,j (x) > 0 for all x. Since X i generates P in Ui the result follows.
Example 2.4. Choose M = R2 {0}. None of the line elds induced by

the foliations F1 , F2 in M at Figure 2.1 is orientable. One can see this by
observing how the tangent vector varies along the curve indicated at F1 .
Proposition 2.2. A manifold M is orientable if, and only if, the plane
eld P (x) = Tx M is orientable.
Proof. Since P is orientable there are a cover {Ui } and vector elds
X 1,i , . . . , X m,i : Ui T Ui such that
Tx M = Span(X 1,i (x), . . . , X m,i (x)) x Ui
and det(X n,k (x))1nm k = i, j > 0, x Ui Uj . For each Ui one choose
an m-form WUi such that if x Ui then {v1 , . . . , vm } is a base of Tx M
with

v1 , . . . , vm
det > 0 WUi (x)(v1 , . . . , vm ) > 0.
X 1,i (x), . . . X m,i (x)
F
2
F
1
Fig. 2.1
Let {i } be a partition of the unity subordinate to the cover {Ui }.

Dene W = i , WUi . Then w is a m-form with w(x) = 0, x. In fact,
i
for x M , w(x) = i (x)WUi (x). Let i be such that x Ui and
{i;xUi }
{v1 , . . . , vm } is a base Tx M satisfying

v1 , . . . , vm
det > 0.
X , . . . , X m,1 (x)
1,i
Let j be such that x Uj . Note that

v1 , . . . , vm X 1,i (x) . . . X m,i (x) v1 , . . . vm
= .
X (x), . . . , X m,j (x)
1,j
X 1,j (x) . . . X m,j (x) X 1,i (x) . . . X m,i (x)

v1 . . . vm
Because P is orientable one has det > 0. So
X 1,i . . . X m,1 (x)
WUi (x)(v1 , . . . , vm ) > 0. Henceforth i (x) WUi (x)(v1 , . . . , vm ) 0, i.
Then

W (x)(v1 , . . . , vm ) = i (x) WUi (x)(v1 , . . . , vm ) > 0
{x;xUi }
and so there is i such that i (x) = 1. It follows that, w(x) = 0, x M

and therefore M is orientable.
Notation: For x M we denote

s,i 1,i
X (x) X (x) . . . X k,i (x)
=
X s,j (x) X 1,j (x) . . . X k,j (x)
and

X s,i (x) X 1,i (x)X 2,i (x) . . . X k,i (x)
= .
X s,j (x) X 1,j (x)X 2,j (x) . . . X k,j (x)
Corollary 2.2. Let P and P be two plane elds in a manifold M such that
a) T M = P P , (i.e., Tx M = P (x) P (x), x M );
b) P and P are orientable.
Then, M is orientable.
Proof. Exercise.
Example 2.5. Let F be the Reeb foliation in the Moebius band (see
Section 1.2.2). Then T F is not orientable. To see this we let M =
[1 + , 1 ] R, > 0 and M /F be the Moebius band. If T F were
orientable, then T F/(M /F ) would be orientable. There is a line eld P in
M /F which is orientable. In fact: it suces to choose P (x) as T F1 where
F1 is the projection of the vertical foliation in M /F . P is induced by the
vertical vector eld X(x, y) = (0, 1). Note that X induces a vector eld in
M /F since

1 0 0 0
DF (0, 1) = = ,
0 1 1 1
hence P is orientable. But then T (M /F ) = P (T F/(M /F )) would be

orientable, a contradiction. The result follows.
Denition 2.5. A foliation F is orientable (resp. transversely orientable)

if its associated plane eld T F is.
Note that if M supports a foliation F which is both orientable and trans-

versely orientable, then M is orientable (as a manifold). If M is orientable,
then a foliation F in M is transversely orientable F is orientable. If F
is a codimension one transversely orientable foliation, then there is a vector
eld X in M such that X F . Warning: The above does not implies that
the time-t map of X preserves F, (i.e., Xt (Fx ) is a leaf for every leaf Fx ,
x M ).
2.3 Orientability of singular foliations
A C r singular foliation on a surface S is a C r -foliation F in the complement

S \ sing(F ) of a discrete set sing(F ) in the interior of S which is either
transverse or tangent to the boundary of S. We denote by Fx the leaf of
F containing x S \ sing(F). One says that F is C r -locally orientable if
there is an open cover {Ui : i I} of S and a C r vector elds Yi in Ui such
that sing(Yi ) = Ui sing(F ) and Tx Fx = Span(Yi (x)) , x Ui \ sing(Yi ),
where sing(Yi ) denotes the set of zeroes of Yi . One says that F is C r
orientable if the cover {Ui : i I} above can be chosen with a single
element U1 = S. This notion of orientability diers from the corresponding
one for non-singular foliations due to the presence of the singularities. One
can easily construct singular foliations in D2 which are not locally orientable
(it suces to complete the ones described in Figure 2.1 to the whole D2 ).
Clearly a C r locally orientable singular foliation is C r orientable. The
converse is false in general but true when S = D2 , the 2-disc in R2 . Indeed,
let F be a C r singular foliation in D2 . For an open set U of D2 one denes
XFr (U ) as the space of C r vector elds in U such that sing(Y ) = U sing(F )
and Tx Fx = Span(Y (x)), x \ sing(Y ). A nite family U1 , , Uk of
open sets in D2 is a chain whenever Ui Ui+1 = is connected for all
1 i k 1. Given a chain U1 , U2 and Y1 , Y2 XFr (U1 ), XFr (U2 ) we
dene U1 ,U2 (Y1 , Y2 ) to be either Y2 (if Y1 , Y2 have the same orientation in
U1 U2 ) or Y2 (otherwise). This denition makes sense because U1 U2 =
is connected. Clearly U1 ,U2 (Y1 , Y2 ) XFr (U2 ) and both Y1 and U1 ,U2 (Y2 )
have the same orientation in Ui U2 . For general chains U1 , , Uk and
Yi XFr (Ui ) (i = 1, , k) we dene Z1 = Y1 , Zi+1 = Ui ,Ui+1 (Zi , Yi+1 )
and U1 , ,Uk (Y1 , , Yk ) = Zk . Under this denition one has
U1 , ,Uk (Y1 , , Yk ) = Uk1 ,Uk (U1 , ,Uk1 (Y1 , , Yk1 ), Yk ). (2.1)

Now let us assume that F is C r locally orientable and let {Y i XFr (Ui ) :
i = 1, , r} be a xed C r local orientation of F . We can assume that all
the Ui s are balls, and so, Ui Uj is either empty or connected for all i, j. De-
ne Y 1 = Y 1 and for i = 2, , r we dene Y i = Ui1 ,Uik (Yi1 , , Yik ),
for some chain Ui1 , , Uik with i1 = 1 and ik = i. The simply connect-
edness of D2 implies that the value of Yi does not depend on the chosen
chain Ui1 , , Uik . Let us prove that if Ui Uj = , then Yi and Yj have
the same orientation in Ui Uj . In fact, let Ui1 , , Uik and Uj1 , , Ujs
be two chains realizing Yi and Yj respectively. Hence Ui1 , , Uik , Uj is a
chain, and so, the invariance of with respect to the chains implies
Yj = Ui1 , ,Uik ,Uj (Yi1 , , Yik , Yj ).
Then Eq.(2.1) and ik = i implies
Yj = Ui ,Uj (Ui1 , ,Uik (Yi1 , , Yik ), Yj ) = Ui ,Uj (Yi , Yj )
proving that Yi and Yj have the same orientation in Ui Uj as desired.

Next we consider a C partition of the unity {Q1 , . . . , Qr } of the cover
U1 , , Ur and dene

k
Y = Qi Yi .
i=1
This vector eld yields a C r orientation of F .
2.4 Orientable double cover
Let P be a k-plane eld in a manifold M m . Dene Bx (M ) the set

of ordered basis of P (x) when x M . Note that {v1 , . . . , vk } =
{v2 , v1 , . . . , vk } Bx (M ). Dene the following relation in Bx (M ):
vs
(v1 , . . . , vk ) x (w1 , . . . , wk ) det > 0. Remember that
ws
k
vs v1 , . . . , vk
= , vi = aij wj .
ws w1 , . . . , wk
i<1
Note that if (v1 , . . . , vk ) x (w1 , . . . , wk ), then (v1 , v2 , . . . , vk ) /x

(w1 , . . . , wk ). The relation x is an equivalence since det AB = det Adet B.
Dene Ox (M ) = Bx (M )/ x with projection
x : Bx (M ) Ox (M ).
Note that
Ox (M ) = {Ox , Ox }
has just two elements. Sometimes we use the notation Ox = Ox . Finally
we dene
= {(x, Ox ) : x M, Ox Ox (M )}.
M
depends on the plane eld P . When P (x) = Tx M , x M ,
At rst M

M is called the orientable double cover of M . In general we call it the
orientable double cover of P .
is a m-dimensional manifold.
Proposition 2.3. M
. By the denition of plane eld P , there is a
Proof. Fix (x0 , Ox0 ) M
Cr
neighborhood U of x0 and k smooth vector elds X 1 , . . . , X k : U T U
such that
P (x) = Span(X 1 (x), . . . , X k (x)), x U.
as:
Dene U
= {(x, x (X 1 (x), . . . , X k (x)); x U },
U
if x0 (X 1 (x0 ), , X k (x0 )) = Ox0 and
= {(x, x (X 1 (x), . . . , X k (x)); x U },
U
if x0 (X 1 (x0 ), , X k (x0 )) = Ox 0 .
We consider the projection

M ; (x, Ox ) = x.
: M
We can assume that U is the domain of a local chart (U, Q) around x0 .
Dene the chart (U , Q)
by Q = Q .
. In fact, let (U, Q), (V, ) be two
The family {(U , Q)} is an atlas of M

such charts with U V = . Then Q1 : Q( U U
V ) ( V ) satises
Q
(y) = Q (y). Hence M
1 1 is a manifold of class C and dimension
r

m. Moreover, : M M is dierentiable and even a local dieomorphism.
Tx M . Dene also a k-plane eld P in M
Note that T(x,Ox ) M given by
)1 (x) (P (x)).
P (x, O2 ) = a(/U
= (M
M , , P) is the orientable double cover of P .
Theorem 2.3. Let M be a manifold and P a plane eld in M with ori-

, P ). Then, M
entable double cover (M is connected if, and only if, P is
not orientable.
Proof. Assume that M is connected. Then we can x a curve c M

joining (x0 , Ox0 ) with (x0 , Ox0 ) The curve c = c in M is closed and
contains x0 . Suppose by contradiction that P is orientable. Let {Ui } the
cover of M and {X 1,i , , X k,i } be the vector elds in Ui generating P
such that

X s,i (x)
det > 0. (2.2)
X s,j (x)
Because c is compact we can suppose that U1 , , Ur is a cover of c. We
can further suppose that x0 U1 and Ox0 = x0 (X 1,1 (x0 ), , X k,1 (x0 )).
given by
Dene a new curve c W
c(t) = (c(t), ct (X 1,i (c(t)), , X k,i (c(t))), if c(t) Ui ,
and i = 1, , k.
Note that c is well dened by Eq.(2.2). In addition, c is continuous
because both c(t) and t c(t) (X 1,i (c(t)), , X k,i (x0 )) are. Dene
B = {t [0, 1] : c(t) =
c(t)}.
We have that B = because x0 B. Moreover, B is closed because c
and c are continuous. Let us prove that B is open. In fact, if t0 B then
c(t0 ) Ui0 for some i0 . If t0 / Int(B) then there is a sequence tn t0
in [0, 1] such that c(tn ) =
c(tn ) for all n. Because c is continuous and
c(t 0) Ui0 we can suppose that c(tn ) Ui0 for all n yielding
c(tn ) = (c(tn ), c(tn ) ((X 1,i0 (c(tn )), , X k,i0 (c(tn ))).
Write c(t) = (c(t), (t)), where (t) a(c(t)) is continuous. Because
c(tn ) =
c(tn ) one has
c(tn ) ((X 1,i0 (c(tn )), , X k,i0 (c(tn ))) = (tn ).
By taking limits the last expression yields
c(t0 ) (X 1,i0 (c(t0 )), , X k,i0 (c(t0 ))) = (t0 )
contradicting c(t0 ) =
c(t0 ). This contradiction shows that c(t) =
c(t)
for all t and then c(1) = c(1). This would imply Ox0 = Ox0 which
is absurd. This proves the rst part. For the converse, we shall prove
not connected then P is not orientable. Recall the projection
that if M
: M M given by (x, Ox ) = x. Let M be a connected component

of M . Observe that (M ) = M . In fact, since is a local dieo-
morphism we have that (M )
) is open in M . Let us prove that (M
is closed in M . Choose xn (M ) x M . By denition there

is a neighborhood U of x and k vector elds X 1 , , X k generating P
in U . Obviously there is yn M such that ( yn ) = xn . Note that

yn = (xn , Oxn ) for some Oxn a(xn ). Without loss of generality we can as-
sume that Oxn = xn (X 1 (xn ), , X k , (xn )) for all n. Passing to the limit
the last expression yields Oxn x (X 1 (x), , X k , (x)) = Ox . Hence
yn (x, Ox ). Since M is closed in M we conclude that (x, Ox ) M .

So, x = (x, Ox ) (M ) proving that (M ) is closed. Because M is
connected we conclude that (M ) = M as desired.
On the other hand, since 1 (x) has two elements for all x M we con-
clude that M has two connected components which we denote by M 2 .
1 , M

This implies that x M and Ox O(x) if (x, Ox ) M1 (x, Ox )
M2 ). It follows that /M 1 is one-to-one. Hence : M 1 M is a dieo-
morphism. Since P is orientable, and P = 1 (P ) we would have that P
is orientable, a contradiction. This proves the theorem.
Corollary 2.3. Every plane eld on a simply connected manifold is ori-

entable and transversely orientable. In particular, all simply connected
manifolds are orientable.
Proof. Let P be a plane eld on a simply connected manifold M . If P

were not orientable then its double cover : M M is connected. Be-
ing M simply connected we would have M = M , where M M is the
universal cover of M . This would imply that M M is non-trivial cover,
a contradiction. This contradiction proves that P is orientable. That P
is transversely orientable follows applying the previous result to a comple-
mentary plane eld of P . The result follows.
Example 2.6. The Reeb foliation in S 3 is orientable and transversely ori-

entable (because 1 (S 3 ) = 1).
Exercise 2.4.1. Show that any foliation in the solid torus D2 S 1 tangent
to the boundary is orientable.
2.5 Foliations and dierentiable forms
Remember that a dierential k-form w of M is a multilinear map associating

to each point p M a linear k-form in Tp M , that is is, w(p) k (Tp M ),
where k (E) denotes the space of k-forms in a vector space E. The space
of all k-forms in M is denoted by k (M ). If w k (M ) and (M ),
then the alternating product w k+ (M ). If w is a k-form, then there

is a derivative dw of w, d = dk : k (M ) k+1 (M ). The form w is closed
if dw = 0, and w is exact if w = d. As it is well-known:
Poincares lemma: An exact form of class C 2 is closed, i.e., d(dw) =

d2 (w) = 0, w k (M ). Conversely, a C 1 closed form is locally exact.
Denote by Z k (M ) = Ker(dk ) is the set of closed k-forms and by
B (M ) = dk1 (k1 (M )) is the set of exact k-forms. Poincares lemma
k
implies that exact forms are closed, namely B k (M ) Z k (M ). The quo-

tient space H k (M ) = Z k (M ) B k (M ) is called the (de Rham) cohomology
k-group of M .
Let w 1 (M n ) be a non-singular 1-form namely, w(p) = 0, p M .
The map p Ker(w(p)) denes a (n 1)-plane eld in M .
We leave the following proposition as an exercise for the reader (use

Frobenius theorem).
Proposition 2.4. The following conditions are equivalent for a smooth

dierentiable one-form w 1 (M ).
(1) w is integrable if.

(2) dw w = 0.
(3) dw = w for some 1 (M ).
Example 2.7. If w is closed, then w is integrable and so w denes a

foliation Fw . This remark apply to the following. Dene M = R2 ,
w = adx + bdy, a, b R. Then w is closed w induces a foliation Fw . The
leaves of this foliation are given by the solution of the dierential equation
adx + bdy = 0 y = b/a.
The general solution of this equation is the straight-line family y = (b/a)
x + K, K R. This gives a foliation of R2 by these straight-lines.
Example 2.8 (Thurston). Let L be a closed manifold with H 1 (L) = 0.

Let be a closed non-exact 1-form of L and f : S 1 R be a dierentiable
map. Denote by d the standard 1-form of S 1 . We dene the 1-form w in
the product M = L S 1 given by
w = d + f ().
Note that
dw = d(d + f ()) = d(f ()) = f () d + f () d
and d = 0. It follows that

dw w = (f () d ) (d + f ())
= f () d d + f () d f () = 0.
It follows that w is integrable, i.e., Ker(w) is tangent to a C 1 codimension
one foliation Fw in M . Note that the sets L0 = {(x, ) M : = 0 },
where 0 f 1 (0) are compact leaves of Fw . In fact, x 0 f 1 (0) and
(x, ) L0 . If v(x,) T(x,) L0 v(x,) = (vx , 0) and = 0 . Hence
w(v(x,) ) = w(vx , 0) = d(0) + f (0 )(vx ) = 0 + 0 (vx ) = 0
proving that L0 is a leaf of w . Clearly L0 is dieomorphic to L and so
L0 is a compact leaf of Fw (recall that L is closed).
Chapter 3
Topology of the leaves
3.1 Space of leaves
Let F be a foliation of a manifold M . Given z M we denote by Fz the leaf

of F that contains z. The relation x, y M , x y x Fy y Fx
is an equivalence. The quotient space OF = M/N is called the space of
leaves of F . Denote by : M OF the projection. We set in OF the
topology making continuous, namely V OF is open 1 (V ) M

is. If A M we dene F (A) = Sat(A) = Fx and call it the saturation
xA
of A. This set is formed by those x such that Fx meets A. The set A is
saturated by F if F (A) = A.
Example 3.1. The leaf space of the foliation F2 in Figure 1.2 is not Haus-
dor. In fact the vertical boundaries of I R correspond to elements in the
leaf space which cannot be separated by open sets.
Proposition 3.1. The saturation F (A) M of an open set A M is

open.
Proof. Choose x F (A). By denition, Fx A = , hence there is

y Fx A. There there exists a nite collection of plaques 1 , 2 , . . . , k
of F in Fx such that i i+1 = , x k y 1 . Let Ui be the
domain of the chart of F dening i . Because y A, A is open, we can
suppose Ui A. We project Ui into an open set of F (A) containing x as
follows: consider U1 U2 which is non-empty since = 1 2 U1 U2 .
Let U1 = U1 , U 2 = U {; be a plaque of U2 with U = }. Note
that 2 U 2 . We have U
2 is open and U2 F(A) (since, for a plaque
we have U 2 = U1 = F(A) = F(a)). Also,

2 3 = . Dene U3 = U {, is a plaque of U3 with U 2 = }. As
53
2 3 = and 2 U2 , we have that 1 U 3 . Hence U 3 is open and

3 F(A). Inductively we have U
U i , i = 1, . . . , k, such that i U
i ,

Ui is open and Ui F (A). Hence x k Uk and F (A) is open.
Corollary 3.1. The projection : M OF is open, (i.e., it sends open

sets into open sets)
Proof. Let A M be open. We have

x F (A) Fx A = y Fx A Fx = Fy
and
y A (x) = (y) (A) (x) (A) x 1 ((A)).
Therefore
F (A) = 1 ((A)).
Because F (A) is open we have that (A) is open with respect to the
quotient topology.
Warning: Not every projection is open. For instance, consider the projec-
tion of the parabola y = x2 , in R2 , into the y-axis.
Denition 3.1. We say that a subset A M is invariant for F (or F -

invariant) if A = F (A) = Sat(A).
Lemma 3.1. If A is F -invariant then A, Int(A) and A are F -invariant.
Proof. We have Int(A) is open. Thus F (Int(A)) is open and therefore

Int(A) F (Int(A)) F (A) = A. Since Int(A) is the biggest open set
contained in A we have F (Int(A)) = Int(A). Assume now that A is F -
invariant. Then M \A is also invariant. Thus Int(M \A) is invariant and
Int(M \A) = M \A M \A is F -invariant. Therefore, A is F -invariant.
Finally, A = A\ Int(A) where A and Int(A) are F -invariant. Hence A is
F -invariant.
Theorem 3.1. Let F be a leaf of a foliation F and be a transverse

section of F intersecting F . Then, one of the following alternatives holds:
(1) F is discrete.
(2) F has non-empty interior in .
Topology of the leaves 55
(3) F is a perfect set, i.e., a closed set without isolated points with
empty interior.
Proof. It suces to prove that if F is not discrete then F is

perfect. Suppose by contradiction that F is not perfect, i.e., F
has an isolated point x0 . Because x0 is isolated in F we have x0 F .
Because F is not discrete, there is x F which is an accumulation
point of {xn } F , xn = x . Because xn F we have that F passes
arbitrarily close to x . Using a suitable plaque sequence we can see that F
passes close to x0 (see Figure 3.1). This is a contradiction and the proof
follows.
x0
xn
F
x*
Fig. 3.1

3.2 Minimal sets
Let F be a foliation in M . A subset M is called minimal for F if

1) is closed and F -invariant.

2) if = satises (1), then = .
Remark 3.1. Zorns lemma applied to the set of closed invariant subsets of
M (ordered by inclusion) implies that there is at least one minimal subset.
Every closed leave is a minimal set. If is minimal and F is a leaf of F in
, then F = . If is minimal and meets a closed leaf F , then = F . In
general the set of minimal sets is pairwise disjoint. The Reeb foliation in
S 3 has a unique minimal set which is its compact toral leaf. The irrational
foliation in T 2 has T 2 as its unique minimal set. As we shall see later on
there is no minimal foliation, (i.e., the whole manifold is minimal) in S 3
of codimension 1 (by Novikovs compact leaf theorem). Analogously there
are no minimal foliations on compact manifolds with nite fundamental
group. Any foliation in the Klein bottle has a compact leaf. Hence there is
no minimal foliation in the Klein bottle. A foliation is transitive if it has
a dense leaf. Minimal foliations are transitive. The converse however, is
not true. There is no transitive codimension one foliations on compact 3-
manifolds with nite fundamental group. In particular S 3 does not support
transitive codimension one foliations. Minimal foliations have no compact
leaves. As we shall see later, a transitive codimension one foliation on a
compact 3-manifold, has no compact leaves as well.
Exercise 3.2.1. Find a non-minimal transitive codimension one foliation on

a compact three-manifold M 3 .
Lemma 3.2. Suppose that F is a foliation in M and is minimal for F .

Then, Int() = = M .
Proof. Clearly = M Int() = .

Conversely, let be minimal with Int() = . On one hand is closed
by denition. On the other hand, is open since = Int() as x
F x = F x Int() = x F(Int()) = Int()). Since M is
connected we conclude that = M and the proof follows.
Proposition 3.2. Suppose that is a p-disc (p =cod F ), is a minimal

subset of F with = and = . If is not a closed leaf then
is perfect.
Proof. Assume that is not a closed leaf and prove that is a perfect
set. For this we proceed as follows. Observe that is compact since it
Topology of the leaves 57
is a p-disc. Let F be a leaf of F . It suces to prove that F is

perfect. By contradiction suppose that it is not so. Then either F is
discrete or F has non-empty interior by Theorem 3.1. In the later case
has non-empty interior (in ) since F is dense and is compact.
This would imply that has non-empty interior and then = M by the
previous lemma. This is a contradiction because = . We conclude
that F is discrete, and so, F is nite. If F were not closed then we
could nd x F \ F . Because F and x F we have x . Because
is minimal the leaf Fx of F containing x is dense in . In particular
Fx Int() = (recall = ). By applying the argument described
in Figure 3.1, we would have that F intersects innitely many times, a
contradiction. This contradiction proves that F is a closed leaf. Since F is
dense in we would have that = F is a closed leaf which is impossible.
This contradiction proves the result.
Denition 3.2. A minimal set of a foliation on a manifold M is exceptional

if it is neither a closed leaf nor the whole manifold M .
A very interesting and deep problem is:
Problem 3.1. Find necessary and sucient conditions for the existence of
exceptional minimal sets.
Lemma 3.3.
In codimension one minimal sets are described below. This is a straight-
forward consequence of the above results and of the well-known description
of perfect sets in the real line. A nowhere dense minimal set of a codi-
mension one foliation is either a closed leaf or an exceptional minimal set.
In particular, an exceptional minimal set of a codimension one foliation,
meets a transverse section in a set homeomorphic to a Cantor set.
Denition 3.3. A leaf L of a codimension one foliation F on a manifold
M is called exceptional if for some (and therefore for any) transverse section
M to F meeting L the intersection L is nowhere dense and without
isolated points. In this case, the closure LR , is homeomorphic to a
Cantor set.
Example 3.2. The irrational foliation in T 2 is minimal, and so, it has no
exceptional minimal sets. In fact a foliation arising from a C 2 vector eld
on a closed surface has no exceptional minimal sets. This is false for C 1
vector elds by the classical Denjoys counterexample. The Reeb foliation
in S 3 has no exceptional minimal sets.
Example 3.3. Let B be the bitorus and consider the representation Q :

1 (B) Di(S 1 ) as described in Section 1.2.6. Recall that the behavior
of the suspended foliation FQ depends on the maps f, g Di(S 1 ) used
in the construction of Q. By a suitable choice of f, g we have that FQ is
a C foliation in some closed Seifert 3-manifold exhibiting an exceptional
minimal set. This example (due to Sacksteder) gives a counterexample for
a possible version of the Denjoys theorem.
Exercise 3.2.2. Are there transitive codimension one foliations with excep-
tional minimal sets?
Chapter 4
Holonomy and stability
4.1 Denition and examples
An important tool for the study of foliations is the notion of holonomy group
dened as follows. Let F be a foliation on a manifold M . Let Ui , Uj be two
charts of F with Ui Uj = . Denote by i : Ui i and j : Uj j
the projection along the plaques. Suppose that every plaque (of F) in Ui
intersects at most one plaque in Uj . Then we can dene
fi,j (x) = j (x ),
where x i and x is the unique plaque of Ui containing x Ui . The
resulting map
fi,j : Dom(fi,j ) i j
is called the holonomy map induced by the two foliated charts
(Ui , Xi ), (Uj , Xj ). Let U1 , , Ur be a nite family of foliated charts such
that every plaque of Ui intersects at most one plaque of Uj (for all i, j). We
can dene the holonomy map f1, ,r : Dom(f1, ,r ) 1 r by
f1, ,r = fr1,r fr2,r1 f1,2 .
Now, let L be a leaf of F and x, y L. Clearly L is connected (by denition)
and so there is a curve c : [0, 1] L joining x and y. This curve can covered
by a nite family of foliated charts U1 , , Ur such that x U1 , y Ur
and every plaque of Ui intersects at most one plaque of Uj (for all i, j).
Without loss of generality we can assume that x 1 , y r . The map
fc = f1, ,r
is the holonomy induced by the curve c. Note that by denition we have
fc (x) = y. One can easily prove that fc does not depend on the foliated
59
cover U1 , , Ur . Moreover, fc only depends on the homotopy class of c.

More precisely, if c, c L are homotopic in L (with xed end points) then
fc = fc in an open subset of 1 containing x. When x = y we obtain a
representation
: 1 (L) Germ()
given by
() = [fc ],
where c is a representative of 1 (L), is a transverse of F containing
x L ( 1 ) and
Germ() = {f : Dom(f ) : f (x) = x}/
is the group of germs of C r maps with xed at x. The equivalence relation
on the space of C r maps {f : Dom(f ) : f (x) = x} is dened
by: f g if and only if f and g coincide in a neighborhood of x.
Denition 4.1. The image Hol(L, , x) = (1 (L)) of is called the

holonomy group of L with respect to and x.
Up to conjugacy by C r germs of dieomorphisms, the group Hol(L, , x)

does not depend on and x L. This allows us to dene the holon-
omy group of the leaf L as the conjugacy class Hol(F, L) = Hol(L) of the
groups Hol(L, , x), under conjugation by C r germs of dieomorphisms. A
representative of Hol(L) is then a group Hol(L, , x). The algebraic and
dynamical properties of the holonomy group Hol(L) can be read on any
representative Hol(L, , x) as we shall see.
We say that the leaf L has no holonomy or is without holonomy or still
has trivial holonomy if Hol(L) = {Id}. A foliation without holonomy is a
foliation whose leaves are without holonomy.
Example 4.1. A simply connected leaf has trivial holonomy. In particular

a foliation by planes R2 is without holonomy.
Example 4.2. There are foliations admitting non-simply connected leaves

without holonomy. Indeed, dene M0 = I T 2 where I is a compact
interval. Then M0 has a boundary formed by two torii T1 (external one) and
T2 (internal one), see Figure 4.1. Gluing T1 with T2 by a dieomorphism
: T1 T2 , we obtain a closed manifold M . The trivial foliation of M0
formed by concentric torii T 2 , I denes a foliation F of M . Any
Holonomy and stability 61
c1
F
T1
T2
Fig. 4.1
leaf of F is a torus. Every torus bundles over S 1 can be obtained in this

way. We observe that F is a foliation without holonomy. This can be seen
as follows. Let F be a leaf of F . Then 1 (F ) = Z2 is the free abelian
group with two generators [c1 ], [c2 ], where c1 , c2 are the meridian curve
and the parallel curve in T 2 respectively. The generator c1 is depicted at
Figure 4.1. Consider the transverse surface S = I c1 in Figure 4.1. Note
that F intersect S in a circle foliation. The holonomy induced by c1 in S is
precisely the rst return induced by this circle foliation. Since this return
map is the identity one has ([c1 ]) = Id. A similar argument shows that
([c2 ]) = Id. Since [c1 ], [c2 ] are the generators of 1 (F ) we conclude that
Hol(F ) = (1 (F )) = 0. This proves that F has no holonomy as desired.
With similar arguments we can prove that all foliation arising from a surface
bundle over S 1 are without holonomy.
Example 4.3 (Holonomy of the Reeb foliation). Let F be the Reeb

foliation in S 3 described in Chapter 1 Example 1.3. Then F has a torus leaf
T and all remaining leaves are planes (and so they have in the holonomy).
To calculate Hol(T ) we proceed as in the previous example. Indeed, as
before 1 (T ) is generated by the meridian curve and the parallel curve
c1 , c2 .
c1
Fig. 4.2
If S is a transverse annulus as in Figure 4.2, is a transverse interval

in S centered at x0 T then the foliation induces a ow on it whose
return map f is as in the right-hand side gure at Figure 4.3.
This map is precisely the holonomy of c2 . Analogously c2 produces a
holonomy having the graph depicted in the left-hand side gure at Fig-
ure 4.3. Now Hol(T ) is generated by the classes of there two maps. Note
that Hol(T ) is abelian since it is the homomorphic image of Z2 (which is
abelian). Because Hol(T ) is torsion free we conclude that Hol(T ) = Z2 .
Example 4.4. A foliation F tangent to a closed non-singular C 1-form

w in a manifold M has trivial holonomy. Indeed, let X be the gradient of
w dened by
wp (vp ) = X(p), vp ,
for all p M and vp Tp M . Clearly X is non-singular since w is. In

addition F is transverse to F . Let F be a leaf of F and c a closed curve in
F . We can assume that c : S 1 F is an immersion. Set I = [1, 1] and
x0
Fig. 4.3 Holonomy of the Reeb foliation.
dene the map : S 1 I S = (S 1 I) by

(, t) = Xf (c()).
It is clear that is an immersion of class C r , r 2. Then w = (w)
is a well dened 1-form in S 1 I. Because dw = d(w ) = (dw) =
(0) = 0 we have that w is closed. Hence w denes a foliation F in
S 1 I. Note that F is conjugated to F S. It follows that the curves
c = S 1 0 and c have the same holonomy. Let us calculate the holonomy
of c . Fix , 0) c and = . Clearly is a transverse of F .
Let f : Dom(f ) be the holonomy of c , p Dom(f ) and
q = f (p). Let be the arc in joining p and q.
Let l be the arc in F joining p, q. Let R be the closed region bounded
by the curves c , l and . Because

0= dw = w = w + w =0+0+ w
R R l
one has

w = 0.

This equality implies that is trivial and so p = q = f (q). We conclude

that c has trivial holonomy. Hence c has trivial holonomy and the proof
follows.
Example 4.5 (Holonomy of suspended foliations). The suspension

of a group action is introduced in Section 1.2.6. Regarding the holonomy
p

q
c*
l
Fig. 4.4
of the resulting foliation we have:
Theorem 4.1. Let B Q F be the suspension of a representation
Q : 1 (B) Di r (F )

and FQ , FQ the corresponding foliations. Then,

(1) FQ is a foliation without holonomy.
(2) Hol(L) Q(1 (B)), leaf L of FQ .
Proof. Let B, F be smooth manifolds and M = B Q F be the suspension

of a representation Q : 1 (B) Di(F ). Recall that M is equipped with

two foliations FQ and FQ which are the projection over B Q F of the

trivial foliations {B } and { F } on B F respectively. Because the

foliation FQ is induced by a bration (with ber F ) we can see that FQ
has no holonomy. So, it is enough to study the holonomy of FQ . For this
we x a leaf L of FQ and choose x0 L. Fix (b0 , f0 ) 1 (x0 ), where
:B F B Q F is the natural projection. It follows from the denition
of FQ that L = (B f0 ). Let c0 be a closed curve containing x0 . We
want to calculate the holonomy h0 : Dom(h0 ) 0 0 of c0 in L,
where 0 is a suitable transverse containing x0 . For this purpose we choose
0 = (b0 F ).
Remark
4.1.
(Bf
0 ) : B f0 L is a cover map.
In fact observe that
(b, f0 ) = (d,
f0 ) {(gb, Q(g)f0 ) : g 1 (B)}
Q(g)f0 ) : g 1 (B)}
= {(g d,
{gb : g 1 (B)} = {g d : g 1 (B)}.

B is the universal cover of B then P (b) = P (d)
Hence if P : B proving
the result.
~ (b~1 ,f )
(b ,f )
~
Bx f
(b~1,f )
~ ,f )
(b ~
c
~
Bx f
~
~ b xF
b xF

lift of c to F x
x
x
L c
Fig. 4.5
By the previous remark we can consider the lift c0 of c0 in B f0 with

c0 (0) = (b0 , f0 ). We can write (t), f0 ). Dene (b1 , f0 ) =
c0 (t) = ( c0 (1).
Fix x 0 and
(b0 , f ) 1 (x).
The curve
c(t) = (
(t), f )
is the lift of c0 to the leaf Fx of F containing x. Hence
h0 (x) = (b1 , f ).
On the other hand, observe that b1 = (1) and It follows that
lies in B.
b1 = g0b0 for some g0 1 (B). Hence
h0 (x) = (b1 , f ) = (g0b0 , f ) = (b0 , Q(g01 )f ).

Since x = (b0 , f ) one has
h0 ( i) = ( i) Q(g01 ),
where i : F b0 F is the natural inclusion. Then,
h0 = (Q(g)),
where g = g01 and : Hol(L) Q(1 (B) is the map dened by
(Q(g)) = ( i) Q(g01 ) ( i)1 .
One can prove without diculty that is an isomorphism. This proves the
result.
4.2 Stability
In this section we consider F a foliation of class C 1 of a manifold M . We

introduce one of the main notions in the study of foliations.
Denition 4.2. A subset B M is stable (for F ) if for every neighborhood

W of B in M there exists a neighborhood W W of B in M such that
every leaf of F intersecting W is contained in W .
A classical problem in the theory of foliations is:
Problem 4.1. Find necessary and sucient conditions for B M to be

stable.
Exercise 4.2.1. Prove that if M is compact then all stable sets of F are
F -invariant.
Exercise 4.2.2. Prove that W in the denition of stable set can be assumed
to be invariant.
Example 4.6. Let w be the 1-form in M = R2 \ {(0, 0)} dened by w =

2 2
xdx + ydy. Clearly w = df where f (x, y) = x +y 2 and so w is exact. Hence
w is tangent to a foliation F whose leaves are concentric circles around
(0, 0). Clearly every leaf of F is stable, has innite fundamental group and
nite holonomy group.
Example 4.7. The compact leaf of the Reeb foliation in S 3 is not stable
and has innite holonomy group.
The above example shows the relation between stability and the nite-
ness of the holonomy group. This relation is the content of the following
result, a rst step towards Reeb local stability theorem (Theorem 4.3),
which is landmark in the theory of foliations.
Theorem 4.2. A compact leaf with nite holonomy group is stable.
Proof. Let F be a compact leaf of a C 1 foliation F with nite holonomy

group Hol(F ). We x x0 F a base point and a transverse 0 of F with
0 F = {x0 }. By assumption there are closed curves 1 , , k containing
x0 such that
Hol(F ) = {[f1 ], , [fk ]},
where fi : Dom(fi ) 0 0 is the holonomy of i and [] denotes the

class in the space of germs of at x0 .
Fix a nite cover U = {U1 , , Ur } of M by charts of F such that
(1) Ui F is a plaque i , i;
(2) Ui Uj = i j = , i, j;
(3) each plaque of Ui intersects at most one plaque of Uj , i, j.
The following notation will be useful:
i =space of plaques of Ui .
i : Ui i the plaque projection.
xi = i (i ). We can suppose that xi is a point of i .
In the case Ui Uj = ( i j = ) we let ci,j i j be a curve

joining xi and xj . Also in this case we let i,j : Dom(i,j ) i j be
the holonomy along the plaques. For all i = 2, , r we let ci F be a
curve joining x1 with xi . The curve ci induces a holonomy hi : Dom(hi )
1 i .
To prove the stability of the leaf F we x a neighborhood W of F .
Without loss of generality we can assume: r1 Ui W , 1 0 , x1 = x0
and D := k1 Dom(fi ) to be a neighborhood of x0 in 1 . The closed curves
of the form
cj,l = cl cl,j cj .
induce a holonomy map
hj,l : Dom(hj,l ) 1 1 .
Since Hol(F ) is the union of [fi ]s we have that hl,j = fp in a neighbor-
hood Dj,l 1 of x1 (= x0 the base point). Dene
D = j,l Dj,l .
Then D is an open neighborhood of x0 in 1 . Of course D D and
hj,l = fp in D . Let D D be a neighborhood of x1 such that y
D fi (y) D, i.
For every y D we dene
Cy = {j1 (hj (fi (y))) : 1 i k, 1 j r},
where h1 =Identity by denition.
Claim 4.1. If P is a plaque of Ul (for some l = 1, , r) and P L =

for some L Cy , then P Cy .
Proof of Claim 4.1: By hypothesis there is L = j1 (hj (fi (y))) such

that P L = . Because L Uj (as it is a plaque of Uj ) and P Ul we
have that Uj Ul = . This implies that cj,l is well dened (recall that this
is a curve joining xi with xj in i j ). By the denition of the holonomy
j,l we have
j,l (hj (fi (y))) = l (P ).
Hence
h1 1
l j,l hj (fi (y)) = hl (l (P )).
But, by denition, h1
l j,l hj is precisely the holonomy hj,l of the curve
cj,l . Hence
hj,l (fi (y)) = h1
l (l (P )).
Because hj,l = fp one has

fp fi (y) = h1
l (l (P )).
As Hol(F ) is a group one has fp fi = fi for some i . Hence

f,i (y) = h1l (l (P )) l1 (hl (fi (y))) = P,
proving P Cy as desired. This proves Claim 4.1.
Now, we let
Ly = {L : L Cy }.
We have the following properties: Ly Fy (=the leaf of F containing y);
Ly is open in Fy (since it is union of plaques of Fy ); Ly is close in M (this
is Claim 4.1); Ly = Fy (because Fy is connected); Ly W (because the
r
plaques forming Cy are contained in Ui W ). The last two properties
1
above imply that Fy W , y D . Dening W as the set of leaves
intersecting D we have that W W is a neighborhood of F such that
every leaf of F intersecting W is contained in W . Since W is arbitrary the
result follows.
Exercise 4.2.3. (Prove or give a counterexample) A compact invariant set

whose leaves have nite holonomy group is stable.
4.3 Reeb stability theorems
The statement of Theorem 4.2 can be improved as follows.

Theorem 4.3 (Reeb Local Stability Theorem). Let F be a compact
leaf with nite holonomy group of a C r foliation F in a manifold M . Then
for each neighborhood W of F there is a C r F -invariant tubular neighbor-
hood : W W F of F with the following properties:
(1) Every leaf F W is compact with nite holonomy group.
(2) If F W is a leaf then the restriction /F : F F is a nite cover
map.
(3) If x F then 1 (x) is a transverse of F .
Proof. Let W be a xed neighborhood of F . We can assume that W is

the domain of a C r tubular neighborhood 0 : W F . Because F is
compact we can further assume that the ber 01 (x) is transverse to F ,
x F . Let W W be given by Theorem 4.2. It follows from the proof
of Theorem 4.2 that all leaves F in W are compact (all of them have the
form F = Ly for some y D and Ly is compact). Dene = 0 /W .
Then : W F is a tubular neighborhood which is invariant and satises
(3). By shrinking W if necessary we can assume that F is transverse to
the ber 1 (x), x F . Since all leaf F W is compact we have that
F intersect each ber nitely many times. The same argument shows that
every leaf F W has nite holonomy group. This proves (1) and (2).
The theorem is proved.
Next we state two useful lemmas.
Lemma 4.1. Let Hom(R, 0) be the group of germs of homeomorphisms in R

xing 0 R. If G is a nite subgroup of Hom(R, 0) then G has at most two
elements. If all the elements of G are represented by orientation-preserving
maps, then G = {[Id]}.
Proof. Suppose that there is [f ] G {[Id]} represented by a local

orientation-preserving homeomorphism f xing 0. On one hand, there
are n0 N and a neighborhood U R of 0 such that f n0 (x) = x for all
x U because [f ] has nite order in G (as G is nite). On the other hand,
there is x0 U such that f (x0 ) = x0 because [f ] = [Id]. We can suppose
that 0 < x0 and that [0, x0 ] U without loss of generality. Because f is
orientation-preserving one has 0 < f n (x0 ) < f n1 (x0 ) < < f ( x0 ) < x0
for all n N Clearly f n (x0 ) [0, x0 ] for all n as [0, x0 ] U . The last ap-
plied to n = n0 yields f n0 (x0 ) = x0 and so x0 < x0 , a contradiction. This
contradiction shows that [f ] = [Id] for all element [f ] G represented by a
local orientation-preserving homeomorphism f xing 0. Let [g], [g ] G be
represented by orientation-reversing local homeomorphisms xing 0. Hence
[g] [g ]1 is represented by g (g )1 which is orientation-preserving. It
follows that [g] = [g ] and so there is only one element of G represented by
an orientation-reversing map. This proves that G has at most two elements
and the proof follows.
Lemma 4.2. Let F be a compact leaf of a codimension one foliation F

dened on a manifold M . Let Fn be a sequence of compact leaf of F accu-
mulating to a point in F . Then neighborhood W M one has Fn W
for all n large.
Proof. Let U1 , , Uk W be cover of F with charts of F such that

Ui F is a single plaque i of Ui , i. For each i we denote by i , the
space of plaques of Ui , and by i : Ui i the projection along the
plaques. Because Fn accumulates a point of F we can assume that Fn U1

contains a plaque arbitrarily close to 1 . From this we can assume that
Fn Ui = for all n, i. Clearly Fn Ui contains a nite number of plaques
as Fn is compact. Let P n,i and pn,i be the maximum and the minimum
of such plaques with respect to the natural order of i (=interval). Dene
Rn = ki=1 (P n,i pn,i ). Clearly Rn Fn is open in Fn (as it is union of
plaques). Let us prove that Rn is closed in Fn . In fact, x n and choose
a sequence xj Rn converging to x F . We can assume that all the xj s
are in a single plaque P n,i0 of Rn . Clearly x Ur for some 1 r k
by the denition of Rn . Hence the plaque r (x) Ur containing x is well
dened. Clearly P n,i0 Ur = and so P n,i0 r (x) = . Thus r (x) is a
plaque of Fn Ur . Since P n,i0 is the maximum of the plaques of Fn Ui0
one has that r (x) is the maximum of the plaques of Fn Ur . In other
words r (x) = P n,r proving x Rn . We conclude that Rn is closed in Fn .
Since Fn is connected we conclude that Rn = Fn . Since Rn i Ui W
we conclude that Fn W . The lemma is proved.
Remark 4.2. The conclusion of the lemma above is false for foliations of
codimension > 1.
For the case of codimension one foliations, the conclusion of the local
stability can be reinforced, by assuming that the compact leaf has nite
fundamental group. In this case, if the manifold is also compact, then all
leaves of the foliation are compact.
Theorem 4.4 (Reeb global stability theorem). Let F be a C 1 codi-

mension one foliation of a closed manifold M . If F contains a compact
leaf F with nite fundamental group then all the leaves of F are compact
with nite fundamental group. If F is transversely orientable then every leaf
of F is dieomorphic to F ; M is the total space of a bration f : M S 1
over S 1 with ber F ; and F is the ber foliation {f 1 () : S 1 }.
Proof. Denote by Fx the leaf of F containing x M and dene
M = {x M : Fx is compact with 1 nite}.
Note that by hypothesis W = . The Reeb local stability theorem implies

that M is open. We can assume that M is connected (otherwise we replace
it by a connected component). Let us prove that M is closed. For this
it suces to prove that M = . Suppose by contradiction that there
exists x0 M . Let U be a chart of F containing x0 , the space of
plaques of U and : U be the projection along the plaques. Note

that is an interval (as F has codimension one) and M U is union of
plaques of F . It follows that (M V ) is a countable family of open
intervals. Let J be one of these intervals and F (J) be the union of the
leaves of F intersecting J. Because J is open and contained in M we have
that F (J) is open in M . We claim that F (J) is closed in M . Indeed,
consider a sequence F (J) xn x M . Assume by contradiction that
x / F (J). Since x M the Reeb local stability theorem implies that there
is a neighborhood R of Fx such that every leaf intersecting R is compact
with 1 nite and contained in R. On one hand we can choose R such that
R J = . On the other hand we observe that Fxn R = for large n
because xn x. Hence Fxn R and also Fxn J = by the denition
of F (J). Hence R J = a contradiction. This contradiction shows that
x F (J), i.e., F (J) is closed. The claim follows. By connectedness we
conclude that F (J) = M . It follows that every leaf of F in M intersect J.
Since every leaf of F in M is compact we conclude that (M U ) is a nite
union of open intervals in . We conclude that x is a boundary point of one
of these intervals. Now we claim that the leaf Fx is closed. Otherwise the
above argument would imply that there exist a foliated chart U such that
M U contains innitely many connected components, a contradiction.
Because M is compact by assumption we conclude that Fx is compact.
Hence there is a tubular neighborhood P : W F of F whose bers
P 1 (y), y F are transverse to F . Because x M there is a sequence
of compact leaves Fn with nite 1 accumulating on x. By Lemma 4.2
we can assume that Fn W . The restriction P/Fn : Fn F is a nite
cover of F . Because Fn has nite fundamental group and P/Fn : Fn F
is a nite cover we conclude that F has nite 1 . We conclude that x M
contradicting x M . This contradiction proves that M = proving
that M is closed. By connectedness reasons we conclude that M = M .
Hence all leaves of F are compact with nite 1 .
Now suppose that F is transversely orientable. We already know that
each leaf L of F is compact with nite 1 . Being F transversely orientable
we have that the holonomy group of L is represented by an orientation-
preserving homomorphism. In other words the subgroup G = Hol(L) of
Di 0 (R, 0) is formed by orientation-preserving maps. Then Lemma 4.1
implies that F is a foliation without holonomy. It follows from the proof
of the Reeb local stability theorem that F is locally a product foliation,
i.e., each leaf L of F is equipped with an invariant product neighborhood
L I such that the leaves of F in this neighborhood have the form L ,
I, with L 0 corresponding to L. On the other hand, M is compact

by hypothesis. Hence there is a closed curve c transverse to F (to nd c we
simply use the non-wandering set of the transverse vector eld associated
to F ). Moreover c can be chosen to intersect some leaf L0 of F in a single
point. Observe that c intersect all leaves of F in a single point. In fact,
consider the set c = {x c : Fx c = {x}}. The set c is not empty by
the existence of L0 . The fact that F is a locally product foliation implies
that l is open and closed in c, therefore c = c and so c intersect each
leaf of F in one point at most. Now let F (c) be the set of points x M
such that Fx c = . It is clear that F (c) is open. One can prove that
F (c) is closed by using the Reeb local stability theorem as before. Hence
F (c) = M proving that all leaves of F intersect c. To dene the desired
bration f : M S 1 we simply dene f (x) to be the intersection point of
Fx with c. The theorem is proved.
Exercise 4.3.1. Prove that there is no codimension one foliation in the closed
3-ball B 3 having B 3 = S 3 as a leaf.
Exercise 4.3.2. (Prove or give a counterexample) Codimension one transi-

tive foliations have no compact leaves.
Exercise 4.3.3. Show that if G is a simply-connected Lie group and M is

a compact manifold of dimension dim M = 1 + dim G then for dim G 2
there is no locally free action of G in M .
Exercise 4.3.4. Let F be a transversely orientable codimension one foliation

on a closed orientable 3-manifold M . If there is a leaf F of F, whose
universal cover is not the real plane R2 then M = S 2 S 1 and F is the
product foliation S 2 . What about the case M is not orientable?
Exercise 4.3.5. Let be a C 2 integrable 1-form in a neighborhood of the

origin 0 Rn . We assume that the origin is a singularity of center type for

n
so that, up to a linear change of coordinates we have = d( 12 x2j )+(...)
j=1
where (...) means higher order terms. A classical result due to Reeb states
that for n 3 there is a neighborhood of the origin where all the leaves
of F : = 0 are dieomorphic to the (n 1)-sphere. This is proved as
follows:
(i) Consider the cylindrical blow-up of the origin given by the map : R
S n1 Rn , (t, x) = t.x. Show that {0} S n1 is a leaf of the lifted
foliation F = (F ) (hint: show that the 1-form = 1t () dened in
(R {0}) S n1 extends to R S n1 as = dt for t = 0 in class C 1 .

Also show that {0} S n1 is a leaf of () and so of .
(ii) For n 3 use the Reeb local stability theorem to conclude.
Exercise 4.3.6. Look (possibly in the literature) for a demonstration of the

following analytic version (also due to Reeb) of the above exercise: If is a
real analytic integrable 1-form in a neighborhood of the origin 0 Rn and
n 3. Suppose that the linear part of is non-degenerate and = df +(...)
for some quadratic analytic function f . Then there is a neighborhood of
the origin where F : = 0 is analytically conjugate to the linear foliation
df = 0.
4.4 Thurston stability theorem
In this section we prove the following theorem known as the Thurston sta-
bility theorem [Thurston (1974)].
Theorem 4.5. Let F be a C 1 transversely orientable codimension one foli-

ation of a compact manifold M tangent to the boundary of M if nonempty.
If F has a compact leaf L with trivial rst cohomology group (over the re-
als), then every leaf of F is homeomorphic to L and, furthermore, M is
homeomorphic either to L [0, 1] with F being the product foliation, or, to
the total space of a bration over S 1 having the leaves of F as bers.
The proof we give here is the one by Schachermayer [Schachermayer

(1978)] which in turns is a simplication of Reeb and Schweitzer [Reeb and
Schweitzer (1978)]. It is based on the elementary lemma below.
Lemma 4.3. Let > 0 and f, g, h : (, ) R be C 1 orientation-

preserving embeddings such that f (0) = g(0) = h(0) = 0 and f (0) =
g (0) = h (0) = 1. If (xn )nN+ (, ) is a sequence such that the limits
f (xn ) xn g(xn ) xn
lim =A and lim =B
n h(xn ) xn n h(xn ) xn
exist, then
f (g(xn )) xn f 1 (xn ) xn
lim = A+B and lim = A.
n h(xn ) xn n h(xn ) xn
Proof. Applying the Mean Value theorem we get
f (g(xn ))xn = f (xn )xn+f (g(xn ))f (xn ) = f (xn )xn +f (n )(g(xn )xn )
for some n in between g(xn ) and xn . Dividing by h(xn ) xn and letting

n we get the rst limit.
Again the Mean Value theorem implies
f 1 (xn ) xn = (f 1 (f (xn )) f 1 (xn )) = (f 1 ) (n ) (f (xn ) xn )
for some n in between f (xn ) and xn . Dividing by h(xn ) xn and letting
n we get the second limit.
Proof of Theorem 4.5. Suppose by contradiction that the conclusion of

the theorem does not hold. Then, by the proof of the Reeb global stability
theorem, the holonomy group Hol(L) of L is nontrivial, i.e., Hol(L) = {Id}.
Put H(L) = Hol(L). Let h : 1 (L) H(L) be the map assigning to a
loop s in L its holonomy map hs . Since F is C 1 , hs is dierentiable and
we denote by hs (0) the corresponding derivative at 0. This yields the map
h : 1 (L) R+ assigning to each s 1 (L) the number hs (0) which is pos-
itive since F is transversely orientable. The composition log h : (L) R
is clearly a group homomorphism. Since H 1 (L, R) = Hom(1 (L), R) where
the later is the group of homomorphism from 1 (L) to R, and H 1 (L, R) = 0
by hypothesis, we obtain that the homomorphism log h is trivial. Conse-
quently, hs (0) = 1 for every s 1 (L).
Let , , , , : (, ) R be C 1 orientation-preserving embed-
dings xing 0 whose germs , , , generates H(L). Because H(L) = 0,
there is a sequence (xn )nN+ (0, ) converging monotonically to 0 such
that
((xn ), (xn ), (xn ), , (xn )) = (0, 0, 0, , 0),
where T (x) = T (x) x for T {, , , , } and x (, ). Dene
n = max{|(xn )|, |(xn )|, |(xn )|, , |(xn )|}
and
NT = {n 1 : |T (xn )| = n } for T {, , , , }.
It follows that N = N N N N . Reordering the generators
+
if necessary we can assume 0 = |(xn )| = n for all n N+ . Since

(x ) (x )
n n
1, , 1
(xn ) (xn )
we can assume (again by passing to a subsequence if necessary) that the
limits
(xn ) (xn ) (xn )
a = lim , b = lim , , l = lim
n (xn ) n (xn ) n (xn )
exist. These limits are not all zero since a = 1.

On the other hand, since hs (0) = 1 for every s 1 (L), we have T (0) =
1 for all T {, , , , }. Then, by Lemma 4.3, the assignments
a, b, , l
extends uniquely to an additive homomorphism : H(L) R which is non-
trivial since a = 1. Then, the composition = h yields a homomorphism
: 1 (L) R which is nontrivial since h is onto and is nontrivial. It
follows that H 1 (L, R) = Hom(1 (L), R) = 0 a contradiction. This nishes
the proof.
Exercise 4.4.1. Let F be an orientable foliation with a compact leaf L F

homologous to zero. Prove that the Euler characteristic of L is zero.

Exercise 4.4.3. Show that if F is a codimension 1 smooth foliation of a

manifold M and L F is a compact leaf with Hom(1 (L), R) = 1 and
H 1 (L, R) = 0 then L has trivial holonomy.
Chapter 5
Haefligers theorem
5.1 Statement
A very nice example of the interplay between Dynamical Systems and

Topology in the Theory of foliations is given by Haeigers theorem. In
few words, it says that an analytic foliation of codimension one, cannot ex-
hibit a closed transverse section homotopic to zero in the ambient manifold.
The reason is that such a null-homotopic transverse curve will imply the
existence of a kind of limit cycle for the foliation.
Denition 5.1. Let F be a codimension one foliation on a manifold M . A

leaf F of F has one-sided holonomy if there are a closed curve c F and
point x0 c whose holonomy map f : Dom(f ) on a transverse
segment intersecting c satises the following properties:
(1) f is not the identity Id in any neighborhood of x0 in .

(2) f = Id in one of the two connected components of \ {x0 }.
The graph of f above may be as in Figure 4.3.
Example 5.1. A leaf with one-sided holonomy cannot be simply con-

nected. The torus ber of a torus bundle over S 1 is a non-simply connected
leaf without one-sided holonomy.
Example 5.2. The Reeb foliation in S 3 is an example of a codimension one

C foliation on a manifold with nite fundamental group with a one-sided
holonomy leaf.
Example 5.3. Real analytic codimension one foliations cannot have one-
sided holonomy leaves.
77
The main result of this section gives a sucient condition for the exis-
tence of one-sided holonomy leaves.
Theorem 5.1 (Haeigers theorem). A codimension one C 2 foliation

with a null-homotopic closed transversal has some leaf with one-sided holon-
omy.
Corollary 5.1. Codimension one C 2 foliations on compact manifolds with

nite fundamental group have one-sided holonomy leaves. In particular,
there are no real analytic codimension one foliations on manifolds with
nite fundamental group.
In fact, every codimension one foliation on a compact manifold has a

closed transverse. If the fundamental group of the manifold is nite then a
suitable power of this curve (as element of the fundamental group) yields
a null-homotopic closed transverse. Then Haeigers theorem applies. The
last conclusion of the above corollary applies to the following case:
Corollary 5.2. There is no real analytic codimension one foliation on S 3 .
The proof of Haeigers theorem is divided in three parts according to

the forthcoming sections.
5.2 Morse theory and foliations
First we recall some classical basic Morse Theory (cf. [Milnor (1963)]).
Let W be a compact 2-manifold with boundary W (possibly empty). Let
f : W R be a C r map r 2. A point p W is a critical point of f
if f (p) = 0. A critical point p is non-degenerated if the second derivative
f (p) is a non-degenerated quadratic form, where
2
(f x1 )(0)
f (p) =
xi xj 1i,j2
for some coordinate system (x1 , x2 ) around p = (0, 0). We shall use the
following lemma due to Morse.
Lemma 5.1 (Morse Lemma). Let p be a non-degenerated critical point

of a C r map f : W R, r 2. Then there is a coordinate system (x, y)
around p = (0, 0) such that one of the following alternatives hold:
(1) f (x, y) = f (0, 0) + x2 + y 2 .

(2) f (x, y) = f (0, 0) x2 y 2 .
Haefligers theorem 79
(3) f (x, y) = f (0, 0) + x2 y 2 .
The level curves of the three alternatives above are depicted in

Figure 5.1.
f(x,y)=f(0,0)+ x2+ y2
f(x,y)=f(0,0)+ x2- y 2
f(x,y)=f(0,0)-x 2-y 2
Fig. 5.1
Motivated by the above, we dene a Morse function as a C 2 map

f : W R, such that all of its critical points are non-degenerated. We
denote by C r (W, R) the set of all C r functions dened on W endowed with
the C r topology (see [Hirsch (1971)]) and by M r (W, R) C r (W, R) the
subset of Morse functions.
Remark 5.1. Notice that a Morse type critical point is isolated from the
set of critical points of a given function. In particular, a Morse function on
a compact manifold has only nitely many critical points.
The following is a classical result in Morse Theory and Dierential

Topology ([Hirsch (1971); Milnor (1963)]).
Theorem 5.2 (Theorem of Morse, [Hirsch (1971)],[Milnor (1963)]).

Given a manifold W , the set M r (W, R) is open and dense in C r (W, R).
Now we describe the foliated Morse Theory. We let M be a manifold

and F a codimension one foliation of class C 2 in M .
Denition 5.2. A C r , r 2 map f : W M is Morse with respect to F,
if for all p M there is a foliated chart U of F containing p such that
if : U R is the projection along the plaques then f M r (W, R).
Morse maps with respect to F are often said to be in general position with
respect to F . A critical point of f with respect to F is a critical point of
f for some (and therefore for every) foliated chart U .
Theorem 5.3. Let F be a codimension one foliation of class C 2 on a
manifold M . Let W be a compact 2-manifold and A : W M be a C r
map. Then there is a C r map f : W M arbitrarily C r close to A such
that
(1) f is Morse with respect to F.
(2) If p, p are dierent critical points of f with respect to F then f (p) and
f (p ) are in dierent leaves of F.
Proof. Fix an open cover Q1 , , Qk of A(W ) by foliated charts of F . We
can assume without loss of generality that the chart i : Qi Dn1 D1
has the form
i = (i1 , , in1 , i ),
where the last coordinate i : Qi R denotes the projection along the
plaques.
Dene Wi = A1 (Ui ) for all i. Hence W1 , , Wk is an open cover of
W . Fix Ui Ui Vi V i Wi such that U1 , Uk is an open cover of
W . For each i we x a C function i : W R satisfying: i [0, 1],
i = 1 in Ui and i = 0 in W \ Vi . See Figure 5.2
Fix > 0 and denote by dr the C r topology in C r (W, M ). We dene
inductively a sequence g0 , g1 , g2 , , gk : W M as follows. First we
dene g0 = A. For a suitable Morse function f1 C r (U1 , R) we dene
g1 (x) as follows:

g1 (x) = 11 (x), , 1n1 (x), 1 (x)f1 (x) + (1 + 1 (x))(1 g0 )(x)
if x W1 and
g1 (x) = g0 (x)
if x W \ W1 .
It follows that
dr (g1 , g0 ) dr (g1 /W1 , g0 /W1 )(rj=1 Kj dj (1 , Id/W1 )) dr (f1 , 1 g0 ),
1111111
1111111
1111111
1111111 graph of i
1111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
Wi
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
Vi

111111111111111111111111111
111111111111111111111111111
Ui
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
111111111111111111111111111
Fig. 5.2
where the constants Kj does not depend on dr (f1 , 1 g0 ). As 1 is xed

we have that dr (1 , Id/W1 ) does not depend on dr (f1 , 1 g0 ). Hence by
Morse Theorem we can choose f1 so that
dr (g1 , g0 ) < /k.
Summarizing g1 satises the following properties:
dr (g1 , g0 ) < /k;

g1 /U1 is Morse (because g1 /U1 = f1 ).
Replacing g0 by g1 in the above construction we can nd g2 such that
dr (g2 , g1 ) < /k;

g2/(U1 U2 ) is Morse.
Repeating the argument we can nd the sequence g0 , g1 , , gk . An

element gi of this satises:
dr (gi , gi1 ) < /k;

gi /(U1 Ui ) is Morse.
The last map gk of the sequence is Morse (because U1 , , Uk is a cover

of W ). Moreover,
dC r (gk , A) ki=0 dC r (gi , gi+1 ) < (/k) k = .

perturb here
F
M M
W W
graph of g k graph of f
Fig. 5.3
Hence the last map gk : W M is Morse respect to F and -close to A in

the C r topology. It remains to choose f close to gk satisfying the property
(2) of the theorem. To modify gk to obtain f satisfying (1) and (2).
Because the set of Morse function is open we we only have to approx-
imate gk by a map satisfying the property (2). The last can be attained
as in the trivial case described in Figure 5.3 where M is the product foli-
ation W I and F is the trivial foliation I. Indeed we only have to
perturb around a critical value as indicated in the gure. The theorem is
proved.
5.3 Vector elds on the two-disc
We denote by D2 the 2-dimensional disc in R2 and by X1 (D2 ) the set of

C 1 vector elds in D2 transverse to the boundary D2 of D2 . The closure
of B D2 is denoted by B. The orbit of x D2 is denoted by O(x).
Consider p D2 , Y X1 (D2 ) and denote by (p) the -limit set of p. Note
that if is an interval transverse to Y and p D2 is regular, then (p)
intersects at most once. In particular, a periodic orbit of Y intersects
once. These facts follow from the trivial topology of the disc D2 . A
singularity of Y is called saddle or center according to the two portrait face
corresponding to Figure 5.1 (left-hand one for center and right-hand one
for saddle). A saddle singularity exhibits two stable separatrices and two
unstable separatrices. A graph of Y is a connected set formed by saddles

and saddles separatrices in a way that if s is a saddle then contains
at least one stable separatrix and one unstable separatrix of s.
Theorem 5.4. Let Y X1 (D2 ) be such that Y is transverse to D2 and

sing(Y ) is a nite set formed by centers and saddles. Suppose that Y has
no saddle-connections. Then, there is x D2 such that:
1) O(x) is a closed curve.
2) There is an interval transverse to Y with the following properties:
2.1) O(x) = O(x) = {x}.
2.2) The rst return map f : Dom(f ) induced by Y in
satises that: f = Id in a connected component of \ {x} and f = Id
in any neighborhood of x in .
Proof. Because Y has no saddle-connection the graphs of Y are as in

Figure 5.4. Clearly the complement D2 \ of a compact invariant set
equals to either a periodic orbit or a graph contains at least one connected
components disjoint from D2 . The union of such connected components
will be denoted by R(). We dene an order < on the set formed by
periodic orbits and graphs of Y by setting:
1 < 2 R(1 ) R(2 ).
A limit cycle of Y will a compact invariant set L with regular orbits of Y
equals to (p) for some p / L. It is easy to prove that a limit cycle L is
either a periodic orbit or a graph. Hence the order < is well dened on the
set of limit cycles of Y .
Lemma 5.2. If 1 > 2 >

. . . is a decreasing sequence of limit cycles of

Y , then = R(n ) is either a periodic orbit or a graph of Y .
n=1
Proof. Since Y is transverse to D2 we can assume that Y points inward

on the boundary D2 of the disc D2 . Clearly, Y has nitely many graphs
as it has nitely many singularities. Hence we can assume that n is a
periodic orbit, n. So, R(n ) is a disc and R(n ) = n , n. There is at
least one regular point in because if s sing(Y ), then s must be
saddle and so at least one of the separatrices of s is accumulated by n .
First we observe that cannot contain periodic orbits unless it is a
periodic orbit. Indeed this follows from Lemma 4.2 but we give a direct

x*

R(
x*
periodic orbit type I

x*
x*
type III
type II
Fig. 5.4
proof here. Suppose that contains a periodic orbit . Pick x and

let x be a transverse of Y containing x. Clearly for all n N large the
set n x consists of a single point xn such that xn x as n . Let
An be the annulus in D2 formed by n and . Since xn x the Tubular
Flow-Box theorem implies that n in the Hausdor topology proving
= as desired.
Second we observe that cannot contain a graph unless it is a graph.
The proof of this is similar to the previous proof. Indeed, let be
a graph. Hence is one of the three types of graphs in Figure 5.4. If
is type II or III then previous argument shows that = and we are
done. Otherwise is type I. In this case does not contain an unstable
separatrix a (say). By the Poincare Bendixson Theorem we have that (a)
is either a singularity or a graph or a periodic orbit. In the later case we
have that is a periodic orbit, a contradiction because it contains the
graph . In the former case we have that is one of the graphs in the
bottom gures (Y has no saddle connections) again a contradiction. Hence
(p ) must be a graph which we still denoted by . cannot be of type
I for, otherwise, the limit cycle sequence n must be contained in R() a

contradiction since n is decreasing. This proves that is type II or III
and we are done.
Now, x p regular. Clearly is invariant and so (p) .
Poincare-Bendixson implies that (p) is either a singularity or a periodic
orbit or a graph. In the last two cases we have that contains either a
periodic orbit or a graph. Hence is either a periodic orbit or a graph
and we are done. We conclude that (p) is a singularity. Analogously (p)
can be assumed to be a singularity Y because it points inward D2 from
D2 . Because Y has no saddle connections we have that the closure O(p)
of the orbit O(P ) is a graph of Y . This graph is evidently contained in .
We conclude that contains a graph and we are done. This proves the
lemma.
Let us nish the proof of Theorem 5.4. Consider the set R of all compact
invariant sets of Y of the form
!
= R(n ) ,
n=0
for some decreasing sequence of limit cycles n of Y . Lemma 5.2 implies
that the order < is dened in R. Lemma 5.2 also implies that any decreasing
sequence 1 > 2 > in R has an inmum in R. The Zorn Lemma
implies that there is a minimal element = in R. By Lemma 5.2 we
have that is either a periodic orbit or a graph. In any case we choose
x as indicated in Figure 5.4. Observe that the closure O(x ) of
the orbit O(x ) is a closed curve. Choose a transverse containing x
as indicated in the gure. Let f : Dom(f ) be the return map
induced by Y in . Because is accumulated by limit cycles of Y one
has that f = Id in any neighborhood of x in (for such limit cycles must
intersect ). On the other hand, consider the connected component c of
{x } contained in . Because the number of graphs of Y is nite, we
can assume by shrinking if necessary that does not intersect any graph
of Y . In particular c does not intersect any graph of Y . Because c R( )
we conclude that the orbit of any point in c is periodic. Hence f = Id in c.
This proves that x = x satises the properties (1),(2) of the theorem.
5.4 Proof of Haeigers theorem
Let F be a codimension one C 2 foliation with a null-homotopic closed

transverse . Since is null-homotopic, there is a C map A : D2 M
such that A((D2 )) = . By Theorem 5.3 we can assume that A = f is in

general position with respect to F . Let F = f (F ) be the foliation in D2
induced by f . By denition a leaf L of F is f 1 (connected component of
F f (D2 )) for some leaf F of F . Note that F is a singular foliation of
class C 2 in D2 and a singularity of F is either a center or a saddle. Clearly
F is C 2 orientable close to the singularities. Far from the singularities we
have that F is C 2 locally orientable by the Tubular Flow-Box theorem.
We conclude that F is C 2 locally orientable. By the results in Section 2.3
we have that F is C 2 orientable, i.e., there is a C 2 vector eld Y in D2
tangent to F . Note that Y is transverse to D2 and contains a nite
number of singularities all of them being centers or saddles. Moreover, Y
has no saddle connections by Theorem 5.3-(2). It follows from Theorem 5.4
that there is x D2 and a transverse satisfying the conclusions (1)-(2)
in that theorem. In particular, if x0 = f (x) then c = f (O(x)) is a closed
curve contained in the leaf F = Fx0 of F and = f () is a transverse
segment of F intersecting c. The holonomy of c is conjugated to the return
map f : Dom(f ) induced by Y in . One can see that c and F
satisfy the properties (1)-(2) of Denition 5.1 by using the property (2.2)
in Theorem 5.4. We conclude that F is a leaf with one-sided holonomy of
F proving the theorem.
Exercise 5.4.1. Let F be a codimension one foliation on S 3 with a compact

leaf L F homologous to zero. Show that L is the torus.
Chapter 6
Novikovs compact leaf theorem
6.1 Statement
The search for codimension one foliations on the three sphere S 3 was ended
by the introduction of the Reeb foliation described in Example 1.3. Never-
theless, this example exhibits some particularities, suggesting some of them
are part of a global phenomena. One of them is the existence of a torus leaf.
The other is the existence of a one-sided holonomy, though this was proved
in Haeigers theorem. Regarding the existence of a compact (torus) leaf,
this also part of a global phenomena. Indeed, in this chapter we shall prove
the celebrated Novikovs compact leaf theorem.
Theorem 6.1 (Novikovs compact leaf theorem). Codimension one

C 2 foliations on compact 3-manifolds with nite fundamental group have
compact leaves.
The proof of this theorem given here is the one in [Haeiger (1967)]
Seminaire Bourbaki 20e annee, 1967-68, Num. 339, p. 433-444. That
proof is based on the following denition.
Denition 6.1 (vanishing cycle). Let F be a C 1 codimension one folia-

tion in a manifold M . A vanishing cycle of F is a C 1 map f : S 1 [0, ] M
(for some > 0) such that if we denote ft (x) = f x (t) = f (x, t),
(x, t) S 1 [0, ], then the following properties hold:
(1) ft (S 1 ) is a closed curve contained in a leaf A(t) of F, t;

(2) ft (S 1 ) is null homotopic in A(t) if and only if t > 0;
(3) f x ([0, ]) is transverse to F , x.
87
Example 6.1. Let F be the Reeb foliation in S 3 and let T be the compact
leaf of F . Any generator of (T ) is represented by a curve contained in
(the image of) a vanishing cycle of F .
Example 6.2. A torus bration over S 1 gives an example of a foliation

with compact leaves having no vanishing cycles.
The proof of Novikovs compact leaf theorem is a direct consequence of

the following two preliminary results.
Theorem 6.2 (Auxiliary theorem I). Codimension one C 2 foliations

on compact 3-manifolds with nite fundamental group have vanishing cy-
cles.
Theorem 6.3 (Auxiliary theorem II). Codimension one C 1 trans-

versely orientable foliations with vanishing cycles on compact 3-manifolds
have compact leaves.
Proof of Novikovs compact leaf theorem. Let F be a codimension

one C 2 foliation on a compact 3-manifold M . Let P : M M be a nite
cover of M such that the lift F of F is transversely orientable. 1 (M ) is
nite since 1 (M ) < 1 (M ) and 1 (M ) is nite. By Auxiliary theorem I
we have that F has a vanishing cycle. Hence F has a compact leaf F by
Auxiliary theorem II. Then F = P (F ) is a compact leaf of F proving the
result.
6.2 Proof of Auxiliary theorem I
Let F be a codimension one C 2 foliation on a compact 3-manifold M . It

is easy to prove that F has a closed transverse . In fact, by passing to a
nite cover we can assume that F is transversely orientable, and so, it has
a transverse vector eld X. Because M is compact we have that X has a
non-wandering point x. Hence there is a piece of orbit of X which starts
and nishes close to x. By modifying a bit such a piece of orbit nearby x
we can construct a closed transverse of F containing x. This proves the
result.
Next we assume that 1 (M ) is nite and let be a closed transverse
of F . Because 1 (M ) is nite we have that there is n N such that the
curve n represent a closed null homotopic transverse of F . Without loss
of generality we can assume that n = 1. Now we proceed as in the proof of
Haeigers theorem: Because is null-homotopic one has that there is a C
Novikovs compact leaf 89
map A : D2 M such that A((D2 )) = . By Theorem 5.3 we can assume

that A = f is in general position with respect to F . Let F = f (F ) be the
foliation in D2 induced by f . Note that F is a singular foliation of class C 2
in D2 and a singularity of F is either a center or a saddle. Clearly F is
C 2 orientable close to the singularities. Far from the singularities we have
that F is C 2 locally orientable by the Tubular Flow-Box Theorem. Hence
F is C 2 locally orientable. By the last example of Section 2.2 we have
that F is C 2 orientable, i.e., there is a C 2 vector eld Y in D2 tangent
to F . Note that Y is transverse to D2 and contains a nite number
of singularities all of them being centers or saddles. Moreover, Y has no
saddle connections by Theorem 5.3-(2). It follows from Theorem 5.4 that
there is x D 2 and a transverse satisfying the conclusions (1)-(2) in that
theorem. Let c0 be the closed curve c0 = f (O(x)). Then c0 Fx0 where
x0 = f (x). Note that c0 is not null-homotopic in Fx0 because its holonomy
map is not the identity in any neighborhood of x0 . Moreover, the closed
curve = O(x) is either a periodic orbit or the closure of a homoclinic loop
of Y . These properties motivate us to dene cycle as a closed curve in
D2 which is either a periodic orbit or the closure of a homoclinic loop of
Y such that f () is not null homotopic in the leaf of F containing it. As
before every cycle of Y bounds a region R() which does not intersect
D 2 . We dene the order < in the set of cycles of Y by setting 1 < 2 if
and only if R(1 ) R(2 ).
Lemma 6.1. Let 1 > 2 > be a decreasing sequence of cycles of Y .

Then there is a cycle of Y such that n > for all n N.
Proof. Because the number of homoclinic loops of Y is nite we can assume

that n is a periodic orbit and that R(n ) is a disc with boundary n for
all n. The sequence R(n ) is a decreasing sequence of compact sets in D2 .
Hence

R(n )
n=1
is a non-empty compact set whose boundary will be denote by . It is clear

that = . Moreover, there is p regular because the singularities of Y
are centers or saddles (no periodic orbit close to a center of Y can be a cycle
of Y ). By Poincare-Bendixson we have that (p) is either a singularity or
a periodic orbit or a graph. In the later two cases we have that contains
either a periodic orbit or a graph. Hence, as in the proof of Lemma 5.2 in
Section 5.3, we have that itself is either a periodic orbit or a graph of
type II or III (see Figure 5.4). If were a periodic orbit with f () null
homotopic in its leaf then f (n ) would be null homotopic in its leaf for all
n large, a contradiction. Hence if is periodic then = is the desired
cycle. Now suppose that is a graph of type II or III. The fact that is
surrounded by cycles of Y implies that f () is not null homotopic in its
leaf. Hence one of the two homoclinic loops forming (say ) satises
that f ( ) is not null homotopic in its leaf. Then, = is the desired
cycle. To nish we assume that (p) is a singularity. In a similar way we
can assume that (p) is a singularity. Hence O(p) is a graph contained in
. As previously remarked this implies the existence of the desired cycle
and the proof follows.
Let us nish the proof of Auxiliary theorem I. By the previous lemma

and the Zorn lemma we have that there is a cycle of Y which is minimal
for the order <. By the denition of cycle we have that f ( ) is not null
homotopic in its leaf. Choose a regular point x and let be a
transverse of F inside R( ) containing x in its boundary. In order to
simplify the notation we shall assume that = [0, 1] with x 0. By
Poincare-Bendixson, because the number of graphs of Y is nite, we have
that all orbit of Y starting at y \ {x} is periodic with period ty . Hence
the set
A = O(x) {Y[0,ty ] : y \ {x}}
is dieomorphic to the annulus S 1 [0, 1]. Put a parametrization P :

S 1 [0, 1] A such that: P0 (S 1 ) = O(x); Py (S 1 ) = Y[0,ty ] for all y \{x};
and P ([0, 1]) is transverse to F for all S 1 . Consider the map g = f P .
Hence g : S 1 [0, 1] M is a limit cycle of F . In fact, g is C 1 since both
P and f are. In addition, g0 (S 1 ) = f (O(x)) Ff (x) is not null homotopic
in A(0) = Ff (x) because is a cycle of Y . Also for y = 0 we have that
gy (S 1 ) = f (a(y)) is null homotopic in its leaf since is minimal with
respect to the order <. Because P ([0, 1]) is transverse to F for all S 1
one has that g (S 1 ) = f (P ([0, 1])) is transverse to F for all S 1 . This
proves the theorem.
6.3 Proof of Auxiliary theorem II
The proof uses the following holonomy lemma. First, we use introduce
short denitions. Given a codimension one foliation F we say that a vector
eld X is normal for F if the trajectories of X are everywhere transverse to
F . Clearly a codimension one foliation has a normal vector eld if and only
if it is transversely orientable. If F is a transversely orientable foliation in
a manifold M we say that a curve c M is normal to F if c is transverse
to F and contained in a solution curve of the normal vector eld associated
to F . Given a leaf A of F , a compact set K and a C 1 map g : K A
we say that g has a normal extension if there are > 0 and a C 1 map
G : K [0, ]) M such that:
(1) G0 /K = g;
(2) Gt (K) A(t) for some leaf A(t) of F with A(0) = A;
(3) x K the curve Gx ([0, ]) is normal to F .
Lemma 6.2 (Holonomy Lemma). Let F a codimension one trans-

versely orientable foliation, A be a leaf of F and K be compact set. If
g : K A is a C 1 map homotopic to constant in A, then g has a normal
extension.
Proof. Let g : K A of as in the statement and denote by X the normal

vector eld of F . For all x K dene the normal curve
x = {Xs (g(x)) : s [0, 1]},
where as usual Xt denotes the ow of X and XB (A) = {Xs (z) : (s, z)
B A}. By hypothesis g(K) is a compact null homotopic subset of A, and
so, it is contained is disc D A. Fix x0 K and dene
G(x, t) = f (Xt (g(x)),
where f = fx : Dom(f ) x0 x is the holonomy induced by a curve
x D joining g(x0 ) to g(x). Since D is contractible we have G(x, t)
does not depend on the chosen curve x . Moreover, since g(K) is compact,
we can assume that there is > 0 such that G(x, t) is dened for every
(x, t) K [0, ]. Let us prove that the map G : K [0, ] M so obtained
is a normal extension of g. It is clear that G is C 1 since g is. First we prove
that G0 /K = g. In fact, if x K then G0 (x) = G(x, 0) = f (X0 (g(x))) =
f (g(x)) = g(x) by the denition of holonomy and X0 = Id. Second we prove
that Gt (K) A(t) for some leaf A(t) of F with A(0) = A. In fact, it is clear
that A(0) = A. Next we observe that Gt (x) = G(x, t) = f (Xt (g(x)))
FXt (g(x) . Since FXt (g(x)) = FXt (g(x0 )) by denition of holonomy we have
that A(t) := FXt (g(x0 ) works. Third we prove that Gx ([0, ]) is normal to
F. In fact, Gx (t) = G(x, t) = f (Xt (g(x))) x which is a solution curve
of X. The lemma follows.
Hereafter we let F be a codimension one C 1 transversely orientable

foliation with a vanishing cycle f : S 1 [0, ] M on a compact 3-manifold
M . We denote by X the vector eld in M transverse to F and by Xt the
ow generated by X. This vector eld exists because F is transversely
orientable. For simplicity we shall assume that = 1. A C 1 curve in a
leaf of F is in general position whenever #1 (p) 2 for all p M and if
x, y Dom() are dierent points with (x) = (y), then (x) and (y)
are not parallel (see Figure 6.1).
leaf of F
Fig. 6.1 General position curve for F .
Lemma 6.3. We can suppose that the vanishing cycle f : S 1 [0, 1] M

of F satises the following additional properties:
(4) f0 (S 1 ) is in general position in A(0).

(5) f x ([0, 1]) is normal to F .
(6) If x, y S 1 and f x (0) = f y (0), then f x ([0, 1]) f y ([0, 1]) = .
Proof. Clearly (5) implies (6) by standard arguments from Ordinary dif-
ferential equations. By moving a bit fx (S 1 ) we can assume (4). To assume
(5) it suces to project f x ([0, 1]) to the solution curve of X passing through
f x (0) via holonomy. This is done as follows (see Figure 6.2): For x S 1 we
dene = {Xs (f x (0)); s [0, 1]} which is the solution curve of X passing
through f x (0). We dene = f x ([0, 1]). By the denition of vanishing

cycle the curve one has F . As X F , we have F . Note that
x . Hence there is a holonomy map g : Dom(g) .
Dene f (x, t) = g(f x (t)). By compactness we can assume that f (x, t)
is dened in S 1 [0, ] for some > 0. We shall assume that = 1 for
simplicity. Let us prove that f is a vanishing cycle of F . First note that
(f )0 (x) = g(f x (0)) = g(x) = x (by the denition of g) and so (f )0 (S 1 ) =
f0 (S 1 ). The last implies that f0 (S 1 ) A(0) is not null-homotopic in A(0).
Moreover, (f )t (x) = f (x, t) = g(f x (t)) Ff x (t) = A(t) by the def-
inition of holonomy. Hence ft (S 1 ) A(t) , t. Now we prove that
Ft (S 1 ) is null homotopic in A(t) for all t > 0 small. In fact note that
ft (x) = g(f x (t)) + g(f x (0)) = f0 (x) and ft (x) + f0 (x). Hence
t0 t0
d(ft (x), ft (x) d(ft (x), f0 (x)) + d(f0 (x), ft (x)) + 0. Thus, ft (S 1 ) is
t0
C 0 -close to ft (S 1 ) in A(t). Then, ft (S 1 ) 0 em A(t) as desired. The proof
follows.
Original vanishing cycle Normal curve

holonomy
f x(0)

Fig. 6.2
Lemma 6.4. We can assume that the vanishing cycle f : S 1 [0, 1] M

of F satises the following additional property:
(7) The lift ft : S 1 A(t) of ft : S 1 A(t) to the universal cover

(t) : A(t) A(t) of A(t) is a simple closed curve, t > 0.
Proof. Dene R = {couples (x, y) of dierent points of S 1 such that

ft (x) = ft (y) for some t > 0}, and rt = #R. We have that r < . In fact,
for all t we dene Bt = {p ft (S 1 ) : #f 1 (p) = 2}, bt = #Bt . b0 <
because f0 is in general position. Note that ft f0 in the C 1 topology as
t 0+ and, because f0 is in general position we have bt = b0 for all t 0.
Hence the map t bt is the constant map t b0 . As r b0 the result fol-
lows. On the other hand, R = for otherwise we are done. Pick (x, y) R
and dene U = {t > 0 : ft (x) = ft (y)}, K = {t 0 : ft (x) = ft (y)}. The
following properties hold:
K is compact (because f x and f y are continuous).

U = (because (x, y) R).
U K (because ft (x) = ft (y) f (t(x) = ft (y)).
U is open: Fix t such that ft (x) = ft (y). Then the curve ((t) ft )/[x, y]
is closed, where [x, y] is a suitable arc in S 1 joining x, y. This curve
is null homotopic in A(t) because its lift ft /[x, y] in A(t) is a closed
curve. By the Holonomy Lemma we conclude that ((s) fs )/[x, y] is
null homotopic in A(s) for s t proving the result.
Because U = is open we can x an interval (t , t ] in U with t /U

and t being arbitrarily close to 0. We also have that t K, because K
is closed and U K. Hence ft (x) = ft (y). We claim that one of the two
arcs [x, y] joining x to y in S 1 satises that the closed curve ft ([x, y]) is not
null homotopic in A(t ). Indeed we have two cases, namely either t > 0 or
t = 0. If t > 0 as t / U we have that ft ([x, y]) is not a closed curve for
each arc [x, y]. Because the closed curve ft ([x, y]) lifts to the non-closed
curve ft ([x, y]) in A(t ) we conclude that ft ([x, y]) is not null homotopic in
A(t) for all [x, y] (recall that A(t ) is the universal cover of A(t )). If t = 0
then we let [x, y], [x, y] be the two possible arcs in S 1 joining x to y. If
both f0 ([x, y]), f0 ([x, y] ) were null homotopic in A(0) then f0 (S 1 ) would be
null homotopic in A(0) as it is the product of f0 ([x, y]), f0 ([x, y] ). The last
contradicts the denition of vanishing cycle. This proves the claim. Hence
we can assume that ft ([x, y]) is not null-homotopic in A(t ). We note that
the closed curve ft ([x, y]) has less than r multiple points. Moreover the
restriction f /(S 1 [t , t ]) : S 1 [t , t ] M is a vanishing cycle of F

with A(t ) close to A(0). Replacing f by f /(S 1 [t , t ]) we have less than
r multiple points for the new vanishing cycle. Repeating the process we
obtain the result.
Lemma 6.5. Let f : S 1 [0, 1] M be a vanishing cycle of F satisfying

the properties (4)-(6) of Lemma 6.3 and (7) of Lemma 6.4. Then, there is
an immersion F : D2 (0, 1] M satisfying the following properties:

(i) Ft D2 = ft , t.
(ii) F (D2 t) A(t), t.
(iii) F x ((0, 1]) is normal to F , x.
(iv) If U = {x D2 : limt0+ F x (t) exists}, then D2 U ; U is open; and
D2 \ U = .
Proof. Let (t) : A(t) A(t) be the universal cover of A(t). We have that
A(t) = R2 or S 2 . The last cannot happen for otherwise the Reeb global
stability theorem would imply M = S 2 S 1 and F is the trivial foliation
S 2 , a contradiction since f0 (S 1 ) is not null homotopic in A(0) (see the
last exercise in Section 4.3). On the other hand, (7) of Lemma 6.4 says that
f1 (S 1 ) is a simple closed curve in A(1). By the classical Jordan Theorem
we have that there is an embedding F : D2 1 A(1) with F1 /D2 = f1 .
We dene F : D2 1 A(1) by F = (1) F . Clearly F is an immersion
as (1) is a cover and F is an embedding. Applying the Holonomy Lemma
to F we can extend F to D2 (t0 , 1] for some t0 > 0 satisfying (1)-(3).
We claim that F can be extended to D2 [t0 , 1] still satisfying (1)-(3).
In fact, rst we show that limtt+ F (x, t) exists for all x D2 . Because
0
t0 > 0 we have that ft0 : S 1 At0 is null homotopic. As before there is an

embedding Ht : D2 A(t0 ) with Ht /D2 = ft . Dene Ht = (t0 )
0 0 0 0
Ht0 . Again by the Holonomy Lemma there are > 0 and an immersion
G : D2 (t0 , t0 + ) M such that:
a) Gt (D2 ) A(t);
b) Gt /D2 = ft ;
c) Gx ((t0 , t0 + )) is normal to F ;
d) Gt0 = Ht0 .
Now we x t (t0 , t0 ) and consider D := Ft (D2 ) and D1 := Gt (D2 ).
Both D and D1 are discs contained in A(t) with D = D1 = ft (S 1 ). If
D = D1 then A(t) would be S 2 a contradiction as before by Reed Stability.
Hence D = D1 and so Ft (D2 ) = Gt (D2 ) for all t (t0 , t0 ). If x D2 and
t (t0 , t0 ), then Ft (x) Ft (D2 ) = Gt (D2 ) Ft (x) = Gt (y(x, t)) for

some y(x, t) D 2 . But Gy(x,t) (t0 , t0 +) and F x ((t0 , 1]) are both normal
to F . Hence y(x, t) = y(x) does not depend on t. Thus, lim F (x, t) =
tt+
0
lim+ G(y(x), t) = G(y(x), t0 ) A(t0 ) proving that lim+ F (x, t) exists x
tt0 tt0
D2 . To nish the proof of the claim we simply dene H : D2 [t0 , 1] M
by

lim F (x, t), if t = 0,
+
H(x, t) = tt0

F (x, t), if t = t0 .
Thus H/D2 (t0 , 1] = F , Ht |D2 = ft , Ht (D2 ) A(t), t and H x ([t0 , 1])

is normal to F . In other words H is an extension of F to D2 [t0 , 1]
satisfying (1)-(3). This proves the claim. If t0 > 0 the Holonomy Lemma
allow us to extend F to D2 (t0 , 1] satisfying (1)-(3). Hence we can
assume that there is F : D2 (0, 1] M satisfying (1)-(3).
Let us prove that F satises the property (4) of the lemma. If x0
U y0 = lim F (x0 , t). Let V be a tubular ow-box for X around
t0+
y0 solution curve of X. Note that y0 OX (x0 ), the orbit of x0 , as X
is non-singular (X F ). Hence, Xt0 (x0 ) = y0 , for some t0 > 0. By the
Tubular Flow-Box Theorem there is a neighborhood B of x0 , such that
Xt0 (B) V . As F1 is continuous there is a neighborhood W of x1 in D2
with x0 = F1 (x1 ) such that F1 (W ) B. See Figure 6.3.
Let us prove that lim+ F (x, t) exists x W . In fact, Consider x
t0

W and x = F1 (x). Note that the curve F x ((0, 1]) has nite length for,

otherwise, it would exist a rst exit point z of F x ((0, 1]) from V . Clearly

z = F x (tz ) for some tz [0, 1]. But F x (tz ) = Ftz (x) A(tz ), F x (tz ) =
Ftz (x ) A(tz ). As z V and F x (tz ) / V (such a point is close to
y Int(V )) we conclude by the Mean Value Theorem that A(tz ) and Fy
have an intersection point. This intersection point implies A(tz ) = Fy .
Because dim F = 2 we can assume from the beginning that Fy = A(tz ) a

contradiction. Hence F x ((0, 1]) has nite length lim F (x , t), x
t0+
W . This proves that U is open.
Warning: The last argument proves that if x U and y = limt0+ F x (t),

then there is a neighborhood W of x in D 2 such that limt0+ F x (t) = y
exists for all x W and Fy = Fy .
To see D2 U it suces to observe that F x (t) = f x (t) for all x
Intersection point
V x
z =F (t )
z
A(tz )
x
F ((0,1]) y
x
F (t z)
x0
B
F1
Fy
x W D2
x
Fig. 6.3
D2 = S 1 limt0+ F x (t) = limt0+ f x (t) exists and belongs to A(0) for

all x S 1 .
Finally we prove that D2 \ U = . If D2 = U we would have that
F0 := limt0+ F x (t) would exist for all x D2 . By the Warning above we
would have F0 (x) A(0) for all x D2 . The resulting map F0 : D2 A(0)
yields a continuous extension of f0 to D2 , a contradiction since f0 (S 1 ) is
not null homotopic in A(0). This contradiction shows D2 \ U = and the
lemma follows.
Lemma 6.6. Let F be the immersion in Lemma 6.5. Then, > 0 there
are 0 < t < t < and an embedding h : D2 int(D2 ) such that
F (t , x) = F (t , x), x D2 .
Proof. By Lemma 6.5 there is y0 D2 \ U . Hence, the limit

limt0+ F y0 (t) does not exist. Nevertheless the compactness of M implies
that there is a sequence tn such that F y0 (tn ) z for some z M . By
using the Tubular Flow-Box Theorem we can assume that F y0 Fz for all
n. In addition we can further assume Fz = A(0) because the leaves of F are

two-dimensional. As F y0 (tn ) A(tn ) we have A(tn ) = A(tm ) for all n, m.
Dening D(t) = Ft (D2 ) we have D(tn ) A(tn ) Fz , i.e., D(tn ) Fz for
all n. Note that z A(tn ) n large for, otherwise, it would exist nk
with z / D(tnk ). By hypothesis F y0 (tnk ) = Ftnk (y0 ) D(tnk ) converges to
z. Because D(t) = ft (S 1 ) for all t we conclude that bnk ftnk sequence
converging to z. But the distance dist(bnk , f0 (S 1 )) goes to 0 as k . As
f0 (S 1 ) is compact and bnk z we would have z f0 (S 1 ) A(0) yielding
Fz = A(0) a contradiction. This proves that z A(tn ) n large. Next we
claim that for all m N one has D(tm ) Int(D(tn )) n large. In fact,
note that D(tn ) = ftn (S 1 ) hence D(tn ) f0 (S 1 ) uniformly as n .
Clearly we can assume from the beginning that A(tn ) = A(0) for all n.
Hence ftn (S 1 ) A(0) = for all n. It follows that for m N xed one has
D(tn ) D(tm ) = for all n large. On the other hand, we can assume
z D(tn ) for all n. From this and D(tn ) D(tm ) = one has either
D(tm ) Int(D(tn )) or D(tn ) Int(D(tm )) for all n large. In the second
case we would have f0 (S 1 ) D(tm ) by taking the limit of the sequence
D(tn ) = ftn (S 1 ). This would imply A(tn ) = A(0) a contradiction. This
contradiction proves D(tm ) Int(D(tn )) for all n large.
The last claim implies that for > 0 xed there are 0 < tn < tm <
such that D(tm ) D(tn ). Choose t = tm and t = tn . Clearly
A(t ) = A(t ). To nd the embedding h we let Ft : D2 A(t ) and
Ft : D2 A(t ) be the lift to the universal cover. They exist because D2
is contractible. Note that Ft (D2 ) Int(Ft (D2 )). Hence for a suitable
base point one has Ft (D2 ) Int(Ft (D2 )). As both Ft , Ft are dieomor-
phisms we can dene
h = (Ft )1 Ft : D2 Int(D2 ).
Hence h is an embedding satisfying
Ft (h(x)) = Ft (x), x D 2 .
By composition with the projection A(s) A(s) for s = t , t one has the
desired property. The lemma follows.
Lemma 6.7. Let f : S 1 [0, 1] M be a vanishing cycle of F , for which

there is an embedding F : D2 [0, 1] M satisfying the conclusion of
Lemma 6.6. Then, there is in the closed transverse of F intersecting A(0).
2
D x t

h
1111
1111
1111 x
0
2 111111111
111111111 N
D x t 111111111
111111111
111111111
111111111
2
D x [t, t]
Fig. 6.4
Proof. Suppose that there is a closed transverse of F intersecting A(0).

Modifying a bit we can assume that there are x0 S 1 and > 0 such
that
f x0 ([0, ]) ,
and
f x ([0, ]) = , x S 1 \ {x0 }.
By hypothesis there are 0 < t < t < satisfying the conclusion of
Lemma 6.6. Let h : D2 Int(D2 ) be the corresponding embedding. In
the cylinder D2 [t , t ] we consider the identication (x, t ) (h(x), t ).
The manifold N obtained from this identication is depicted in Figure 6.4.
Note that N is either a solid torus or a solid Klein bottle depending on
whether h preserves or reverses the orientation in D2 . In any case we let
: D2 [t , t ] N be the quotient map. Denote by P : N M the map
dened by
P (z) = F (x, t),
where (x, t) 1 (z). P is well dened. In fact, if (x, t) = (y, s) then

s = t , t = t and x = h(y). Because F (y, s) = F (y, t ) = F (h(y), t ) =
F (x, t ) = F (x, t) we obtain the result. In addition P is an immersion
since F also is. On the other hand, we can assume that and the normal
vector eld X of F points in the same direction. Now, as contain the
normal curve f x0 ([0, ]) (and does not intersect any other normal segment)
we have that intersects P (N ) as in Figure 6.4. Now it suces to observe
that cannot exit P (N ) because it cannot intersect P ((S 1 [t , t ])). This
proves that cannot be a closed curve, a contradiction. This contradiction
proves the result.
Proof of Auxiliary theorem II: The vanishing cycle f : S 1 [0, 1] M

of F can be assumed to exhibit an immersion F satisfying the hypothesis
of Lemma 6.7. The conclusion of this lemma says that the leaf A(0) cannot
intersect a closed transverse of F . And this implies that A(0) is a compact
leaf. The result follows.
6.4 Some corollaries of the Novikovs compact leaf theorem
We observe that if M and F are orientable and transversely orientable then

the quotient manifold N in the proof above is a solid torus. In such a case
it can be proved without diculty that N is a Reeb component of F. This
remark is summarized in the following result.
Theorem 6.4. A codimension one transversely orientable C 2 foliation on

a compact orientable 3-manifold with nite fundamental group has a Reeb
component.
Corollary 6.1. If F is a codimension one C 2 foliation without compact

leaves of a compact 3-manifold M , then the leaves of F are 1 -injectively
immersed in M .
Proof. Suppose by contradiction that there is a leaf F such that

Ker(1 (F ) 1 (M )) = 0, where 1 (F ) 1 (M ) is the homomorphism
induced by the inclusion F M . To get the contradiction it suces by
Auxiliary theorem II to prove that F has a vanishing cycle. For this we pro-
ceed as follows. As Ker(1 (F ) 1 (M )) = 0 there is a curve F which
is null homotopic in M but not in F . Because is null homotopic in M we
have that there is a map f : D 2 M with = f (D2 ). We can assume
that f is in general position with respect to F . Hence the induced foliation
F in D2 is a singular foliation tangent to a vector eld Y . Note that the

singularities of Y are either saddles or centers and there is in the saddle
connection for Y . Clearly the closed curve D2 is a cycle of Y . Hence the
set of cycles of Y is not empty. By Lemma 6.1 such a set is inductive with
respect to the inclusion order. A minimal element in this set produces a
vanishing cycle for F (see the proof of Auxiliary theorem I in Section 6.2).
This yields the desired contradiction and the proof follows.
Corollary 6.2. Let F be a codimension one C 2 foliation without compact

leaves of a compact 3-manifold M . Then the lift of F to the universal cover
of M is a foliation by planes.
Proof. Let F be the lift of F to the universal cover : M M of M .

Suppose by contradiction that there is a non-simply connected leaf F of
F . Hence there is a closed curve c F which is not null homotopic in
F . Obviously c is null homotopic in M . Hence the closed curve c = (c)
is null homotopic in M . Because c is not null homotopic in F we have
that c is not null homotopic in the leaf F = (F ) of F . This proves
that F is not 1 -injectively immersed in M . Then F has a compact leaf
by Corollary 6.1, a contradiction. This contradiction proves that all the
leaves F of F are simply connected. Thus F = R2 or S 2 . If some F is S 2
then F has a compact leaf with nite fundamental group. By Reeb global
stability theorem it would follow that all the leaves of F are compact, a
contradiction. This contradiction proves that all the leaves of F are planes
as desired.
Remark 6.1. Corollary 6.2 shows that closed 3-manifolds supporting codi-
mension one C 2 foliations without compact leaves are irreducible, namely
every tamely embedded 2-sphere in the manifold bounds a 3-ball. In par-
ticular such manifolds are prime, i.e., they are not non-trivial connected
sum. We observe that compact 3-manifolds supporting Reebless foliations
may be non-irreducible as shown the trivial foliation {S 2 } of S 2 S 1 .
Nevertheless the 2-sphere bundles over S 1 are the solely closed 3-manifolds
which are not irreducible and supports Reebless foliations.
Remark 6.2. The results in this section hold true for C 1 foliations.
Exercise 6.4.1. Let F be a codimension one foliation on S 3 with a compact

leaf L F homologous to zero. Show that L is the torus.


Chapter 7
Rank of 3-manifolds
The notion of rank of a manifold was introduced by J. Milnor, improving

original ideas of H. Hopf, in the search of non-homotopic invariants for
manifolds.
Denition 7.1. Let M be a dierentiable manifold. The rank of M is

the maximum number k N such that there exist continuous vector elds
X1 , . . . , Xk on M with the property that [Xi , Xj ] = 0, i, j (i.e., the vector
elds commute) and X1 , . . . , Xk being linearly independent at each point
of M .
The Poincare-Hopf-Euler Theorem states that any continuous tangent

vector eld on S 2 must have some singularity so that rank (S 2 ) = 0. the
following remarkable result is due to E. Lima:
Theorem 7.1 (Limas theorem, [Lima (1965)]). The rank of the 3-

sphere S 3 is one.
Notice that, since a C 1 vector eld on a compact manifold is always

complete we may state:
{A compact manifold M has rank k} {M admits a locally free

action : Rk M M of the additive group (Rk , +)}
Sketch of the proof of Limas theorem:

First we observe that rank (S 3 ) 1 as it is easily proved by observing
that X(1, x2 , x3 , x4 ) = (x2 , x1 , x4 , x3 ) is tangent to S 3 and non-singular
(outside of the origin 0 / S 3 ).
Assume by contradiction that rank (S 3 ) 2. By the above remark
there exists a locally free action : R2 S 3 S 3 . The action generates a
103
codimension one foliation F (assumed to be C 2 ) on S 3 . By Novikovs com-

pact leaf theorem F exhibits some Reeb-component. Thus we are nished
once we prove the following:
Lemma 7.1. Given any pair of commuting continuous vector elds X, Y

2
in the solid torus D S 1 such that X and Y are tangent to and linearly
2
independent along S 1 S 1 = (D S 1 ), then there exists some point
p D S whereX and Y are linearly dependent.
2 1
2
Proof. The boundary torus (D S 1 ) has isotropy group of the form
r Z + s Z for some r, s R2 C with r/s / R so that we may re-
parameterize as ((rt1 , st2 ), ) (t1 , t2 ) R2 , in such a way that we may
assume

X|S 1 S 1 = y +x and Y |S 1 S 1 =
x y z
for coordinates (x, y, z) R3 with
2
D = {(x, y, z) R3 ; z = 0, x2 + y 2 1} .
We shall therefore prove that any continuous extension 1 , 2 of the

2
vector elds 1 = y x

+ x y
, 2 = z on S 1 S 1 to D S 1 must exhibit
some point where 1 and 2 are linearly dependent. This is done as follows:
we may assume that 1 and 2 are orthonormal extensions as it is easy to
see.
2
Such an extension may be regarded as a path homotopy a : D [0, 1]
2
Go2,3 of the path a = a( , 0) : D Go2,3 with a constant; where G2,3 is
the space of orthonormal oriented pairs of vector on R3 .
By its turn Go2,3 may be identied with the real projective space of
dimension 3, RP (3) as follows: to any element (v1 , v2 ) Go2,3 we associate
a vector (v1 , v2 ) R3 as follows.
Denote by A(v1 , v2 ) the matrix whose columns one v1 , v2 and the vec-
torial product v1 v2 R3 . Then A(v1 , v2 ) is orthogonal and exhibits
some eigenvector u(v1 , v2 ) such that A(v1 , v2 ) u(v1 , v2 ) = u(v1 , v2 ). Let
(v1 , v2 ) be the 2-dimensional subspace of R3 orthogonal to u(v1 , v2 )
The restriction A(v1 , v2 )|(v1 ,v2 ) is an orthogonal linear map of R2 so
that it is a rotation of an angle say (v1 , v2 ) [0, ].
If = 0 we dene (v1 , v2 ) = 0 R3 (in this case A(v1 , v2 ) = Id). for
(v1 , v2 ) (0, ) we choose (v1 , v2 ) (v1 , v2 ) with |(v1 , v2 )| = and
Rank of 3-manifolds 105
same direction and orientation that any v A(v1 , v2 ) v for v (v1 , v2 )

{0}.
Finally, if (v1 , v2 ) = . Then v A(v1 , v2 ) v = 0, v (v1 , v2 ). In
this case we cannot dene an orientation for (v1 , v2 ), what corresponds to
identify the vectors v and v (v1 , v2 ) having |v| = .
Thus above dene gives an homeomorphism : Go2,3 B 3 (0, )/ of
Go2,3 on to the quotient space B 3 (0, )/ of the closed ball {(x1 , x2 , x3 )
3
R3 ; j=1 xj } on R by the equivalence relation that identies the
2 3
points v and v for v B 2 (0, ).

Clearly B 3 (0, )/ is homeomorphic to RP (3) (recall that RP (3) =
R RP (2)).
3
2
The path a = a( , 0) : D Go2,3 is according to this identication, a
diameter of B 3 (0, ) parallel to the x3 -axis, from down to up (orientation).
Thus this path, once projected into RP (3), is not homotopic to a constant.
This proves the theorem.
Remark 7.1. The original proof of Lima is from 1963 and does not make
use of Novikovs compact leaf theorem. Actually, the above proof shows:
Theorem 7.2 (E. Lima, 1963). A compact simply-connected manifold

of dimension three has rank one.
Remark 7.2. The complete solution to the problem of describing the rank
of closed 3-manifolds was given by Rosenberg-Roussarie [Rosenberg and
Roussarie (1970)] where they prove that a rank two 3-manifold must be a
non-trivial ber bundle over the circle with a torus ber.
Exercise 7.0.1. Is there any locally free action of the ane group A(R) on
the 3-sphere?


Chapter 8
Tischlers theorem
8.1 Preliminaries
C2
Let M be a compact manifold admitting a submersion f : M S 1 . We
consider the angle-element 1-form H 1 (S 1 , R) and take = f () its
lift to M . We obtain then a closed 1-form, without singularities, of class
C 1 in M . Since is integrable, it denes a foliation F of codimension
1, class C 1 in M . Let now p M be any point. Since is not singular
there exist neighborhoods p Up M and C 1 vector-elds Xp in Up such
that Xp = 1 in Up . Using partition of the unity we obtain nally a
global vector-eld X in M with the property that X = 1. Since M is
compact, X is complete dening therefore a ow : R M M . From
X = 1 we conclude that the ow is transverse to F . Since d = 0 we
have that LX () = d( X) + iX (d) = 0 so that preserves the foliation
F (each dieomorphism t : M M takes leaves of F onto leaves of F).
We conclude that F is invariant by a transverse ow. Tischlers theorem
states the converse of this fact:
Theorem 8.1 (Tischler-1970, [Tischler (1970)]). Let M be a closed

dierentiable manifold. The following conditions are equivalent:
(i) M supports a foliation F , of class C 1 and codimension 1, invariant by

a transverse ow C 1 .
(ii) M supports a closed 1-form of class C 1 without singularities.
(iii) M bers over the circle S 1 .
Taking into account the Theorem of Sacksteder (according to which a

foliation of class C 2 , codimension 1 and without holonomy is topologically
107
conjugate to a foliation dened by a closed non-singular 1-form (cf. [Sack-

steder (1965)]) we obtain in class C 2 the following equivalent condition:
(iv) M admits codimension one foliation without holonomy.
A demonstration of Tischlers theorem uses strongly the fact that in a closed

manifold M we can nd closed dierentiable 1-forms 1 , . . . , H 1 (M, R)
such that given a base 1 , . . . , of the free part of H1 (M, Z) we have

i = ij delta of Kronecker. Thus the closed 1-form closed in M
j

writes = j j + df for some function f : M R, where {1 , . . . , }
j=1
generates the group of periods Per() < (R, +) of . If is non-singular

then, since Q = R, we can obtain perturbations = j j + df of
j=1
such that is non-singular and Per( ) Q and hence for some integral
multiple k we will Per(k ) Z. Clearly k = dg for some submersion
g : M R/Z = S 1 .
8.2 Proof of Tischlers theorem and generalizations
In this section we state the basic results we need in order to prove Tis-
chlers theorem. Throughout this section F will denote a (non-singular)
codimension one smooth foliation on a connected manifold M of dimension
n 2.
Denition 8.1. Let : C M M be a smooth ow on M . We say that

p
is a ow transverse to F if the vector eld Z = (where t C is
t t=0
the complex time) is transverse to (the leaves of) F .
We say that F is invariant under the ow if each ow map t : M M

takes leaves of F onto leaves of F .
We shall say that F is invariant under the transverse ow of Z if Z is a
complete vector eld on M , whose corresponding ow is transverse to F
and F is invariant under .
Example 8.1. Let M be a n-torus, M = Rn / where Rn is some

lattice. Let F be the foliation on Rn by hyperplanes parallel to a given
Rn . Then F induces a foliation F on the quotient M =
direction Z
Tischlers theorem 109
Rn / which is called a linear foliation on the Torus M . Such a foliation

is invariant under a transverse ow given by a vector eld Z whose lift
As it is easily checked, F is given by a (non-singular) closed
to Rn is Z.
smooth 1-form on M , with constant coecients. The following (classic
real) result states the existence of as a general fact:
Proposition 8.1. Let F be a smooth codimension one foliation invariant

by a transverse smooth ow of Z on M . Then F is given by a (non-
singular) closed smooth 1-form characterized by:
t2
(t (x)) Z(t (x)) dt = t2 t1
t1
x M , t1 , t2 R.
Proof. We follow the original construction in [Plante (1972)]. We con-
struct locally as a time form for Z. Given any point p M choose a
distinguished neighborhood : U M Rn1 R such that takes F |U
into the horizontal foliation on Rn1 R. We may also assume that (p) = 0
and (most important) (t (p)) Rn1 {t}, t with t (p) U (here we
use the fact that is transverse to F and leaves F invariant). Dene now
U := d( ) where : Rn1 R R is the projection (x, y) = y.
Given two such distinguished charts j : Uj M Rn1 R with Uj
connected and having connected intersection U1 U2 = then if we put
j := Uj = d9 j ) we obtain in U1 U2 :

1 U U = d( 1 )U U = d( 2 )U U
()
1 2 1 2
1 2
= 2 U1 U2 .
Remark 8.1. Notice that j (t (pn )) Rn1 {t} implies j (t (q))

Rn1 {t} , q Lpj Uj where Lpj = leaf of F through pj ,. Therefore
j (t (q)) = (aj (q, t), t), q Lpj Uj and henceforth ( j )(t (q)) =
(aj (q, t), t) = t, q Lpj Uj , t 0, so that nally
( j )(t (r)) = t, r Uj , t 0.
In this way we obtain a well-dened closed one-form on M which satises

(t (p)) Z(t (p)) = d( )(t (p)) Z(t (p))
d
= ( )(t (p)) =
dt
d
= (t) = 1, p M, t R .
dt
Corollary 8.1. Let F be a codimension one (non-singular) smooth folia-

tion on a compact (connected) manifold M . The following conditions are
equivalent:
(i) F is invariant under some smooth transverse ow.
(ii) F is given by a closed smooth one-form on M .
Proposition 8.2. Let F , , Z, be as in Proposition 8.1 but assume M

is compact. Given any leaf L0 of F there exist a dierentiable cover
: L0 R M, (x, t) = t (x);
and an exact sequence of groups

#
0 1 (L0 R) 1 (M ) A 0,
where A is a nitely generated free abelian group. Moreover, L0 is compact

if, and only if, A is a lattice on R.
Proof. Dene

H= [] 1 (M ); =0

then H is a normal subgroup of 1 (M ) and it is free because

= n 1 (M ), n Z.
n
Put A := 1 (M )/H then A is nitely generated and also A is abelian

because H [1 (M ), 1 (M )] (the group of commutators) because

= + = + =

, 1 (M ).
Let P : M M be the smooth cover of M , corresponding to H. Let also

F , and respectively.
t be the lifting of F , and t to M
Remark 8.2. F = P (F ),
= P () are usual pull-backs. R
: M

M is dened by
t (p) := t
(P (p)), , t R,
p M
that is, for each

M,
t (p) is the lifting by P of the curve t (P (p)) on
M.
This lifting is well-dened because of the following:

Let , be simple piecewise smooth paths on R with (0) = 0 = (0) and
(1) = t = (1).
Put c = 1 then" c is closed. Since is closed and p : R M is
smooth we have that c p () = 0. Therefore,

= 0, that is, = 0.
(p )# c p (c)
This says that p (c) H. But p (c) = p ()1 p () so that p () and

p () are paths whose lifts by P exhibit the same nal points. Therefore
we may dene R M
: M in a natural way. It is now easy to check
is (locally) smooth in each variable x M
and t R separately. By
Hartogs Theorem is smooth as a map M R M . Finally we have by
construction P (t,
x) = t (P (x)) so that
P (t,
(s,
x)) = t (P ((s,
x))) = t (s (P (x)))
= t+s (P (x)) = P (t
+ s, x).
(s,
This implies that (t, x)) = (t
+ s, x) so that is actually a ow on

M.
=
Let therefore Z be the corresponding smooth vector eld. It is
t t=0
= Z, that is, Z is a lift of Z.
then clear that P Z "
By construction if ) is such that
1 (M = 0 then
is (homotopic

to) the zero element so that = df for some smooth function f: M
R.
Lemma 8.1. We have f( .

t (x)) = t + f(x) t R, x M
Proof. Indeed,
d
t (x))) = df(
(f ( t (x)) Z( t (x)) = ( t (x)) Z( t (x)) = 1
dt
and also

t (x))t=0 = f(x).
f(
0 M
Given any leaf L0 of F on M let L be a leaf of F such that P (L
0) =

L0 . Dene the map g : L0 R M by setting g(x, t) = t (x).
0 R onto M
Lemma 8.2. g is a smooth dieomorphism of L .
Proof. We have

g t
(x0 , t0 ) = (x0 ) = Z( t0 (x0 )).
t t t=t0
Also

g
(x0 , t0 ) = t0 (x))
( = (x0 , t0 ).
x x x=x0 x
Since the ow of Z is transverse to F it follows that the ow is transverse
to F so that g is a local dieomorphism in L 0 R.
Now we notice that if g(x1 , t1 ) = g(x2 , t2 ) then t1 (x1 ) = t2 (x2 ) and

f ((t 2 , x2 )) so that
1 , x1 )) = f ((t
t1 t2 (x1 ) = x2 and t1 + f (x1 ) = t2 +
f(x2 ).
Now, x1 and x2 belong to the same leaf L x0 = L
0 of F so that f (x1 ) =
f (x2 ), this implies t1 = t2 and therefore x1 = x2 .
Therefore g is also injective and it is a dieomorphism of L 0 R onto its

image g(L0 R) M . It remains to prove that g(L0 R) = M .
It is enough to prove that this image of g is closed. Take any point x1 M
belonging to the closure of g(L 0 R) in M. Let B 1 x1 be any open ball
in the leaf L 1 x1 . Let U be the cylinder U = t (B1 ), and take any
tR
g(L
x U 0 R).
We have x 0 ) for some s R and also there exists r R such that
s (L
x 1 ). Thus x1
r (B 0 ) and hence x1 g(L
sr (L 0 R). This proves
the lemma.
Now we may prove:
0 M
Lemma 8.3. P |L0 : L L0 M is a bijection and therefore a
dieomorphism.
Proof. The map P is injective, for if x1 , x2 L 0 are such that P (x1 ) =

P (x2 ) then we may take a path : [0, 1] L0 of class C 1 with (0) = x1
and (1) = x2 .
This gives a projected path = P : [0, 1] L0 M which is closed,
i.e., 1 (L0 ).
We have
d
(f ((t))) = ((t)) (t) so that
dt

((t)) (t) = 0, t [0, 1] and therefore
((t)) (t) = 0, t [0, 1].

This gives
1
0= ((t)) (t) dt =
0
, i.e., x1 = x2 .
and therefore [] H 1 (M ). This gives (0) = (1) in M
Let now t : L0 R L R be given by t (x, s) := (x, s + t). Let also

G : L0 R M be dened by

G(x, s) := g (P |L0 )1 (x), s .
Notice that

0 R.
(P |L0 )1 (x), s L
Consider the following diagram
P
M M G
L0 R

t $ t $
$ t
P
M M G L0 R
The left side is commutative by construction.
Now we observe that given (x, s) L0 R we have
1
G(t (x, s)) = G(x, s + t) = g P |L0 (x), s + t
1
=s+t P |L0 (x)
1
=t s P |L0 (x)
t (G(x, s)) = (
G(Mt (x, s)) = t G)(x, s).
Therefore the whole diagram is commutative.
Dene now
:= P G : L0 R M
by requiring tht the diagram below is commutative:

M L0 R

t $ $ t

M L0 R
In other words:
t (x, s) = (x, s + t).
Lemma 8.4. : L0 R M is a covering map.
Proof. We know that P is a covering map P : M M . It is therefore

enough to show that G : L0 R M is a covering map. Actually g : L 0
RM 0 L0 , so that G : L0 R
is a dieomorphism and so is P : L
L0
is a dieomorphism.
M
Clearly (x, t) = t (x, 0) = t (x), t R x L0 . Therefore

satises the rst condition in the statement of Proposition 8.3.
If for any [] 1 (M ) we have [] = # ([]) in 1 (M ), for some []
1 (L0 R) then

= = () = (t ) ()
L0 R
= 0
for L0 R
and =0along L0
) for some
so that [] H. Conversely, if [] H then = P# ( )
1 (M
and therefore we have = g# ( 1 (L
) for some 0 R) so that =
)) = (P G)# () where = P
P# (g# ( 1 (L0 R) is obtained
in a natural way.
Therefore we have proved the following:
Lemma 8.5. The sequence below is exact

#
0 1 (L0 R) 1 (M ) A 0.
Remark 8.3. Another way of seeing the " above equivalence is the following:
if [] H then 1 (M ) is such that = 0. Therefore we " may consider
the lifting of by to L0 R obtaining a path such that (t ) () = 0
and therefore is closed that is, [] = # ([]) for [] 1 (L0 R).
Assume that A has rank one, A Z. We may take a transformation

T : L0 R L0 R which corresponds to a generator of A = 1 (M )/H
(notice that the covering : L0 R M has group isomorphic to 1 (M )/H
#
because of the exact sequence (0 1 (L0 R) 1 (M ) 1 (M )/H
0).
T
L0 R L0 R
The diagram " commutes.
M

Let T0 := T L0 {0} : L0 {0} L0 R.
Lemma 8.6. t0 R {0} such that T (L0 {0}) = L0 {t0 }.
Proof. Write T (x, t) = (a(x, t), b(x, t)) so that T =

b(x,t) (a(x, t)) = t (x) b(x,0) (a(x, 0)) = 0 x L0 . ()
We have a(x, 0) L0 , x L0 therefore (since t is transverse to F ) we
must have from (*) from (b(x, 0) is constant x L0 and therefore if we
put t0 = b(x, 0) then
t0 (a(x, 0)) = x, x L0 .
That by
T (L0 {0}) L0 {t0 } .
If t0 = 0 then b(x, 0) = 0, x L0 and T (x, 0) = (a(x, 0), 0) and also (*)
a(x, 0) = x x L0 .
Thus T (x, 0) = (x, 0), x L0 . This is not possible for T is a non trivial
covering transformation.
Dene now a map f : M R/t0 Z S 1 by setting f (x) := s (mod t0 )

where x s (L0 ). Notice that given any x1 M , since : L0 R M ,
(x, t) = t (x) is a covering, it follows that x1 s (L0 ) for some s R.
Now, if s1 , s2 R are such that x sj (L0 ), j = 1, 2, then x1 , x2 L0
with x = s1 (x1 ), x = s2 (x2 ) so that (x1 , s1 ) = x = (x2 , s2 ). Since
the group of covering maps of is generated by T we must have (x2 , s2 ) =
T n (x1 , s1 ) for some n Z so that s2 = s1 + n t0 , so that s2 = s1 (mod t0 ).
Therefore f : M S 1 = R/t0 Z is well-dened and clearly smooth.
Since t takes leaves of F onto leaves, f is constant along the leaves of F.
Thus f is a smooth rst integral for F.
Assume now that rank(A) = 0. In this case H = 1 (M ) and P : M M
is the universal covering of M . Since A = {0} we have a dieomorphism

M L0 R which is not possible because M is compact.
Conversely, assume now that L0 is a compact leaf of F . Since F has trivial
holonomy the Stability Theorem of Reeb (see [Godbillon (1991)]) implies
that all the leaves of F are compact and F is a Seifert (smooth) bration.
Now, the group A acts on L0 R taking leaves of (F ) onto leaves of (F )
in a natural way as in Lemma 8.5. Therefore, since F is a compact foliation,
the leaves of () are closed on L0 R and therefore the action of A must
be discrete so that indeed, A must correspond to a discrete subgroup of R

and therefore rank(A) 1 as it is well-known (see Remark 8.4 below).
Remark 8.4. We remark that, according to what we have seen, we have

an homomorphism of groups
: 1 (M ) (R, +)

[]

whose kernel is H so that there exists an injective homomorphism : A =

1 (M )/H R, so that A is naturally identied to a certain subgroup of
(R2 , +).
Therefore we have proved that rank(A) = 1 L0 is compact all leaves

of F are compact. This ends the proof of Proposition 8.2.
Corollary 8.2. Let M be compact, F, t , A as in Proposition 8.2. Then

1 rank(A) rank(H1 (M, Z)). Moreover, if M is an orientable compact
manifold and rank(H1 (M, R)) 1 then F is a foliation by compact leaves.
Next step is the following:
Proposition 8.3. Let F , t , , A and M be as in Proposition 8.2. Let

Per(L0 ) := {t R; t (L0 ) = L0 } and t : L0 R L0 R be given by
t (x, s) = (x, s + t).
(i) If T : L R L0 R is a covering transformation of the covering
: L0 R M
(x, t) t (x)
then T (L0 {t}) = t0 (L0 {0}) for some t0 = t0 (T ) R.

(ii) The correspondence T t0 (T ) denes an isomorphism A Per(L0 ).
(iii) Per(L0 ) is the group of periods of .
Proof. As we have seen in the proof of Proposition 8.2 above for each
covering transformation T of we must have T (L0 {0}) = L0 {t0 (T )}
for some t0 (T ) R. Moreover t0 (T ) = 0 if, and only if, T is the identity.
It is also possible to see that t0 (T ) depends only on T , not on the choice of
the leaf L0 M . Therefore we have T (L {0}) = t0 (T ) (L {0}) for any

leaf L of F .
The mapping : A R, T t0 (T ) is therefore such that (A) Per(L0 ):
given any x0 L0 we have T (L0 {0}) = L0 {t0 (T )} T (x0 , 0)
L0 {t0 (T )} if we write T (x, t) = (a(x, t), b(x, t)) then T (x, 0) =
(a(x, 0), t0 (T )) and therefore
x0 = 0 (x0 ) = (x0 , 0) = T (x0 , 0)
(a(x0 , 0), t0 (T )) = t0 (T ) (a(x0 , 0))
so that t0 (T ) (L0 ) = L0 and then t0 (T ) Per(L0 ). Thus we have : A
Per(L0 ) R.
Lemma 8.7. is an injective group homomorphism.
Proof. Given S, T A be have (S T ) = t0 (S T ) and by denition

S T (L0 {0}) = L0 {t0 (S T )}.
But, on the other hand,
S T (L0 {0}) = S(T (L0 {0})) = S(L0 {t0 (T )})
S T (L0 {0}) = S(L0 {t0 (T )}).

Now, for any leaf L of F we have S(L {0}) = L {t0 (S)}. Therefore
S T (L0 {0}) = L0 {t0 (S) + t0 (T )}.
This implies that t0 (S T ) = t0 (S) + t0 (T ). The injectivity of we have
already checked.
Finally we claim that is surjective. Indeed, given any t0 Per(L0 )

and any x0 L0 we may consider paths := st0 (x0 ) in M and in L0 ,
joining l0 t0 (x0 ) to x0 because t0 Per(L0 ).
The homotopy class
[] = [ ] 1 (M )
is such that if T A corresponds to [] then T (L0 {0}) = L0 {t0 (T )}
where t0 (T ) is given by

t0 (T ) = = + =

1 1
d
= (st0 (x0 )) (st0 (x0 )) dx = t0 1 ds = t0 .
0 ds 0
Thus t0 (T ) = t0 and is surjective. This shows (i) and (ii).
Lemma 8.8. Per(L0 ) is the group of periods of which is dened by

Per() := ; [] 1 (M ) < R.

"
Proof. Let t0 Per(L0 ) and [] = [ ] as above, then [] = t0 so
"
that t0 Per(). Conversely, given any period t0 Per() say t0 =
for some [] 1 (M ) we may perform small homotopies so that [] is of
the form [] = [1 1 r r ] with j segment of orbit of t and
j contained in a single leaf of F , j {1, . . . , r}. Using the ow we may
obtain a homotopy between r1 r and some path of the form .
Using the ow: r1 r is homotopic to some path of the form .
Therefore we may assume that r = 1, and = 1 1 . Therefore

t0 = = 1 (t0 ) = t0 (x0 )
1
and 1 (0) = x0 belong t0 a same leaf of F and therefore t0 Per(L0 ). This

proves (iii) and Proposition 8.3.
Corollary 8.3. Let F , , t , A, Per(), M compact be as above. The

leaves of F are compact if, and only if, Per() R has rank one and
denes a lattice on R. In any other case the leaves of F are not closed.
Proof. We have already proved that rank(A) 1 and also rank(A) = 1

and if, and only if, F is a compact foliation. Moreover rank(A) 2 implies
Per() R is not discrete in fact it is dense, what implies (see Remark 8.5)
that A acts in the leaves of (F ) in L0 R with non-discrete dynamics.
This implies that the leaves of F are not closed.
Remark 8.5. Let M be a compact dierentiable manifold supporting a

non-singular codimension one smooth foliation invariant by a transverse
ow. Then 1 (M ) is not nite, indeed rank(H1 (M, Z)) 1.
of M is also compact
Indeed, if 1 (M ) is nite then the universal covering M
so that the closed one-form lifts into a closed non-zero smooth 1-form
on M which is exact, = df for some smooth function f: M R. Since
is compact f must exhibit some critical point is constant and has
M
some singularity, contradiction.
Now we are in conditions to prove Tischlers theorem.
Proof of Tischlers theorem. We may assume that M is orientable and

oriented. According to what we have seen above the foliation F is given
by a non-singular smooth closed one-form in M . We may nd a basis
{wj = (j )} of the group the De Rham cohomology group H 1 (M, 1 )
given by (classes of) closed 1-forms in M such that
for some loops 1 , . . . , r corresponding to a basis of the free part of

H 1 (M, Z) we have

wi = ij .

r
We may therefore write = j wj + df for some j R, and some
"
j=1
f : M R smooth. Then j = j so that {1 , . . . , r } Per(). Indeed
Per() is generated (as a group) by the j s, j = 1, . . . , r, i.e., Per() =

r
{1 , . . . , r } . Let now (1 , . . . , r ) Rr be such that := j wj is
j=1
close enough to so that it is also non singular (recall that is non singular
and M is compact) and the subgroup {1 , . . . , r } of R is a rank 1 discrete
lattice (it is enough to choose {1 , . . . , r } Q of rank 1). Thus denes
a bration of M over the circle S 1 = R/ , = {1 , . . . , r } .


Chapter 9
Plantes compact leaf theorem
This chapter is dedicated to an exposition of the celebrated result by J.

Plante [Plante (1975)], asserting that every codimension one foliation on a
closed manifold M exhibiting a leaf of subexponential growth also exhibits
a leaf of polynomial growth of degree max{0, 1 (M ) 1}, where 1 (M )
is the rst Betti number of M . In particular, the foliation has a compact
leaf provided that it exhibits some leaf of subexponential growth on compact
M having 1 (M ) 1.
9.1 Growth of foliations and existence of compact leaves
In opposition to a n-dimensional version of Novikovs compact leaf theorem

there are examples of foliations C 2 of codimension one of S n , n 4
without compact leaves. In particular, the minimal set of such a foliation
is exceptional (cf. [Sacksteder (1965)]). It was J. Plante, in an outstanding
work, who initiated the modern comprehension of such facts relating the
concepts of growth of leaves and existence of exceptional minimal sets (cf.
[Plante (1973)]). Let us recall such concepts:
9.1.1 Growth of Riemannian manifolds
Let (M, g) be a connected oriented Riemannian manifold of class C r , r 1.

Given any point x M the growth function of M at x is dened by x (r) :=
volume of the closed metric ball B[x; r]. The growth type of x does not
depend on the choice of x M . In this way we may introduce of polynomial
growth, exponential growth, ... for (M, g). If M is compact then it has
polynomial growth of degree zero. In the case r = we have the following:
121
Proposition 9.1 (Moussu, Pelletier [Godbillon (1991)]). For

any r 0 and any x M the closed ball B[x; r] is a standard Whit-
ney domain: the boundary B[x; r] contains a compact subset K with zero
(m 1)-dimensional measure such that B[x; r] K is a submanifold with
boundary of M . Moreover, the function x (r) = vol(B[x; r]) is dieren-
tiable with respect to r and its derivative at the point ro is the volume of
the (m 1)-dimensional sphere B[x; ro ].
vol(B[x;r])
It follows from the above result that if lim inf > 0 then M
r+ vol(B[x;r])
has exponential growth.
9.1.2 Growth of leaves

The notion of growth for the leaves of a foliation on a compact manifold
may be introduced in a geometric way regarding the growth of the volume
of the balls in the leaves, and it will be related to the growth of the orbits
of the holonomy pseudogroup of the foliation, as we will see. The main
remark is the following:
Proposition 9.2 ([Godbillon (1991)]). Given two Riemmanian met-

rics in a compact manifold M equipped with a C 1 regular foliation F ,
the metrics induce on each leaf L of F , complete quasi-isometric metrics.
Therefore, the growth type of the leaf L does not depend on the choice of
the ambient metric.
In the non-compact case however, we may x the metric and consider

the growth type of the leaves with respect to this xed metric. Let (M, g)
be a Riemannian manifold, perhaps non-compact, and let F be a (regular)
C 1 foliation of codimension k on M . Assume that M is oriented and F is
transversely oriented. For each x M denote by Lx the leaf of F through
x. The metric on M induces a metric gx along the (immersed) leaf Lx .
Denition 9.1. The growth type of the leaf Lx with respect to the metric
g is the growth type of the Riemmanian manifold (Lx , gx ).
Therefore, compact leaves have polynomial growth of degree zero.
9.1.3 Growth of orbits

Let X be a Hausdor topological space and a collection of homeomor-
phisms g : U V , where U, V are open subsets of X. Denote by Dom(g)
Plantes compact leaf theorem 123
and Range(g) the domain and the range of g respectively.
Denition 9.2 ([Plante (1972)]). is a pseudo-group of local homeo-

morphisms of X if:
(i) For any g we have g 1 and Dom(g) = Range(g 1 ), and
Dom(g 1 ) = Range(g).
(ii) If g1 , g2 and g : Dom(g 1 ) Dom(g2 ) Range(g1 ) Range(g2 ) is a
homeomorphism such that g Dom(gi ) = gi , i = 1, 2, then g .
(iii) Id : X X belongs to .
(iv) If g1 , g2 then g1 g2 , with Dom(g1 g2 ) g11
(R(g2 ))Dom(g1 ).
(v) If g and U Dom(g) is an open subset then g U .
The orbit of x in the pseudogroup is dened by (x) := {g(x) X, g

, x Dom(g)}. Assume now that is nitely generated by a (symmetric)
nite subset o .
Denition 9.3. For x X and n N we dene n (x) := {y X, y =

g1 ... gk (x), k n, gj o , j = 1, ..., k}. The growth type of the orbit
of x in is the growth type of the function x (n) := n (x) as n N.
9.1.4 Combinatorial growth of leaves

Let M be a compact manifold and F a foliation of codimension k on M .
Given a nite cover U = {U1 , ..., Ur } of M by distinguished neighborhoods
for F we denote by U the holonomy pseudogroup associated to this cover.
Then U is nitely generated. Given a point x U1 and n N the value
of the growth function U ,x (n) is equal to the number of plaques of U that
can be joined to the plaque P1,x of U1 , by a chain of plaques with at most
n plaques.
Proposition 9.3 ([Godbillon (1991)]). Let U and V be nite covers by

distinguished neighborhoods of the manifold M . The growth type of the
functions x,U and x,V is the same. If F is of class C 1 then the growth
type of the functions x,U is the same for all points x is a same xed leaf
of F .
We dene therefore the combinatorial growth type of a leaf L of a C 1

foliation F on a compact manifold M as the growth type of the function
x,U where x L is any point and U is any nite cover of M by distinguished
neighborhoods. According to what we have seen, in this compact case, the
growth type of a leaf L of F is equal to the growth type of the orbit of any
point x L in the holonomy pseudogroup F of F. In case the manifold

is compact we also have:
Proposition 9.4 ([Godbillon (1991)]). Let F be a C 1 transversely ori-

ented foliation on a oriented compact Riemannian manifold M . The (geo-
metric) growth type of any leaf L of M is equal to the combinatorial growth
type of L.
9.1.5 Growth of groups

Let G be a nitely generated abstract group. There exists therefore a subset
n
Go G such that any element g G writes as g = g where A is nite
A
set, n Z and g Go , A. The set of generators Go is symmetric
if g Go then g 1 Go . We will assume that Go is symmetric. For any
n N we dene the subset Gn of elements of G which can be expressed as
a word of length at most n in the generators. That is,
Gn = {g G, g = g1 ... gk , k n, gj Go , j = 1, ..., k}.
The growth function of G is g(n) = Gn , n N. This denition can
be extended as follows [Plante and Thurston (1976)]: Let d be any left-
invariant metric on G. We assume that G is discrete so that for any g G
there exists g > 0 such that the metric ball BG (g; g ) G contains only
the element g. The growth function of the pair (G, d) is therefore dened as:
for t > 0, (t) = BG (e; t) where e G is the identity. If (t) < , t 0
then we say that is the growth function of (G, d). In case G has a
symmetric set of generators S we may consider any function n1 : S R+
such that n11 (0, r] < , r 0. For any g G we dene the function

n(g) := min{ n1 (si ), g = si , si S}. Clearly n(.) denes a left-
i=1 i=1
invariant metric d on G by setting d(g, h) := n(g 1 h). If S = G then any
left invariant metric on G is obtained this way. Let us precise our main
denition:
Denition 9.4 ([Plante and Thurston (1976)]). The pair (G, d) has
polynomial growth of degree k if there exists a polynomial p(x) of degree k
such that (t) p(t), t 0, where (.) is the growth function of (G, d).
We may also consider polynomials of the form ax , 0, R.
Proposition 9.5 ([Plante and Thurston (1976)]). Let S be a nite

symmetric set of generators of G and n1 : S R+ , n1 1. Denote by
n(.) the metric above corresponding to n1 1. The growth of G is polyno-

mial with respect to some left-invariant metric d if and only if, (G, n) has
polynomial growth.
Example 9.1. Some examples of growth of groups are given below:

A nitely generated abelian group has polynomial growth.
A non-cyclic free group has exponential growth once we have (n) =

n1 m(2m 1)n 1
2m (2m 1)k =
k=0 m1
Let M 2 be a closed surface orientable of genus g 2 then 1 (M ) has
exponential growth.
A situation of particular interest is the case of groups of polynomial

growth, on which we have essential contributions of J. Plante, J. Wolf and
J. Milnor (cf. [Wolf (1968)], [Milnor (1968)], [Plante (1975)]).
We have:
1. A nilpotent group of nite type has polynomial growth.
2. A solvable group of nite type G which does not have a nilpotent
subgroup of nite index has exponential growth. In case G has polynomial
growth then G is polycyclic (G is polycyclic if G = Gk Gk1
G0 = {e} with Gk /Gk1 cyclic).
3. J. Tits has shown ([Tits (1972)]) that the converse of 1. is true: a group
of nite type having polynomial growth has a nilpotent subgroup of nite
index.
9.2 Holonomy invariant measures
Let X be a Hausdor topological space and a pseudogroup of local home-

ormorphisms of X. Denote by c (X) the ring of subsets of X generated by
the compact sets. A measure on c (X) is -invariant if:
(i) is non-negative, nitely additive, nite on compact sets.
(ii) g and any measurable set A Dom(g) we have (g(A)) = (A).
Consider now the case M is a C manifold and F is a (regular) foliation
of codimension k 1 on M , assumed to be transversely oriented, of class
C . The holonomy pseudogroup dened by F will be denoted by (F ).
Denition 9.5 (Plante, [Plante (1975)]). A foliation F is said to have

a measure preserving holonomy, or holonomy invariant measure, if its holon-
omy pseudogroup has a non-trivial invariant measure which is nite on
compact sets. The support of an F -invariant measure is the set of points

x M such that: given any k-dimensional disk transverse to F, D k F ,
with x Int(Dk ), we have (Dk ) > 0. Since is F -invariant, supp() is
closed and F -invariant.
Example 9.2. Assume F has a closed leaf L on M . Given any transverse

section M transverse to F we dene a measure on as follows: A
, (A) := {A L}. This denes an holonomy invariant measure. Let
now F be given by a closed holomorphic 1-form on a complex manifold
M . The holonomy pseudogroup is naturally a pseudogroup of translations
(F ) (C, +) and any leaf of F has trivial holonomy. Any Borel measure
on C which is invariant by translations is also (F )-invariant. Another
situation comes when F is real given by a closed C k-form on M . In this
given any transverse section M , transverse to F , the restriction
case,
is a volume element (F is transversely oriented) which is positive on
open sets. The fact that is closed implies that the induced transverse
measure is F -invariant. Assume now that M, F , are holomorphic. Using
the complex structure J : T M T M we may consider the real part and
the imaginary part Re(), Im() of . Take = Re() Im(), this is a
2k-form real form, which is closed and denes F as a real codimension 2k
foliation. Therefore the restriction 2k , where = 2k is regarded as a
2k-dimensional real submanifold transverse to F , is positive on open sets
and denes an F -invariant transverse measure.
Example 9.3. It is a fairly well-known fact that a compact manifold M

supporting a codimension one C 1 Anosov ow t : M M has fundamen-
tal group with exponential growth [Thurston (1972)]. Such a result is not
stated for codimension one holomorphic foliations (see [Ghys (1995)]). We
shall consider an example of such a situation [Scardua (1997)]. Let M be a
compact complex manifold of dimension n equipped with a closed holomor-
phic 1-form and f : M M an automorphism such that f = . for
some C \ S 1 . We put (x, t) := t.(x) on M C so that d =

for = dtt . The 1-form is closed and holomorphic in M C so that
according to [Scardua (1997)] = 0 denes a codimension one holomor-
phic foliation F on M C which is transversely ane. F is non-singular
provided that is non-singular on M . The action
: Z (M C ) M C , (n, (x, t)) = (f n (x), n .t)
where n Z, (x, t) M C , is a locally free action generated by the
automorphism : M C M C , (x, t) = (f (x), 1 .t). Since
(x, t) = (x, t) and = it follows that F induces a codimension

one transversely ane holomorphic foliation F of the quotient manifold
V n+1 = M / = M /Z. We apply this construction in a concrete situation:
Take A : C2 C2 as the linear automorphism given by

11
A= .
12

Then A has eigenvalues s = 32 5 andu = 3+2 5 . The corresponding

eigen-spaces are generated by vs = (2, 1 5) and vu = (2, 1 + 5) respec-
tively. The stable linear foliation
and the instable linear foliation are given
by the 1-forms s = 2dx+(1+ 5)dy and u = (1+ 5)dx2dy respectively.
Take F u = 0 on C2 . We consider the action of the integer lattice Z2 on
u :
C2 obtained in the natural way and put M = C2 /Z2 = C C . The map A
leaves Z2 invariant so that it induces an automorphism F : M M , which
is indeed given by F (z, w) = (zw, w2 ) for coordinates z = e2ix , w = e2iy
on C C .
Now we consider the Z-action on C given by : Z C C , (n, t)
s .t. Then M = C /Z C /Z = M /Z obtained this way is a compact
n
surface equipped with an automorphism f : M M induced by F : M

M indeed, F (1 1 2 2
s .z, s .w) = s .(zw, w ).
u satises A () = 1
Now, the 1-form =
s . and corresponds to a
Darboux 1-form = (1+ 5) z 2 w on M = C C . Therefore, we have
dz dw
F = 1
s . and nally since = it follows that induces a closed
holomorphic 1-form on M with the property that f () = 1 s .. Thus,
according to the above construction, the manifold V 3 = M C /Z obtained
by quotienting M C with the action of Z given by the action of f on M
and of the homotheties t s .t on C , is a compact complex 3-manifold
equipped with a transversely ane codimension one holomorphic foliation
F coming from the linear unstable foliation Fu on C2 . The foliation F
exhibits exponential growth (for any metric on the compact manifold V 3 )
because An expands vu .C by a factor nu . On the other hand, [Scardua
(1997)], the leaves of F on V are dense, biholomorphic to C C or to
(C /Z) C .
Example 9.4. Let G be a Lie group which has polynomial growth in some
left invariant metric. Let : G M M be a locally free smooth action
on a manifold M . There exists a Riemannian metric on M which restricts
to the -orbits as the induced metric coming from G. Thus denes a
foliation F on M , whose leaves have polynomial growth for this metric. For
instance we may take any locally free holomorphic action : Cn M M
where M is a complex manifold and the euclidian metric on Cn . The
foliation by -orbits on M has polynomial growth for a suitable metric on
M . For n = 1 we have a holomorphic ow whose orbits have polynomial
growth for a given metric on M .
Example 9.5. Here we complexify an original example in [Plante (1975)].

Let G be a simply-connected complex Lie group and H < G a closed
(Lie) subgroup of (complex) codimension one. Given any discrete subgroup
< G the group H acts on the quotient G/ by left translations generating
a foliation F of codimension one. The leaves of F are the orbits of the above
action. Since G is simply-connected the universal covering G G/ lifts
F into a foliation F on G whose leaf space is the Riemann surface H \ G.
The exact homotopy sequence of the bration G H \ G shows that (for
H
H connected) the manifold H \ G is simply-connected since G is simply-
connected. Therefore, H \ G is either dieomorphic to CP (1), C or D.
Therefore the action of on H \ G denes a global holonomy of F as a
subgroup of Di(N ) for N {CP (1), C, D}, so that this global holonomy
group is either a subgroup of Moebius maps, ane maps or SL(2, R). If
is uniform, that is, the quotient G/ is compact, then G is unimodular
and the action of G on H \ G has an invariant measure i H is unimodular
i there exists a -invariant measure. Therefore, when G/ is compact F
admits an invariant measure i H is unimodular.
The existence of holonomy invariant measures is a consequence of subex-

ponential growth for the leaves as stated below:
Theorem 9.1 (Plante, [Plante (1974)]). Let F be a C 2 foliation of

codimension k 1 on the compact manifold M . Assume that F exhibits a
leaf L having subexponential growth. Then there exists a nontrivial holon-
omy invariant measure for F which is nite on compact sets and which
has support contained in the closure L M of L.
It is also known that if a codimension one (real) foliation of class C 2 on

a compact manifold admits a non-trivial holonomy invariant measure then
F has a leaf with polynomial growth [Plante (1974)]. Is this also true for
complex foliations?
9.3 Plantes theorem
Let F be a C 0 codimension k foliation of a closed Riemannian n-manifold

M . Given an open subset U M we denote by F /U the restriction of F
to U , namely, the foliation whose leaves are the connected components of
the leaves of F intersected with U . Recall that the leaves of F /U are called
the plaques of F in U . Denote by D r the closed unit disk in Rr , r N.
We shall call a cover U = {Ui : i I} of M nice if the collection
Int(U) = {Int(Ui ) : i I} is an open cover of M and the following prop-
erties hold for all i, j I:
(1) Ui is dieomorphic to Dnk Dk and F /Ui is given by the trivial

codimension k foliation {Dnk : Dk } of Dnk Dk .
(2) Every plaque of F in Ui intersects at most one plaque of Uj .
To any such cover U we can associate the disjoint union X = iI Xi

where each Xi is the leaf space of F /Ui . Clearly each Xi is dieomorphic
to Dk so X is a compact metric space. By a plaque of F /U we mean a
plaque of F in Ui for some i I. A chain of plaques is a nite collection of
plaques {P1 , , Ps } of F /U satisfying Pi Pi+1 = for all 1 i s 1.
Given a leaf L of F , a nice cover U, x L X and n N we dene
Ln (x) as the set of all y X for which there are 1 s n and a chain of
plaques {P1 , , Ps } such that x P1 and y Ps . Clearly #Ln (x) <
where # denotes the cardinality operation.
Denition 9.6. We shall say that L has exponential growth if there are a
nice cover U and x X L such that
log(#Ln (x))
lim inf > 0.
n n
Otherwise we say that L has subexponential growth. If instead there are
p N, K > 0, a nice cover U and x X L such that
#Ln (x) Knp , n N,
then we say that L has polynomial growth of degree p.
It turns out that the above denitions depend neither on the point x L
nor on the nice cover U.
Now we state and prove the following result due to Plante [Plante
(1975)]. Recall that the rst Betti number of M is the rank 1 (M ) of
the free part of its rst integer homology group.
Theorem 9.2 (Plante [Plante (1975)]). Every transversely oriented

C 1 codimension one foliation of a closed manifold M exhibiting leaves of
subexponential growth also exhibits leaves of polynomial growth of degree
max{0, 1 (M ) 1}.
Proof. Fix a foliation F and a leaf L of F as in the statement. Then,

there is a nice cover U and x X L such that
log(#Ln )
lim inf =0
n n
where we have written Ln instead of Ln (x) for simplicity.
Let us prove the identity
#Ln+1 #Ln
lim inf = 0.
n #Ln
Suppose that it fails. Then, there is d > 0 such that #Ln+1 (1 + d)#Ln
for n large. By induction we obtain #Ln+k (1 + d)k #Ln for n large and
k N. Taking log and lim inf k in this inequality we obtain log(1+d) 0
which is absurd since d > 0. This proves the identity. Then,
#Ln+1 #Ln1 #Ln+1 #Ln
lim inf = lim inf
n #Ln n #Ln
#Ln #Ln1
+ lim inf =0
n #Ln
and so there is a sequence ni such that
#Lni +1 #Lni 1
lim = 0. (9.1)
i #Lni
For all i we put the normalized counting measure i in X supported on
Lni , i.e.,
#(A Lni )
i (A) = .
#Lni
Since X is compact, and so the space of all Borel probability measures
is compact in the weak topology, we can assume that there is a Borel
probability measure in X to which the sequence i converges. In the
sequel we shall prove an invariant property for this measure.
Given x Xi we denote by Pi (x) the plaque of F in Ui containing x.
Denote by Dom() and Rang() the domain and range operations. Every
pair (i, j) with Ui Uj = denes a C 1 map ij : Dom(ij ) Xi Xj ,
Dom(ij ) = {x Xi : Pi (x) Pj (y) = for some y Xj } and ij (x) = y.
Denote by the collection of all such maps. Notice that each is

invertible with inverse ji whenever = ij . Therefore 1 for all
.
We claim that is -invariant, i.e.,

f d = f d( ),
for all continuous map f and all , where is the pullback of

under . Indeed, for all such f and we have,

1
f d f d( ) = lim f (y) f ((y)) .
i #Lni ni ni
yL yL
But

f (y) f ((y)) = f (y) f (y)
yLni yLni yLni y 1 Lni

= f (y) f (y) = (f (y)),
yLni \ 1 Lni y 1 Lni \Lni yLni 1 Lni
where above denotes symmetric dierence. Then,

ni 1 ni
f d f d( ) sup |f (y)| lim #(L L ) . (9.2)
y i #Lni
Let us prove the inclusion
Ln 1 Ln Ln+1 \ Ln1 , n N, .
Indeed, we have Ln Ln+1 and Ln Ln+1 by denition. If y Ln \
1 Ln then y Ln+1 and if y Ln1 then we would have (y) Ln and
so y 1 Ln which is absurd. Then Ln \ Ln Ln+1 \ Ln1 . Analogously
we prove Ln \ Ln Ln+1 \ Ln1 and the inclusion follows.
Since Ln1 Ln+1 , (9.2) and the previous inclusion imply

ni +1
#Lni 1
f d f d( ) sup |f (y)| lim #L =0
y i #Lni
proving the claim.
Denote by N the nerve of Int(U), namely, the collection of all subsets
J of I such that either J = or J = and iJ Ui = (we have written Ui
instead of Int(Ui ) for the skae of brevity). Clearly if J N every subset
of J belongs to N thus it the nerve an abstract nite simplicial complex.
Its vertices are the elements of U and there is an edge (1-simplex) joining
Ui to Uj if and only if Ui Uj = . A chain of N is a nite ordered subset

C = (Ui1 , , Uis ) of U such that Uij Uij+1 for all 1 j < s (we then say
that s is the length of C). If additionally Uis Ui1 = then we say that
C is a closed chain. Since N is nite there is d > 0 such that every pair of
vertices can be connected by a chain of length at most d. To each closed
chain C it corresponds the homotopy class [C] which form together the
homotopy group 1 (N ). A standard trick in algebraic topology shows that
the set G of homotopy classes corresponding to closed chains of length at
most 3d generates 1 (N ). Moreover, the homotopy class of a closed chain
of length at most nd can be written as the product of at most n elements of
G. Finally let us mention that for suitable U the homotopy groups 1 (M )
and 1 (N ) are homeomorphic (e.g. [Bott and Tu (1982)] p. 148 or [Rotman
(1988)]). We shall assume hereafter that this is the case.
Next we use to construct a homomorphism : 1 (N ) R. First
observe that every closed path in M is homotopic to a closed path of the
form
c = 1 1 2 2 r1 r1 r r
where i i are contained in some Uj U, i Xj is transverse to F
and i is tangent to F for all i {1, , r}. Then, we can dene
r
1 1
([C]) = (i (0, 1)) + (i ({0})) + (i ({1}))
i=1
2 2
where c is a closed path as above contained in C and the signal above de-
pends on whether the transversely orientation of F is preserved or reverted
by i . Denote by p the rank of the nitely generated subgroup (1 (N ))
of R. Since 1 (M ) and 1 (N ) are isomorphic we have p 1 (M ).
So, to prove the theorem, it remains to prove that every leaf intersecting
the support of has polynomial growth of degree max{0, p 1}.
Previously let us remark that for every subgroup G of the additive group
R, every base {g1 , , gp } of G and every > 0 there is a polynomial Q(x)
of degree max{0, p 1} satisfying the property below:
#{g G (, ) : g = n1 g1 + + np gp , n1 , , np Z,
(9.3)
|n1 | + + |np | n} Q(n),
for all n N. To see it dene the maps f : G Zp , f (n1 g1 + + np gp ) =
(n1 , , np ), and g : Rp R, g(x1 , , xp ) = x1 g1 + + xp gp . Clearly g
is linear and a left inverse of f therefore
{g G(, ) : g = n1 g1 + +np gp , n1 , , np Z, |n1 |+ +|np | n} =
{v Zp g 1 ((, )) : $v$ n}
where $(v1 , , vp )$ = |v1 | + + |vp |. But now it is easy to nd a poly-
nomial Q(x) of degree max{0, p 1} satisfying
#{v Zp g 1 ((, )) : $v$ n} Q(n).
This proves (9.3).
Next take a leaf W intersecting the support of at some point x. Again
we write W n instead of W n (x) for simplicity. Dene Win = W n Xi and

set i (n) = #Win for all i I. Certainly we have #W n = iI i (n). We
concentrate in an specic i (n) and set Win = {x1 , , xi (n) }. It follows
from the denition that there are a positive integer n1 n and a chain of
plaques (U1 , , Un1 ) such that x U1 and x1 Un1 . Take xj for some
2 j i (n) and a path j Xi from xj to x1 . Again by denition there is
another chain of plaques (U 1 , , U nj ) from x to xj with nj n. Evidently
U 1 = U1 and U nj = Un1 so the chain Cj = (U 1 , , U nj , Un1 1 , , U1 }
is closed. Moreover, ([Cj ]) = (j ([0, 1)) thus ([Cj ]) = ([Ck ]) for all
1 j < k i (n) since x belongs to the support of the -invariant measure
.
Now choose G = (1 (N ), a base {g1 , , gp } of G and = (Xi )
(which is positive for Xi intersects the support of ). It follows from the
remark above that there is a polynomial Q(x) of degree max{0, p 1}
satisfying (9.3). Notice that each ([Cj ]) belongs to G (, ) since j
Xi . Moreover, [Cj ] is represented Cj which is a closed chain of length
nj + n1 2n thus
[Cj ] = 1j ljj
% &
where 1 lj 2n d (here [] denotes integer part). Since {g1 , , gp } is
a base of G, and each (kj ) belongs to G, we can select integers njk,s with
1 k lj and 1 s p such that

p
(kj ) = njk,s gs
s=1
so

lj

p
lj

p
([Cj ]) = (kj ) = njk,s gs = mjs gs ,
k=1 s=1 k=1 s=1
where

lj
mjs = njk,s .
k=1
But G is a nite set, so there is an upper bound O for the set of integers
{|njk,s |}. Then,
' (
2n
|mjs | max{|njk,s |}lj Olj O
d
so each ([Cj ]) belongs to
) ' (*

p
p
2n
g G (, ) : g = ns gs , ns Z, |ns | pO .
s=1 s=1
d
Applying (9.3) we conclude that

2n
i (n) Q pO .
d

Then, we are done since i I is arbitrary and #W n = iI i (n).
Since leaves of polynomial growth of degree 0 are compact we obtain

the following corollary.
Corollary 9.1 (Plante, [Plante (1975)]). Every C 0 transversely ori-

ented codimension one foliation exhibiting a leaf of subexponential growth
on a closed manifold with rst Betti number 1 has a compact leaf.
Chapter 10
Currents, Distributions, Foliation

Cycles and Transverse Measures
10.1 Introduction
The next chapters of this text are dedicated to some other topics in the
Global Theory of Foliations. In this chapter, special attention is paid to
the consequences of the Theory of Currents on foliated manifolds. We will
therefore exploit aspects, already mentioned in the rst part, of growth of
leaves and of groups as well as the existence of invariant transverse measures
and of foliation cycles for a given foliation. Despite its certain informality
our approach and exposition aim to clear the key-points of some central
results of the classical theory (e.g. the bijection between transverse invari-
ant measures and foliation cycles and homological versions of Novikovs
compact leaf theorem) allowing this way the link between the classical real
framework and the so called Complex World, where the foliations are fre-
quently singular and therefore the ambient manifold may not be compact.
After constructing the bases of the theory of currents and foliation cycles
in the real case we address the problem of giving a non-geometrical (?)
proof of Novikovs compact leaf theorem. The central idea/philosophy is
that such a proof may be somehow adapted to the complex setting. Refer-
ences for these two parts should be essentially contained in the works of J.
Plante, D. Sullivan, S. Schwartzmann, D. Ruelle, A. Haeiger (for the real
classic part) and M. McQuillan, M. Brunella, for the existing complex part;
and may be found in the end of this text ([McQuillan (1998)], [McQuillan
(2001)], [Brunella (1997)], [Brunella (1999)], [Demailly (1997)]).
135
10.2 Currents
This section is inspired in the expositions of [Schwartz (1966)], [De Rham

(1955)] and [Griths and Harris (1978)]. The study of currents associated
to foliations has proved to be very useful in the comprehension of topological
dynamical phenomena related to foliations (cf. [Sullivan (1976)], [Schwartz-
mann (1957)], [Plante (1975)], [Haeiger (1981)] et al). In this chapter we
try to illustrate some of these applications. We shall begin with the ba-
sic denitions which are involved, with motivations coming from particular
situations already well-known. The rst step is to introduce the concept of
current. We denote by Cc (Rn ) the vector space of the functions C of
compact support f : Rn R. Endow Cc (Rn ), as usual, with the topology
of the uniform convergence in compact sets (for f and its derivatives of all
orders). A distribution in Rn is then a linear functional T (Cc (Rn )) ,
that is, a linear application T : Cc (Rn ) R which is continuous in the
C topology in Cc (Rn ). We denote now by Apc (Rn ) the R-vector space
of dierential p-forms of class C and compact support in Rn , equipped
with topology inherited from Cc (Rn ) in the natural way. Then Apc (Rn )
is complete and we can consider its topological dual Dnp (Rn ). In what
follows we take p + q = n.
Denition 10.1. A current of degree q on Rn is an element C Dq (Rn ).

Thus, a current of degree q on Rn is a linear continuous form on the space
of dierential forms of class C and degree p = n q having compact
support in Rn . Also we shall say that C is a current of dimension p.
10.2.1 Examples
Some basic examples are listed below:

1. A current of degree n in Rn is simply a distribution in Rn .
2. Let N p Rn be an oriented
submanifold of R . The integration along
n
N p denes a current C() := , Apc (Rn ) of dimension of dimension

N
p.

3. Let = j dxj (in ane coordinates (x1 , . . . , xn ) Rn ) be a dier-
j
ential q-form with locally integrable coecients (J L1loc (Rn )). To we
Currents, distributions, foliation cycles and transverse measures 137
Np
Fig. 10.1
Z
can associate a current of degree q (and dimension p) C() := ;
Rn
Apc (Rn ).
P
r
4. Given a singular p-chain = aj Nj in Rn we can (as in 2. above)
j=1
define an integration current by setting
Z
C() := , Apc (Rn ).

In {Apc (Rn )} we consider the exterior derivation of forms
d : Apc (Rn ) Ap+1
c (Rn )
7 d
and induce, in natural way, a derivation operator in Dq (Rn ):
d : Dq (Rn ) Dq+1 (Rn )
C 7 dC
dC() := C(d), Acq+1 (Rn ).
In a natural way we obtain a complex of cochains {d : Dp (Rn ) Dp+1 (Rn )}
(naturally) associated to the complex of De Rham with compact support
of Rn
{d : Apc (Rn ) Ap+1
c (Rn )}.
In particular, d(dC) = 0 for every current C in Rn .
We can localize the notions above in an obvious way: given open
subset U Rn we introduce the spaces Apc (U ) and Dq (U ) := (Apc (U ))
where the topology we consider is the natural inherited from the topology
of uniform convergence in compact parts (for functions and its derivatives
of all orders) in Cc (U ). Given a diffeomorphism C F : U V between
open subsets of Rn we have a natural homeomorphism F : Apc (V ) Apc (U )

which is also linear. Thus, we can introduce the spaces of currents Dp (M n )
in a dierentiable manifold M n . Let us see properties of the correspond-
ing complexes of currents {d : Dq (M ) Dq+1 (M )} and of De Rham
{d : Apc (M ) Ap+1
c (M )} in M . We recall:
A complex of cochains is a collection {dk : Ak Ak+1 }kZ of abelian
groups Ak and group homomorphisms dk : Ak Ak with the property that
dk+1 dk = 0. In particular we can consider the quotient groups
Ker(dk : Ak Ak+1 )
H k :=
Im(dk1 : Ak1 Ak )
called the Cohomology groups of the complex considered. The De Rham
cohomology groups with compact support of M (dierentiable manifold) de-
k
noted Hc,DR (M ) are dened this way (from {d : Akc (M ) Ak+1
c (M )}) for
k 1 recalling that, by denition,
+ C
0
Hc,DR (M ) := f : M R; f has compact support and df = 0}
is the number of compact connected components of M ; also we have

k
Hc,DR (M ) = 0, k n + 1 (n = dim M ) and we have the following:
)
R if M is orientable
Hc,DR (M ; R) =
n
0 if M is non-orientable.
Remark 10.1. We can also work with general dierential forms (not nec-
essarily with compact support) of class C in M obtaining the De Rham
k
complex of M , whose cohomology is denoted by HDR (M ), and the maximal
order cohomology is given by:
)
n 0 if M is non-compact or non-orientable
HDR (M ) =
R if M is compact and orientable.
Let us return to the currents in M . As we have seen in the above examples,

there exists a natural inclusion of the space of q-forms of class C in M in
the space of currents of degree q in M

C () := , Apc (M )
M

A (M ) q-form C
q
in M .
Such inclusion gives indeed a homomorphism of complexes

{ip : Aq (M ) Dq (M )} that induces by its turn a homomorphism in the
cohomology groups
q
ip# : HDR (M ) H q (D (M ))
where H q (D(M )) denotes the order q cohomology group of the complex of

currents D (M ) of M .
Theorem 10.1 (Theorem of De Rham, [De Rham (1960)], [De

Rham (1955)]] Given a differentiable oriented manifold M we have natu-
ral isomorphisms between the singular cohomology singular groups of Rham
and of currents in M :
q q
Hsing (M, R) HDR (M ) H q (D (M )).
10.3 Invariant measures
Let F be a foliation of class C dimension p and codimension q of a man-

ifold M n . There exists a cover U = {Uj }jN of M with the following
properties:
1. U is locally finite: given a compact K M we have #{j N; Uj
K 6= } < .
2. Uj is connected and F |Uj is trivial: there exists a diffeomorphism
j : Uj j (Uj ) Rn such that j takes F onto the horizontal foliation
in Rp Rq = Rn .
3. In each Uj we have an embedded disc Dq j Uj which is
transverse to the plaques of F in Uj and parametrizes this space of plaques.
We shall call U a regular cover of M for the foliation F . We also as-

sume, with no loss of generality, that j (Uj ) = Rn and that M =
S 1
j ((1, 1)n ) and we can then take j 1 n
j ((1, 1) ) and also re-
jN
name Uj = 1 n
j [(1, 1) ] in a way that:
4. Each leaf of F cuts some transverse disc Tj ; and if Ui Uj 6=
then each plaque of F |Ui meets at most one plaque of F |Uj defining local
diffeomorphisms C say gij : i j with the property that in Ui Uj
disco j
M
Uj
j
q
m p q
j Uj
p
Fig. 10.2
we have yj = gij yi where yj = is the projection of Uj onto j (via the

chart j ).
q
M
j
Uj p
j
i q
Ui i
p
Fig. 10.3
Clearly we have the following condition of cocycle:

1
5. Ui Uj = gij = gji and
Ui Uj Uk = gij gjk = gik
in the corresponding domains.
Denition 10.2. The holonomy pseudogroup of F for a regular cover U

is the pseudogroup U of local dieomorphisms C of the manifold U ,
disjoint sum of transverse discs j , generated by the local dieomorphisms
gij .
We recall the following definition:
Definition 10.3. Let X be a topological space Hausdorff and a collection

of local homeomorphisms g : U V where U, V X are open subsets of
X. Let us denote by Dom(g) and Im(g) the domain and the image of g
respectively. We say then that is a pseudogroup of local homeomorphisms
of X if:
(i) g we have g 1 , Dom(g) = Im(g 1 ) and Im(g) = Dom(g 1 );

(ii) If g1 , g2 and g : Dom(g1 ) Dom(g2 ) Im(g1 ) Im(g2 ) is a home-
omorphism such that g|Dom(gj ) = gj , j = 1, 2 then g .
(iii) Id : X X belongs to .
(iv) If g1 , g2 then g1 g2 com Dom(g1 g2 ) g11 (Im(g2 ))Dom(g1 ).
(v) If g and U Dom(g) is open then g|U .
Under these conditions we define the orbit of a point x X in the

pseudogroup by (x) := {g(x) X, g and x Dom(g)}.
We denote by c (X) the ring of subsets of X generated by the compact
sets. A measure in c (X) is said to be -invariant if:
(vi) is non-negative, finitely additive, and finite in compact sets.

(vii) g and any measurable subset A Dom(g) we have (g(A)) =
(A).
In the above case, of the holonomy pseudogroup of the foliation F relative

to the regular cover U we conclude that, in fact, U is a pseudogroup of local
diffeomorphisms C of U . In case we have another regular cover of M
relative to F , say Ue = {Uej } e
jN if we suppose that U is thinner than U, (i.e.,
for each index j N there exists an index (j) N such that U ej Uk(j) ),
and also that U ej is uniform in Uk(j) (i.e., each plaque of Uk(j) meets at
most one plaque of U ej ) then we obtain a natural identification between

the corresponding holonomy pseudogroups Ue U . This shows the
following (exercise!):
Lemma 10.1. All the holonomy pseudogroups U , where U is cover regular

of M for a foliation F , are naturally equivalent.
We may then introduce the (well-defined whether M is compact or not)

holonomy pseudogroup of the foliation F . This way we can formalize the
following notion:
Denition 10.4. A foliation F of a manifold M is said to admit a holon-

omy invariant transverse measure (or simply invariant transverse measure)
if its holonomy pseudogroup has some invariant measure (non-trivial) which
is nite in compact sets. The support of an invariant measure is the set
of points x M such that: given any transverse disc to F of dimension
q = codimension of F , Dq M with x Int(Dq ), we have (Dq ) > 0.
The support of , denoted by supp(), is closed and (since is invariant)
it is saturated (invariant) by F.
n
D
q
M
x F
Fig. 10.4
Let us see some examples illustrating the notions above:
10.3.1 Examples
Let us now see some examples of foliations and invariant transverse mea-
sures.
1. Let F be a foliation of codimension 1 given by a non-singular closed
1-form of class C , in M . Then it is to see from the Poincare Lemma
that the holonomy pseudogroup of F is naturally a group of translations
F (R, +) and any leaf of F has trivial holonomy group (a translation
with a nite xed point is the identity). Therefore, any Borel measure in
R invariant by translations is also F -invariant. Suppose now that F is of
codimension k and given by a closed k-form in M . In this case given any
transverse k-disc to F say Dk M , the restriction |Dk is a volume form
(assume also that F is transversely oriented this way) which is positive in
open sets. The fact that is closed implies that the transverse measure
this way induced is F -invariant.
2. Suppose that F admits a closed leaf L0 M . Given a transverse disc
to F , say D M we define for any A D a measure (A) := {A L}.

Clearly we obtain this way an invariant transverse measure for F ; also
we have supp() = L0 .
3. Let F be a foliation defined by the fibration M B of M over a
manifold B; then the transverse measures for F correspond to the measures
over B, which are finite on compact sets.
4. Let F be a foliation of M and f : M f M a proper application of class
C and transverse to F . Denote by Fe the lift f F to M f, if F admits
an invariant transverse measure then F admits an invariant transverse
measure e := (D), where D
e := f () defined naturally by (D) e = f 1 (D)
as in the figure below.
~ ~
M M
D D
Fig. 10.5
5. According to Joe Plante [Plante (1975)] if F is foliation of class C 2

of a compact manifold M admitting a leaf L0 with subexponential growth
(geometrical, Riemannian) then F admits an invariant transverse measure
, finite in compact sets, whose support supp() L0 .
6. Let us consider now more in details the case of suspensions: Let
: E B be a fiber bundle of class C with typical fiber F ; base B,
F
projection and total space E. We say that a foliation F of E is transverse
to the fibers of E if:
(a) Given x E we have which Lx is transverse to the fiber F(x) =

1 ((x)) and in fact dim F + dim F = dim E.
(b) The restriction |L : L B, where L is an arbitrary leaf of F , is a
covering map.
Remark 10.2. We observe that if the fiber F is compact then (b) follows
from (a); even for B non-compact (see [Camacho and Lins-Neto (1985)]
page 94).
F (x)
L x
(x) B
Fig. 10.6
Since each restriction |L : L B is a covering map we can dene a

representation : 1 (B) Dif (F ) from the fundamental group of the
base B into the group of (global) dieomorphisms of class C (suppose
F of class C ), of the ber F as follows: Fixed base points b0 , b0 B.
Given a path : [0, 1] B, (0) = b0 , (1) = b0 we dene, for each
y Fb0 , the point f (y) Fb0 , as the nal point nal y (1) of the lift
y : [0, 1] L0 of , by the covering map |Ly : Ly B, with origin at the
point y = y (0).
For b0 = b0 we identify Fb0 = Fb0 F and we obtain representation

: 1 (B, b0 ) Dif (F )
[] f[] .
The image of this representation is called global holonomy of F . By means
of a constructive process one may prove the following
I. Let F and F be foliations transverse to the bers of a ber bun-
dle : E B. Then the groups of global holonomy of F and F are
F
conjugate (in Dif (F )) if and only if F and F are conjugate by a bred
dieomorphism : E E.
II. Given a ber bundle space : E B there exists a foliation F
F
transverse to the bers of the bundle if and only if the structural group of
Fb Fb
0 0
f (y )
~ y
y
Ly

bo b0 B
Fig. 10.7
the bundle is discrete.

III. Given a representation : 1 (B) Dif (F ) of the fundamental
group of a manifold B in the group of C diffeomorphisms of a manifold
F with image G < Dif (F ) there exist a foliation F of class C of the
manifold E, a structure of fiber bundle space : E B such that F is
F
transverse to the fibers of the bundle and whose global holonomy (of F ) is
conjugate to G. By (I) F is unique up to natural equivalence.
We shall call such a foliation F the suspension of the representation
: 1 (B) Dif (F ).
We recall that a group G is amenable if the space B(G) := {f : G R; ||f ||
is bounded}, equipped with the norm of the supreme, admits a positive
linear functional : B(G) [0, +) with (1) = 1 and G-invariant, (i.e.,
(f Lg ) = (f ), f B(G). Such a functional is called continuous
invariant mean (cf. [Hirsch (1971)]). It is proven that if G is a finitely
generated group and with subexponential growth then G is amenable (cf.
[Thurston (1974)]) and that solvable finitely generated groups (for example)
are amenable.
Let finally F be a foliation transverse to the fibers of a fiber bundle
: E B and suppose that the group of global holonomy of F is
F
amenable then, if a fiber F is compact, F has invariant transverse mea-
sures.
10.4 Current associate to a transverse measure
This section is based in the exposition from [Godbillon (1991)]. Let F

be a foliation of codimension q in M n , class C , admitting an invariant
transverse measure . Let us see how to associate to a current C in
M ; we begin taking regular covering U = {Uj }jN of M relative to F and

considering the holonomy pseudogroup U . As before U = j denotes
jN
the space of plaques of F relatively to U; we can then disintegrate the a
measure as follows:
denes a Borelian measure over U invariant by U .

Let aj = 1 be a partition C of the unity, strictly subordinate to
jN
the cover U of M .
Given Anqc (M ) of class C and degree n q (n = dim M ) and
compact support in M we can consider the product aj Anq c (j ) as
a continuous function on j provided that F is oriented and we consider
in Uj (and therefore in j ) the orientation induced by F . In fact, we can
consider a function y aj
Py
Fig. 10.8
dened in terms of local coordinates (x, y) in Uj that make F |Uj {y = cte};

the plaques of F |Uj are of the form Py Rnq {y} and a transverse j
of the form j {x = 0}.
We integrate and sum these functions

obtaining
! the value

C () := aj (y) d(y).
jN j Py
Fig. 10.9
Using the fact that is invariant by the local diffeomorphisms gij : i

j we conclude that in fact the value of C () does not depend on the
P
aj = 1 (partition of the unity) neither on the regular cover U with
jN
respect to F (there is no need to suppose M compact).
Definition 10.5. C is the current associate to the invariant transverse

measure for F .
The following result is central in the theory:
Proposition 10.1. C is a closed current.
Proof. Using the above notations we have that

!
XZ Z
C (d) = (aj d)(y) d(y).
jN j Py
!
P P
On the other side, d = d aj = d(aj ) so that
j j
!
XZ Z
C (d) = d(aj d)(y) d(y).
jN j Py
Assume now that aj has compact

support in Py (not compact) so that,
by the Theorem of Stokes, d(aj ) = 0, y and thus C (d) = 0.
Py
By denition the derivative dC D(M ) is dened by dC () := C (d)
where Ac (M ). Therefore dC = 0, that is, C is a closed current.
Let us see some important consequences of this result:
As we have already seen, there exists an isomorphism of (De Rham)

cohomology groups
q
Hsing q
(M, R) HDR (M ) H q (D (M ));
therefore each closed current C Dq (M ) denes a class [C] in the space
q
HDR (M ). By the Duality Theorem of of Poincare, if M is orientable, we
, -
q p
have a natural isomorphism HDR (M ) Hc,DR (M ) = topological dual
space of the cohomology group (of degree p) with compact support in M ,
of De Rham. Hence, we can associate to C a class [[C]] in
)
, - Hp (M, R) if M is compact
p
Hc,DR (M )
H q (M, R) if M is orientable.
Thus, for M compact (respectively orientable) we have associate an in-
variant transverse measure for F , the homology class (respectively class of
cohomology) of this measure.
Let us see some examples:
10.4.1 Examples
1. If N p M n is an oriented submanifold compact of dimension p invariant
by F then the class of the current of integration corresponding to N is the
class [N ] of homology of N in Hp (M, R); note that N is a compact leaf of
F.
2. Let F be a foliation of dimension p and codimension q of M n . Assume
that F and M are oriented and that is F transversally oriented. The
dierential form of degree q in M such that for each transverse disc to
F , Dq M we have |Dq is the form of volume (positive for the induced
orientation in Dq ) is a transverse volume form of F in M . We can choose
a continuous vector eld XF of p-vectors on M such that in each point
x M we have Tx F = oriented space generated by XF (x).
M
Lx
x
TxF
Fig. 10.10
In this case we can obtain a transverse volume form positive F for F in

M of class C such that F (XF ) = 1 in M . We shall say that F is
normalized for XF . In a general way, given transverse volume
Z q-form
for F in M the associated current to is defined by C() = and
M
the homology class corresponding to C in H q (M, R) is the corresponding
q
class of in HDR (M ).
Pr
3. Let = aj Nj , aj Z; be a singular p-chain in M n and denote by
j=1
C the current of integration defined by in M ; if is closed ( = 0) then
C is closed (dC = 0) as consequence of the Theorem of Stokes. The class
of C in Hp (M, R) is the class of in this same space. In this example we
are not necessarily assuming the existence of a foliation in M which leaves
invariant.
4. Let now F be a foliation transverse to the fibers of the bundle
: E B in E; given an invariant transverse measure we have that
F
corresponds (in bijective way) to a Borelian measure 0 over the fiber
F which is invariant by the global holonomy Hol(F ) Dif(F ) of F , and
finite in compact sets of F . Let C be a current corresponding a ; then by
construction we have which
Z Z ! Z
C() = d(y) = f (y)d0 (y)
B F(y) B
where f : B R is defined by the integration of along the fibers (cf.

the figure below).
Thus, in order to study the class of C in H q (M, R) we can report to the

(class of the) measure 0 in H (B, R).
Fig. 10.11
Suppose now that the ber F is compact and let us study the homology class
of (one ber) [F0 ] in Hq (M, R). Take a tubular neighborhood : W F0
of this ber in E such that the projection has as bers (transverse discs)
the leaves of F |W .
E
F 0
W
F ( x)
Lx
x
em W
Fig. 10.12
We can assume W M compact and using bump functions we obtain

closed form Apc (M ) such that supp W and = 1, x F0
Dx
where Dx = r1 (x) is a ber of by x F0 .
But then we have C() = 0 ((W )) > 0 so that C() = 0. On the other
p
hand, since is closed we have that its class in Hc,DR (M, R) is dual to the
class of F0 in Hq (M, R) so that if [F0 ] = [0] in Hq (M, R) then [] = [0] in
x Dx Lx E
W
F0
Fig. 10.13
p
Hc,DR (M, R) and so C() = 0 giving a contradiction. This shows that the
class of F0 is not zero in Hq (M, R) . Since F0 is arbitrary we conclude the
same for any fiber of : E B. The same proof gives us:
6. Let F and M be oriented and N q M n compact submanifold without
boundary and transverse to F . If there exists invariant transverse measure
for F with supp N 6= then [N ] 6= [0] in Hq (M, R).
10.5 Cone structures in manifolds
In this section we shall follow [Sullivan (1976)]. Let E be a real locally

convex topological vector space. Given a convex cone C E we say that
C is a connected convex compact if there exists linear functional : E R
such that
1. (x) > 0, x C\{0}.
2. 1 (1) C is compact; called the base of the cone.

We denote by C the set of radii of of C; C is direct identification with its
base.
Definition 10.6. A cone structure in a closed subset F of a manifold

C M is a continuous field of convex compact sets cones, say {Cx }xF ,
in the vector spaces Xp (x) of tangent p-vectors in M (for x F ).
The continuity of the field {Cx }xF is defined in terms of the movement

of its bases C x for a suitable metric in the radii (see [Sullivan (1976)]).
Thus the p-form C in M is said to be transverse to the cone structure
Fig. 10.14
{Cx }xF if (x)(v1 , . . . , vp ) > 0, (v1 , . . . , vp ) Cx Xp (x) non-zero and

x F . Such transverse forms can always be constructed and determine
currents: given point x M we dene the Dirac current associate to the
xed p-vector X(p) Xp (x) by X,x : (X)(x); by choosing X(x) in
the cone Cx Xp (x) we obtain a collection of such Dirac currents which
gives us a closed convex cone of currents called cone of currents of structure
associate to the cone structure {Cx }xF . When F is compact the cone of
currents of structure associate is a compact convex cone. We shall call the
structural cycles of a cone structure in a manifold the structural currents
which are closed (in the sense of currents). It is proven (cf. [Sullivan
(1976)] 2)
that if F M is compact then any structural current C writes
as C = f d where is a measure 0 in F and f is an integrable
F
function -integrable taking values in Xp (M ) = {p-vectors in m} and such
that f (x) Cx (cone structure given originally, x F ).
Chapter 11
Foliation cycles:
A homological proof of Novikovs
compact leaf theorem
Let F be an oriented foliation of class C , dimension p and codimension

q in M oriented, XF a continuous eld of p-vectors generating T F and F
transverse volume form normalized for XF . Clearly F denes (via XF )
a foliation current of dimension p over M ; for each x M we denote by
CF (x) the convex cone in Tx M generated by the elds of p-vectors tangent
to F in x and denote by CF the cone structure over M obtained this way;
an element of the cone of currents of structure associate to CF is called a
foliation current of F. In other words, a foliation current of F is an element
do convex cone closed do space of currents of dimension p over M which is
generated by Dirac currents of the form X,x : (X)(x) where x M
and X is a p-eld tangent to F.
Denition 11.1. A foliation cycle of F in M is a foliation current of F

which is closed in the sense of currents, that is, a structural cycle of CF .
Owing to the preceding chapter if is an invariant transverse measure for

F in M then a current associate to is a foliation cycle of F . The converse
in the compact case was proven by D. Sullivan (cf. [Sullivan (1976)]):
Let F be a foliation C of M compact and suppose F and M oriented.
Then each foliation cycle for F in M comes (via the construction already
presented) of a unique invariant transverse measure for F.
We dene the support of a current in the obvious way and we can then ob-
serve that if C is a foliation cycle for F , coming from an invariant transverse
measure in M , then supp(C) = supp() M ; in particular supp(C) is
closed and F -invariant in M .
153
11.0.1 Examples
Some examples of currents are given below:

1. All examples of currents (foliation) constructed from invariant measures
in 2.4 give examples of foliation cycles.
2. If is a Borelian measure (positive

not necessarily F -invariant) over M
a current of integration C : (XF )d is a foliation current for F ;
M
by the above Theorem of Sullivan C is a foliation cycle if, and only if,
is F -invariant.
3. Let F foliation transverse to the bers of the bundle : E B with

F
global holonomy Hol(F ) Dif(F ), then given any Borelian measure 0 in
B a current C associate to the measure dened by 0 em E is a foliation
current (as in 2. above) which is a foliation cycle if, and only if, 0 and
Hol(F )-invariant.
4. Let F foliation of codimension 1 in M ; according to Haeigers theorem

(Chapter 5) if F has a transverse closed curve homotopic to zero in M
then there exists leaf L0 of F and loop (of holonomy) 0 1 (L0 ) with
holonomy f0 Dif((, ), 0) such that f |(,0] is the identity and f |(0,+)
is increasing.
Fig. 11.1
Foliation cycles: A homological proof of Novikovs compact leaf theorem 155
Such a leaf L0 we will call in general a ressort leaf, more precisely we have:
Definition 11.2 ([Godbillon (1991)] Chapter IV, page 228). A leaf

L0 of a transversely orientable codimension one foliation F is ressort (from
the French word for spring) if it contains a loop 0 whose corresponding
holonomy map h0 Diff(R, 0) is not decreasing and is, in at least of the
sides of R \ {0}, a contraction accumulating L0 at itself.
Let us see how the existence of an invariant measure for F restricts the
existence of ressort leaves.
Proposition 11.1. Let F be a transversely orientable codimension one

smooth foliation on a compact manifold M . The support K of an invariant
transverse measure for F does not contain a ressort leaf.
Proof. First we observe that K is invariant (because is invariant), closed

(by definition) and any leaf contained in K is dense in K (indeed, K is
support of ). Since the measure is finite in compact sets we conclude
that K does not contain a ressort leaf.
The above proposition is enough for our purposes, i.e., the homological
proof of Novikovs compact leaf theorem. Nevertheless, under the hypoth-
esis of Proposition 11.1, it is possible to say more.
Theorem 11.1. Let F be a smooth transversely orientable codimension one

foliation on a compact manifold M , equipped with an invariant transverse
measure for F . Let K = supp() M be the support of . Then:
(1) Either K = M , or K is a union of compact leaves and finitely many
(2) If K = M then all the leaves of F have trivial holonomy.
(3) If F has class C 2 then K contains no exceptional minimal set, only
compact leaves.
The proof of Theorem 11.1 requires some features from and more knowl-
edge on the structure of codimension one foliations as Dippolitos theory
on semi-stable leaves and Cantwell-Conlons theory on minimal sets of C 2
codimension one foliations on closed manifolds. This is partially done in
the appendix (see Appendix A). For Dippolitos theory we give the main
steps, while for Cantwell-Conlons theory we suggest their book [Candel
and Conlon (2000)].
11.0.2 Homological proof of Novikovs compact leaf

theorem
Note that above we strongly use the fact that F is of codimension 1. Sup-
pose now that dim F = 2 and dim M = 3 so that F of codimension 1. We
will also assume M compact and that F has a vanishing cycle, say, in the
leaf L0 of F . We will show how to construct the foliation cycle for F; there
is no loss of generality if we assume that the vanishing cycle is simple: recall
that (cf. Chapter 6) a vanishing cycle of F in the leaf L0 consists of a lace
(closed) 0 : [0, 1] L0 such that it extends to a continuous application
C0
: [0, 1] [0, 1] M with the following properties:
(i) Given t [0, 1] the application t : [0, 1] Lt , t (s) = (t, s)
denes a loop in the leaf Lt of F .
(ii) 0 is the loop originally given in L0 .
(iii) 0 is not homotopic to zero in L0 but t is homotopic to zero in
Lt t (0, 1].
(iv) Fixed s [0, 1] the curve Cs : [0, 1] M t t (s) is transverse
to the foliation F .
curve C x
Fig. 11.2
The vanishing cycle is called simple when also we have

t of the leaf
(v) the lift of t , denoted by t , to the universal covering L
Lt is, for each t = 0, the closed curve (because t 0 in 1 (Lt )) which is
simple (that is, without self-intersection).
^t
^t
L
Cannot occur
{ Lt
t
t >0
Cannot occur
{ Fig. 11.3
t
^t
L
Lt
We can approximate continuous functions by functions C 1 so that we can

assume that : S [0, 1] M is of class C 1 .
Note that the universal covering L b t of Lt is necessarily (diffeomorphic to)
R because otherwise F would have some leaf covered by S 2 , this leaf would
2
be compact and being orientable it would be the sphere with g 0 aisles; if

g = 1 the universal covering is R2 and if g 2 then the universal covering
(as a Riemman surface) is the unit disc D R2 so we must have g = 0 and
the leaf would be diffeomorphic to S 2 .
By Reeb global stability theorem F would be a compact fibration over
the circle S 1 with fibers S 2 and in this case it could not have vanishing
cycle (all the leaves would be simply-connected).
Now, since each leaf Lt is covered by L b t R2 each (simple) curve t
b b b
in Lt is boundary of a disc Dt Lt this allows us to obtain an immersion
C 1 , A : D2 (0, 1] M of the solid cylinder (not compact) D2 [0, 1] in
M with the following properties:
(vi) At |S 1 =D2 = t , t (0, 1] and the image At (D2 ) Lt ,
t (0, 1].
(vii) Given an oriented transverse ow X F chosen from the

beginning from the transverse orientability of F in M we have which
At : D2 Lt dene, for 0 < t 1, lift of A1 : D2 L1 by the

transverse ow X
M
2 A
D {t}
t
X
Lt
D2 (0,1]
X
Lt
t
At X X
2 A1
D 1
L1
X F
folhas de F{
Lt
L1
Fig. 11.4
Since 0 is not homotopic to zero in L0 and since for each x S 1 the curve
[0, 1] M , x A(x, t) has a limit when t 0+ we conclude that
(viii) The set W = {x D2 ; t A(x, t) has a limit when t 0+ }
which is an open neighborhood of S 1 in D2 with S 1 W D2 .
We denote by Ct , for each 0 < t 1, the foliation current of F dened
by the integral

1
Ct () := , where Apc (M )
vol(Dt ) Dt
(note that still we are not using the fact that M is compact, which will be
used in what follows), and where Dt = At (D2 ) Lt (note which At : D2
Dt Lt M is a C 1 immersion). We obtain then, using the terminology of

[Sullivan (1976)], a family of Plante of foliation currents {Ct }t(0,1] defined
by the properties below which can be easily verified:
(ix) Each Ct has mass 1 so that {Ct }t(0,1] is a pre-compact family of
currents (weak topology).
(x) Each accumulation C of {Ct }t(0,1] is necessarily a foliation cycle of
F : in fact if tn 0 is such that Ctn C then the fact that the quotient
length(Dtn )
0 implies that the mass of the derivative dCtn satisfies
Area(Dtn )
mass (dCtn ) 0 and therefore mass(dC) = 0. We obtain then foliation
cycles C for F in M 3 , if we suppose M 3 compact, from the existence of a
vanishing cycle for a leaf L20 of F , foliation of codimension 1 in M 3 .
Such facts have been generalized by D. Sullivan for higher dimension

with the notion of vanishing cycle of dimension = dim F 1 (cf. [Sullivan
(1976)]).
We can now conclude the following:

F foliation orientable and transversally orientable C 2 of

codimension 1 of M 3 compact, F with a leaf L0
containing a vanishing cycle

F has a foliation cycle C whose support contains a leaf L0

" #
F admits an invariant transverse measure whose support
contains L0 (in fact supp() = supp(C))

since F is of codimension 1 and M compact we have that K = supp()
(contains L0 ) does not contain ressort leaf and is contained in the union of
minimal sets of F in M . Thus K is a union of compact leaves and hence
L0 is a compact leaf of M that is,

M has L0 as compact leaf.
This ends the homological demonstration of Novikovs compact leaf
theorem.


Appendix A
Structure of codimension one

foliations: Dippolitos theory
In what follows, we study the structure of codimension one foliations. Our

aim is to introduce the results of Dippolito and Cantwell-Conlon on semi-
stability, minimal sets and structure of codimension one foliations. On this
course we pave the way to prove Theorem 11.1.
A.1 Semi-proper leaves, Dippolitos semi-stability
In what follows F is a codimension one transversely orientable foliation on

a manifold M , not necessarily compact. We start with the denition of
proper leaf.
Denition A.3. A leaf L of F is proper if it is locally path-connected.
It is an exercise to show that a leaf L of F is proper if, and only if, its
topology coincides with the one induced by M . Or else, that a leaf L is
proper if, and only if, L does not cluster on itself (see [Dippolito (1978),
p. 408] or [Candel and Conlon (2000), Def. 4.3.3]). A compact leaf is
always proper. Now we dene semi-proper leaves. There are many ways of
dening semi-proper leaves. We follow the nice exposition in [Seitoh (1983)]
page 96. Since F is smooth and transversely orientable, we can choose a
one-dimensional smooth foliation T which is everywhere transverse to F .
Denote by (T ) : M M , the unit tangent bundle to T . Given a leaf L
of F , the sides of L are the leaves L of the foliation (T ) (F ) that project
onto L, i.e., (T )(L) = L.
Denition A.4. A side L of a leaf L of F is a proper side if a transversal

curve : [0, 1] M to L, with starting from L in the direction L, satises
(0, )L = for some > 0. A leaf with a proper side is called semi-proper.
In this case this side is called positive side of the leaf.
161
In other words, a leaf is semi-proper if it is proper or does not cluster

on itself from one side, which is called positive (see [Candel and Conlon
(2000), p. 118]).
The following equivalent formulation is left as an exercise to the reader:
Lemma A.1. For a leaf L of a codimension one transversely orientable

foliation F , the following conditions are equivalent:
(1) L is semi-proper.
(2) Either L is proper or it is exceptional and such that for every leaf T of
T , the intersection points L T are the extremities gaps of L T , i.e.,
of the connected components of T \ (L T ).
Notice that, because F is oriented by T , the extremities of the gaps of

L T are all on a same side of L, the side where L is semi-proper.
Now we state a notion of semi-stability:
Denition A.5. A semi-proper leaf L of F is semi-stable on the positive

side if there exists an open connected subset U M , which is a foliated
product by F times T , having L as a positive boundary leaf. A semi-proper
leaf L is attracting on the positive side if there is an open connected subset
U invariant by F , having L as positive boundary leaf, and such that L is
the adherence of every leaf of F in U .
The equivalent formulation found in [Candel and Conlon (2000)] is:
Proposition A.2. A leaf L of F is semi-stable if it is semi-proper and on

the proper side of L and in a transverse arc J on this side, that meets L
only at a point x0 L, the xed points of the holonomy of L cluster at x0
(by denition, a point x J is a xed point of the holonomy if, for every
loop L based at the point x0 , the corresponding holonomy map h either
is not dened at the point x J or it xes the point x J, i.e., h (x) = x,
see [Candel and Conlon (2000), p. 134]).
Now we introduce the notion of foliated product:
Denition A.6. Given an open F-invariant subset U M , where F is

a transversely oriented codimension one smooth foliation on M m , we shall
say that U is a foliated product (with respect to a one-dimensional positively

oriented transverse foliation T to the foliation F ) if the restriction T U is
a bration by open intervals over a certain manifold of dimension m 1.
Structure of codimension one foliations: Dippolitos theory 163
Exercise A.1.1. Show that in the definition above there is a smooth diffeo-
morphism taking U onto the product N (1, 1), where N is a m 1-
dimensional manifold. Also show through an example that the foliation F
does not need to be diffeomorphic to the horizontal product foliation on
N (1, 1).
Let us now recall Dippolitos semi-stability theorem in [Dippolito (1978)]

(see also [Candel and Conlon (2000), sections 5.2, 5.3], [Godbillon (1991),
section IV. 4] and [Hector and Hirsch (1987), Chapter V, sections 3, 4]).
Theorem A.2 (Dippolitos semi-stability). Let L be a leaf of F which

is semi-proper on the positive side. If the holonomy group of L is not
attracting on this positive side, then there exists a sequence {Un }nN of
nested open connected subsets Un M , such that:
(1) Each Un is a foliated product of F times T ;

(2) L is the leaf positive boundary of each of the open sets Un ;
T
(3) Un =
nN
Therefore, if the holonomy group of L is not attracting on the positive side

then L is semi-stable on the positive side.
Even though it may seem natural, the well-known proof of Dippolitos

semi-stability requires some deeps result, which is developed in the next
section. This is also useful in the proof of Theorem A.2.
A.2 Completion of an invariant open set
We shall present the celebrated Dippolitos structure theorem. In order

to do this we need to introduce the notion of completion of an invariant
open subset of a closed foliated manifold. We follow [Godbillon (1991)]
and [Candel and Conlon (2000)]. We begin with a brief discussion of the
standard theory of saturated open sets, that we use in the sequel. A more
detailed exposition is found in the book of Candel and Conlon ([Candel and
Conlon (2000)] Chapter 8).
We consider a codimension one, non-singular foliation F of class C in
a closed C manifold M . Given such a pair (M, F ), we denote by O(F ) the
set of all open F -saturated subsets of M . We assume that F is transversely
oriented, and we let L be a one dimensional oriented foliation, defined by
a smooth non-singular vector field transverse to F .
Let U O(F ) be connected; x a Riemannian metric on M and take

its restriction to U . Let d : U U [0, ) be the induced topological
metric, and denote by U . its completion with respect to this metric. One
has (see propositions 5.2.10 to 5.2.12 in [Candel and Conlon (2000)]):
Proposition A.3. i) The space U . is a complete connected, C manifold

with nitely many boundary components, and its interior Int U . is dieo-
morphic to U .
ii) The manifold U . has a foliation F. induced from that in U , and the
inclusion i : U M extends to a C immersion .i : U. M that carries
.
leaves of F dieomorphically onto leaves of F .
. ) be the image of the boundary of U
iii) If we let U = .i( U . , then U is a
union of leaves of F , and if L is a leaf in U then .i (L) consists of one
1
..
or two leaves in U
iv) There is also an induced oriented foliation L. on U . , dened by a vector
. .
eld transverse to F , which is carried by i into the foliation L.
Denition A.7. The manifold U . is the (abstract transverse) completion of

U O(F). The set (U ) is the border of U ; the leaves in U are the border
leaves of U .
That is, a border leaf L U is the image under .i of a leaf in the

boundary of the completion U . , and there are at most two such leaves in
U. corresponding to L.

A biregular cover of (M, F , L) is a cover M = U by open sets U
equipped with coordinates (x , y ) U , such that F U and LU are
trivial, the plaques of F in U are the level sets of y and the plaques of L
in U are the level sets of x . Every foliated manifold (M, F, L) admits a
biregular cover, which can be taken to be nite if M is compact.
Recall that (cf. Denition A.6) U O(F) is a foliated product (with
respect to L) if the restriction LU bers U by open intervals over some
(m 1)-manifold N . Since L is dened by a vector eld, this bundle is
trivial. Thus U is dieomorphic to N (0, 1), but the restriction F U is
not necessarily a product foliation.
One has that a leaf L F is semi-proper if and only if it is a border
leaf of some U O(F ) (see [Candel and Conlon (2000), Lemma 5.3.2]).
Therefore each component of U . is identied with a semi-proper leaf of F ;
some pairs of components may be identied with a same leaf.
Using this we can restate Dippolitos semi-stability theorem as follows

([Dippolito (1978)], see also [Candel and Conlon (2000), sections 5.2, 5.3],
[Godbillon (1991), section IV. 4] and [Hector and Hirsch (1987), Chapter
V, sections 3, 4]).
Theorem A.3 (Semi-stability theorem of Dippolito). Let L be a

semiproper leaf which is semi-stable on the proper side defined by the trans-
verse arc J = [x0 , y0 ). Then there is a point y1 J \ {x0 } such that
the F -saturation U = SatF ((x0 , y1 )) is a foliated product having as bor-
der leaves the (distinct) leaves through x0 and y1 . Also, there exists a
sequence {yk }k=1 (x0 , y1 ] converging monotonically to x0 , such that the
b
leaf Lk yk is carried by the L-fibration b L homeomorphically onto
: U
L, for all k 1.
For the interested reader, we shall give a sketch of the proof of the above
theorem. Our proof will then be based on an equivalent construction of the
completion of an open invariant subset, which we pass to present.
A.2.1 Completion of an invariant open set - revisited
Now we shall recall an equivalent way of introducing the completion of an

invariant open subset of a foliated manifold. More precisely, we have: M
a closed (compact and without boundary) manifold, F a transversely ori-
ented codimension one foliation of class C r , r 1. Also consider a C r one-
dimensional foliation transverse to F and oriented according to the trans-
verse orientation of F . We take a finite biregular covering U = {U1 , ..., Us }
of M , with respect to F and T . We can assume that for each i {1, ..., s},
there is a diffeomorphism i : Vi Rm = Rm1 R, defined in some neigh-
borhood Vi of Ui in M , such that (U1 ) = (1, +1)m . Moreover, this
diffeomorphism takes F and T onto the horizontal and vertical foliations
respectively, always preserving the orientations. The following lemma is
proved in the standard manner, we leave it to the reader:
Lemma A.2. If U M is a proper invariant open subset then U can be

chosen such that the following properties are also true:
(1) The external plaques 1 (Rm1 {a}, a = 1 are not in the bound-
ary of U ;
(2) There is N r s such that the open set Ui is contained in U if, and
only if, i > r.
(3) For i r the intersection U Ui contains a countable number of con-

nected components. These are denoted by Uij , except for two of them,
which adhere to the upper and lower plaques (i.e., the external plaques
in the obvious sense). These two plaques are denoted by U + and U
respectively.
Now we denote by Pij the upper plaque of Ui bounding Uij and by Pij+
the lower plaque of Ui bounding Uij . Finally, we denote by Pij the upper
plaque of Ui bounding Ui and by Pi= the lower one for Ui+ .
By gluing the manifolds U .i = Ui for i > r, the manifolds U . :=
i
. . = P U for i r, by the
Ui Pi , Uij := Pij Uij Pij and U
+
i
+
i
+ +
i
gluing dieomorphisms j1 i , we obtain a m-dimensional manifold with
boundary U . , of class C r , equipped with a codimension one foliation F,.
.
transversely oriented, with a 1-dimensional transverse foliation T , of class
C r . The result is:
Proposition A.4. The above construction gives a manifold U . and foli-

. .
ations F , T , which are homeomorphic to the completion of the invariant
open subset U by F and T , dened by a Riemannian metric on M , as
introduced above.
We denote by U . the boundary of U . . Notice that the lifted foliation

F. is tangent to this boundary. Therefore, the foliation T. is transverse to
U . for the subset of the boundary where the orbits of T.
. . We can put + U
.
points inwards to U and by U . the other part.
We end this paragraph with the statement of Dippolitos structure the-
orem:
Theorem A.4. Let F be a codimension one foliation on M , U O(F) a

proper subset, with completion U . , equipped with the foliation F. and trans-
.
verse foliation T . Then, there are m-dimensional submanifolds with corners
K and B of U . (one of them may be empty), invariant by T. such that:
. = K B.
(1) U
(2) K is compact and connected while B has no compact connected compo-
nent.
(3) C = K U . is a compact m 1-dimensional subvariety and it is the
saturated of the boundary C by T. .
(4) B has only nitely many connected components, each one having as
boundary a connected component of D.
(5) The intersection A = B + (Ub ) is a m 1-dimensional subvariety

of the boundary U and the orbits of Tb define a fibration by compact
b
intervals of B onto A.
Proof. Let us use the notation introduced in the second construction of

the completion of an invariant open subset above. We denote by V U b the
union of open subsets Ubij where i r. Then V Ub is an open, Tb -invariant
subset. Moreover, it is easy to see that given the set Ubij , its projections
+ b
onto the external plaques Pij along the orbits of T , define a fibration by
compact intervals, of V onto the intersection V + U b . Since U b \ V is
compact, we can choose a compact submanifold with boundary S + U b,
of dimension m 1, such that:
b S has a finite number of connected components, each of these
(1) + U
corresponds to a connected component of the boundary S. Each such
component is non-compact. They are called ends.
b is a neigh-
(2) The intersection of S with each connected component of + U
borhood of the intersection (Ub S) U
+ b.
b S and B as the part of

With this, we define A as the adherence of + U
V above A. We also get K as the adherence of U b B, if we choose S such
that K is connected.
b while the connected components of B are

We call K above as the core of U
b
the arms of U defined by the core K.
Using this notion we can redefine an invariant connected open subset
b is fibred
U O(F ) as a foliated product (by F and T ) if its completion U
in compact intervals by the orbits of Tb .
Then we have:
Lemma A.3. Given a connected open subset U M which is a foliated

product for F and T and its boundary leaves L1 , L2 , we have that:
The limit sets of L1 and L2 coincide and are contained in the limit set
of any leaf L of F in U .
If one of this leaves is exceptional then the same holds for the other and
we have L1 = L2 L.
The proof is quite standard and we leave it to the reader.

A.3 Proof of Dippolitos semi-stability theorem
Now we are in conditions to prove Dippolitos semi-stability theorem.
Proof of Theorem A.2. We preserve the notation of the preceding sec-

tions. Given a point p L we denote by T + (p) the positive semi-orbit
of the transverse foliation T , starting at p. It is easily veried that, the
saturation of an open arc J T + (p), with origin at p, and not intersecting
L, is a connected open set U M , having the leaf L as a positive boundary
leaf. Now we use the Structure theorem (Theorem A.4). Denote by U . the
completion of U and consider the intersection K(L) := K L, . where K
. . +.
is the core of U , and L is the leaf of U that corresponds to L. Then
K(L) L . is a compact, connected, m 1-dimensional submanifold with
boundary. We may also assume that K(L) is not empty, indeed that it
. above p.
contains the base point p of L,
Now a few considerations more and we are done. Denote by H :=
Hol(F , L, p) the holonomy group of L based at p, with transverse section
given by the orbit T (p) of T , through p. If H is not attracting from the
positive side, then there is a monotonous sequence (pn )nN of points pn
J, converging to p, such that each pn is xed by all the elements of the
group H which are dened at pn (this means that the element admits a
representative which is dened in an interval contained in J, that contains
p and pn ). Because K(L) is compact, for each n big enough, we can lift
K(L) to a submanifold with boundary Kn in the leaf L 7pn pn , that projects
.
dieomorphically onto K(L), along the orbits of T . The intersection of
Kn with each arm of U . is connected. Therefore, we can also lift to L 7 pn
each of the submanifolds of L . corresponding to the basis of these arms.
This way we obtain by means of projection along the orbits of T. , a global
dieomorphism of L . Therefore the open subset Un U bounded
. n onto L.
by the leaves Ln and L, which corresponds to the saturation of the subarc
Jn J between p and pn , is a foliated product with respect to F and T .
The proof is nished.
A.4 Guided exercises: Cantwell-Conlons theory
In what follows we give a sketched proof of Theorem 11.1.
Exercise A.4.1. In this exercise we shall prove the following result due to
Cantwell-Conlon (see [Candel and Conlon (2000)] Chapter 8, pp. 190-192):
Theorem A.5 (Cantwell-Conlon [Cantwell and Conlon (1981)]).

Let F be a codimension one smooth foliation of a connected manifold M
and U M an invariant connected open subset. Then we have:

(1) The union of minimal sets of the restriction F U is closed in U and
contains only finitely many exceptional minimal sets of F ;
U

(2) The union of closed leaves of F U is closed in U ;
(3) Given a leaf L of F in U , the closure L contains only a finite number
of minimal sets of F U .
In order to prove this result we introduce a pre-order in the leaf space

M/F of a codimension one foliation F on a manifold M . Given leaves
L1 , L2 of F we put L1 L2 if the closures are related by L1 L2 . If we
consider the class of a leaf L as the set [L] of all leaves having the same
closure than L then we have in the quotient space an order relation. We
can also write L1 < L2 to denote that L1 ( L2 .
Recall that we are dealing with codimension one foliations:
Lemma A.4. The leaves in the class [L] of a leaf L, have all the same
nature regarding F . More precisely, the leaves in [L] are either all proper,
or all locally dense, or all exceptional.
We define the inferior structure of a leaf L as the union inf(L) of all leaves
L such that L < L. The superior structure sup(L) is defined in a natural
analogous way.
The following is left as an exercise.
Lemma A.5. For a leaf L we have L = [L] inf(L) as a disjoint union.

The superior structure sup(L) is an open invariant subset of M .
The next result states an isolation property for exceptional minimal

sets.
Proposition A.5. If a leaf L is exceptional such that the class [L] is a

minimal set, then [L] sup(L) is an open subset of M .
Proof. We leave the details to the reader. We will show that if {Ln }nN
is a sequence of leaves in the complement Ln M L and converging to L
then, for n big enough, we have Ln sup(L). Notice that such sequences
exist provided that L is nowhere dense in M . Denote by U the connected
component of the open subset M inf(L) that contains L. We can assume
that the leaves Ln are contained in U . The class of the leaf L is denoted
by [L]. Let us denote by Wn the connected component of the open subset

U [L], containing the leaf Ln . Then Wn has a boundary leaf say L0n in
the class [L].
Now we have distinct cases to consider:
1. If one of the open sets Wn , say Wno contains innitely many leaves
L0n , these leaves converge to L0no and for n big enough are all contained in
sup(L) (it is enough to choose a sub-sequence of leaves such that the leaves
in this subsequence intersect a same arm of a Dippolitos decomposition (cf.
Theorem A.4) of the completion W 7no of Wno , whose saturated is contained
in sup(L)).
2. If one of the open sets say Wn1 is a foliated product then we have
L = L0 n1 Ln1 (this is the content of Lemma A.3).
Finally, we nish by observing that the open set U [L] has only a nite
number of connected components which are not foliated products.
Proof of Theorem A.5. Item (1) is a direct consequence of the fact that
sup(L) is an open invariant subset of M . Dippolitos semi-stability theo-
rem A.2 then shows that the exceptional minimal sets are isolated implying
(2). The other properties then follow.
Exercise A.4.2. Finally we are in conditions to prove Theorem 11.1, whose

statement we repeat here:
Theorem 11.4 Let F be a smooth transversely orientable codimension one
foliation on a compact manifold M , equipped with an invariant transverse
measure for F . Let K = supp() M be the support of . Then:
(1) Either K = M , or K is a union of compact leaves and finitely many

(2) If K = M then all the leaves of F have trivial holonomy.
(3) If F has class C 2 then K contains no exceptional minimal set, only
compact leaves.
Proof of Theorem 11.1. In order to prove the rst part we state:
Claim A.1. A leaf L0 K cannot accumulate from a proper side, at a

semi-proper leaf L1 which is attracting on this side.
Proof. The claim is a consequence of Dippolitos semi-stability result

(Theorem A.2) and of the following statement, whose proof we leave to
the reader:
Given a locally compact topological space X and a pseudogroup of local

homeomorphisms of X preserving a locally finite measure we have: a
point x K := supp() which is fixed by a contraction : A B of ,
has a discrete and closed orbit (x) by in X .
We recall that : A B is a contraction at the fixed point x if (A) =
T n
B A and (A) = {x}.
nN
Moreover, the support cannot contain either a leaf L0 that accumulates at

a semi-proper leaf L1 from the non proper side.
Now we recall that an exceptional minimal set is contained in the ad-
herence of every leaf which is close enough (Proposition A.5). This shows
that if K 6= M then K is contained in a union of compact leaves and
finitely many exceptional minimal sets (Theorem A.5). This ends the proof
of Theorem 11.1.


Bibliography
R. Bott, W. L. Tu, Dierential forms in algebraic topology. Graduate Texts in

Mathematics, 82. Springer-Verlag, New York-Berlin, 1982.
M. Brunella; A Global Stability Theorem for transversely holomorphic foliations;
Annals of Global Analysis and Geometry 15, 179-186, 1997.
M. Brunella; Courbes entieres et feuilletages holomorphes, Enseign. Math. 45
(1999), 195-216.
C. Camacho, A. Lins Neto; Geometric theory of foliations, Translated from the
Portuguese by Sue E. Goodman. Birkhauser Boston, Inc., Boston, MA
(1985).
Candel, A., Conlon, L., Foliations. I. Graduate Studies in Mathematics, 23. Amer-
ican Mathematical Society, Providence, RI, 2000.
J. Cantwell and L. Conlon: Poincare-Bendixson theory for leaves of codimension
one, Trans. Amer. Math. Soc. 265 (1981).
P. Dippolito: Codimension one foliations of closed manifolds, Ann. of Math. 107
(1978), 403453. MR0515731, Zbl 0418.57012.
J.-P. Demailly; Varietes hyperboliques et equations dierentielles algebriques, Gaz.
Math. No. 73 (1997), 323.
de Melo, W., Palis, J., Geometric theory of dynamical systems. An introduction.,
Translated from the Portuguese by A. K. Manning. Springer-Verlag, New
York-Berlin (1982).
do Carmo, M., P., Riemannian geometry, Translated from the second Por-
tuguese edition by Francis Flaherty. Mathematics: Theory & Applications.
Birkhauser Boston, Inc., Boston, MA (1992).
G. De Rham; Varietes Dierentiables, Paris: Hermann, 1960.
G. De Rham; Varietes dierentiables. Formes, courantes, formes harmoniques.
Paris: Hermann 1955.
R. Edwards, K. Millet, D. Sullivan; Foliations with all leaves compact; Topology
16, (1977), 13-32.
D.B.A. Epstein; Foliations with all leaves compact; Ann. Inst. Fourier, Grenoble
26, 1 (1976), 265-282.
C. Ehresmann; Sur les sections dun champ delements of contact dans une variete
dierentiable. C. R. Acad. Sci. Paris 224, (1947). 444445.
173
C. Ehresmann, G. Reeb; Sur les champs delements of contact, C.R. Acad. Sci.
Paris, 218, 1944.
E. Ghys; Holomorphic Anosov Systems, Inv. Math. 119, 585-614 (1995).
C. Godbillon; Feuilletages: Etudes Geometriques, Birkhauser, Berlin 1991.
P. Griths, J. Harris; Principles of Algebraic Geometry, John Wiley & Sons,
New-York, 1978.
R.C. Gunning; Introduction to holomorphic functions of several variables, vol. I,
Wadsworth & Brooks/Cole, California 1990.
A. Haeiger; Structures feuilletees et cohomologie a valeur dans un faisceau of
groupodes. Comment. Math. Helv. 32 1958 248329.
A. Haeiger: Travaux de Novikov sur les feuilletages, Seminaire Bourbaki 20e
annee, 1967-68, Num. 339, p. 433-444, 1967-68.
A. Haeiger; Some remarks on foliations with minimal leaves. J. Dierential
Geom. 15 (1980), in the. 2, 269284 (1981).
V. L. Hansen; Braids and Coverings. London Mathematical Society Student Texts
(No. 18), Cambridge University Press, 1989.
Hector, G. and Hirsch, U.; Introduction to the geometry of foliations. Part B.
Foliations of codimension one., Second edition. Aspects of Mathematics,
E3. Friedr. Vieweg & Sohn, Braunschweig (1987).
Herstein, I., N., Topics in algebra, Second edition. Xerox College Publishing,
Lexington, Mass.-Toronto, Ont. (1975).
M. Hirsch; Stability of compact leaves of foliations. Dynamical systems (Proc.
Sympos., Univ. Bahia, Salvador, 1971), pp. 135153. Academic Press, New
York, 1973.
M. Hirsch, S. Smale, Dierential Equations, Dynamical Systems and Linear Al-
gebra, Accademic Press, San Diego, CA, 1974.
M. Hirsch, W. Thurston; Foliated bundles, invariant measures and at manifolds.
Ann. Math. (2) 101 (1975), 369390.
H. Holmann; On the stability of holomorphic foliatons; Springer Lect. Notes in
Math. 198 (1980), 192-202.
H. Hopf, Uber die Abbildungen der dreidimensionalen Sphare auf die Kugelache,
Math. Ann. 104 (1931), 637-665.
E. Lima; Commuting vector els on S 3 ; Ann. of Math., vol. 81, 1965, p. 70-81.
David W. Lyons; An Elementary Introduction to the Hopf Fibration, Mathematics
Magazine Vol. 76, No. 2 (Apr., 2003), pp. 87-98.
M. McQuillan; Diophantine approximation and foliations, Publ. Math. I.H.E.S.
87 (1998), pp. 121-174.
M. McQuillan; Non-commutative Mori Theory, Prepublication
I.H.E.S./M/01/42, August 2001.
J. Milnor: Morse Theory; Annals of Mathematics Studies, Princeton University
Press, 1963.
J. Milnor; Growth of nitely generated solvable groups. J. Dierential Geometry,
2, 1968, 447449.
R. Moussu, F. Pelletier; Sur le theoreme of Poincare-Bendixson. Ann. Inst.
Fourier (Grenoble) 24 (1974), in the. 1, 131148.
P. Painleve; Lecons sur la theorie analytique des equations dierentielles, Paris,
Bibliography 175
Librairie Scientique A. Hermann, 1897.

J. Plante; Anosov Flows, Am. Journal of Mathematics, 94 (1972), pp. 729-754.
J. Plante; A Poincare-Bendixson Theorem, Topology 12, pp. 177-181, 1973.
J. Plante; On the existence of exceptional minimal sets in foliations of codimen-
sion one. J. Dierential Equations 15 (1974), 178194.
J. Plante; Measure preserving pseudogroups and a theorem of Sacksteder. Ann.
Inst. Fourier (Grenoble) 25 (1975), in the. 1, vii, 237249.
J. Plante; Foliations with measure preserving holonomy; Ann. of Math., 102,
1975, pp. 327-361.
J. Plante, W. Thurston; Polynomial growth in holonomy groups of foliations.
Comment. Math. Helv. 51 (1976), in the. 4, 567584.
G. Reeb; Sur les varietes integrales des champs delements of contact
completement integrables. C. R. Acad. Sci. Paris 220, (1945). 236237.
G. Reeb; Varietes feuilletees, feuilles voisines. C. R. Acad. Sci. Paris 224, (1947).
16131614.
G. Reeb; Remarque sur les varietes feuilletees contenant une feuille compacte to
the groupe of Poincare ni. C. R. Acad. Sci. Paris 226, (1948). 13371339
G. Reeb; Sur certaines proprietes topologiques des varietes feuilletees; Actualites
Sci. Ind., Hermann, Paris, 1952.
H. Rosenberg, R. Roussarie, D. Weil; A classication of closed orientable 3-
manifolds of rank two, Ann. of Math. 91, no. 2, (1970), 449-464.
D. Ruelle; Statistical Mechanics, New York, Benjamin 1969.
D. Ruelle, D. Sullivan; Currents, Flows and Dieomorphisms, Topology, vol. 14,
nthe 4, 1975.
H. Rummler; Comment. Math. Helv., 54, 1979.
P. Molino; Riemannian Foliations; Progress in Mathematics, Birkhauser, Boston
1988.
G. Reeb, P. Schweitzer; Un theoreme de Thurston etabli au moyen de lanalyse
non standard. (French) Dierential topology, foliations and Gelfand-Fuks
cohomology (Proc. Sympos., Pontifcia Univ. Catolica, Rio de Janeiro,
1976), p. 138. Lecture Notes in Math., Vol. 652, Springer, Berlin, 1978.
Rotman, J., An introduction to algebraic topology. Graduate Texts in Mathemat-
ics, 119. Springer-Verlag, New York, 1988.
R. Sacksteder; Foliations and pseudogroups. Amer. J. Math. 87, 1965, 79102.
R. Sacksteder; On the existence of exceptional leaves in foliations of codimension
one; Ann. Inst. Fourier, Grenoble, 14, 2, 1969, pp. 224-225.
B. Scardua, Transversely ane and transversely projective foliations, Ann. Sc.
Ecole Norm. Sup., 4e serie, t.30, 1997, p.169-204
H. Seifert, Closed integral curves in 3-space and isotopic two-dimensional defor-
mations, Proc. Amer. Math. Soc. 1, (1950). 287302.
L. Schwartz; Theorie des distributions, Nouvelle Edition, Paris, Hermann, 1966.
S. Schwartzmann; Asymptotic cycles, Ann. Math., 66, pp. 270-284, (1957)
P. Scott; The geometries of 3-manifolds, Bull. London Math. Soc. 15 (1983),
401-487.
W. Schachermayer; Addendum: Une modication standard de la demonstration
non standard de Reeb et Schweitzer Un theoreme de Thurston etabli au
moyen de lanalyse non standard, Dierential topology, foliations and

Gelfand-Fuks cohomology (Proc. Sympos., Pontifcia Univ. Catolica, Rio
de Janeiro, 1976), pp. 139140. Lecture Notes in Math., Vol. 652, Springer,
Berlin, 1978.
P. A. Schweitzer, Counterexamples to the Seifert conjecture and opening closed
leaves of foliations, Ann. of Math. (2) 100 (1974), 386400.
Akira Seitoh; Remarks on Stability for Semiproper Exceptional Leaves, Tokyo J.
Math. VOL. 6, No. 1, 1983.
Y. Siu, Techniques of Extension of Analytic Object, Lecture Notes in Pure and
Appl. Math. 8, Marcel Dekker, Inc., New York, 1974.
D. Sullivan; Cycles for the Dynamical study of Foliated Manifolds and Complex
Manifolds, Inventiones Math., 36, pp. 225-255, (1976).
W. P. Thurston, Anosov Flows and the Fundamental Group, Topology, Vol. 11,
pp.147-150. Pergamon Press, 1972.
W. P. Thurston; A generalization of the Reeb stability theorem, Topology 31, pp.
347-352, 1974.
W. Thurston; A local construction of foliations for three-manifolds, Proc. Sym-
posia in Pure Math. 27 (1975), A.M.S., 315-319.
W. Thurston; Existence of Codimension-One Foliations, Annals of Mathematics
Second Series, Vol. 104, No. 2 (Sep., 1976), pp. 249-268.
D. Tischler; On bering certain foliated manifolds over S 1 , Topology, vol. 9 (1970),
pp. 153-154.
J. Tits; Free subgroups in linear groups, J. of Algebra, 20, 1972, pp. 232-270.
J. Wolf; Growth of nitely generated solvable groups and curvature of Riemannian
manifolds, Journ. Di. Geom., 2, 1968, pp. 421-446.
Index
attracting leaf, 162 structure theorem of, 163

Dirac, 152
biregular cover, 164 distinguished map, 3
border double of a foliation, 6
of an invariant open subset, 164
bration, 8
Cantwell-Conlon, 155, 161, 169 ow, 16
theory, 168 periodic, 16
theory of, 155 foliated product, 162, 164
completion foliation, 13
abstract transverse completion, 164 suspension of a, 25
of an invariant set, 165 equivalence, 4
complex of cochains, 138 foliation current, 153
cone structure, 151 foliation cycle, 153
critical point, 78 linear, 6
non-degenerated, 78 Reebless, 13
with respect to a foliation, 80 suspension of a, 24
current, 136, 153 transverse to a bre bundle, 23
Dirac current, 152 transversely
curve ane, 4
in general position with respect to orientable, 4
a foliation, 92 turbulization, 18
normal to a foliation, 91
graph, 83
De Rham group
complex, 138 amenable, 145
Dippolito, 155, 161, 163 Cohomology group, 138
decomposition, 170 De Rham Cohomology group, 138
semi-stability theorem, 168 representation, 20
semi-stability theorem of, 163, 165 growth
structure theorem, 166 combinatorial growth type, 123
177
exponential growth, 129 holonomy invariant measure, 125

function, 121 holonomy invariant transverse
growth function of a group, 124 measure, 142
growth of a group, 124 invariant by a pseudo-group, 125
growth type of an orbit, 123 measure preserving holonomy, 125
polynomial growth, 129 support of a, 142
subexponential growth, 129 support of a measure, 126
type, 122 Milnor, 103
minimal set, 55
Haeiger, 78 exceptional, 57, 58
holomorphic ow, 128 Morse function, 79
holonomy with respect to a foliation, 80
global, 24, 25
global holonomy, 144 nerve, 131
group, 24, 59, 60 Novikov, 87
Holonomy lemma, 91
map, 59 orbit
of the Reeb foliation, 61 of a point in a pseudo-group, 123
one-sided, 77 periodic, 16
trivial, 60
without, 60 Plante
Hopf, 103 family of, 159
Hopf bration, 17 plaque, 1
holonomy along the, 68
isotropy subgroup, 14 projection, 67
space of, 67
leaf, 2 plaques, 129
attracting, 162 proper
border leaf, 164 side of a leaf, 161
inferior structure of, 169 pseudo-group
ressort leaf, 155 of local homeomorphisms, 123
semi-proper leaf, 161 pseudogroup
semi-stable, 162 holonomy pseudogroup, 125, 140,
superior structure of, 169 141
Lie group, 13 of local homeomorphisms, 141
action, 14
Lima, 103 rank
limit cycle, 77, 83 of a manifold, 103
rank of manifold, 18
map Reeb
in general position with respect to component, 12
a foliation, 80 foliation, 6, 911
normal extension, 91
Maurer-Cartan form, 27 Seifert
Maurer-Cartan formula, 27 conjecture, 12
measure manifold, 16
Index 179
semi-proper 163
leaf, 161 Haeigers theorem, 78
semi-stable Limas theorem, 103
leaf, 162 Novikovs compact leaf, 87
set of Cantwell-Colon, 169
minimal, 55 of De Rham, 139
side of Plante, 129
positive, 161
of Sacksteder, 107
singularity
structure theorem of Dippolito, 163
center, 82
Tischlers theorem, 107
saddle, 82
Tischler, 32, 107
stable
compact leaf, 67, 69 Tits, 125
leaf, 67 transverse
subset, 66 map, 76
submersion, 4
suspension, 145 vanishing cycle, 87
simple, 156
theorem vector eld, 12, 1618, 32
Dippolito semi-stability theorem, normal to a foliation, 90

Geometry Dynamics and Topology of FoliationsA First Course

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Geometry Dynamics and Topology of FoliationsA First Course

Transféré par

Droits d'auteur :

Formats disponibles

GEOMETRY,

This page intentionally left blank

Carlos Arnoldo Morales Rojas

Library of Congress Cataloging-in-Publication Data

British Library Cataloguing-in-Publication Data

Copyright 2017 by World Scientific Publishing Co. Pte. Ltd.

This page intentionally left blank

The Geometrical Theory of Foliations is one of the elds in Mathematics

introduction to the general theory of foliations. There, one can nd a very

In the last part of the book we give an exposition of some important

This page intentionally left blank

2. Plane elds and foliations 39

3. Topology of the leaves 53

3.1 Space of leaves . . . . . . . . . . . . . . . . . . . . . . . . 53

4. Holonomy and stability 59

6. Novikovs compact leaf 87

7. Rank of 3-manifolds 103

8. Tischlers theorem 107

9. Plantes compact leaf theorem 121

10. Currents, distributions, foliation cycles and transverse measures 135

11. Foliation cycles: A homological proof of Novikovs com-

Appendix A Structure of codimension one foliations: Dip-

This page intentionally left blank

1.1 Denition of foliation

There are essentially three ways to dene foliations. Let M be a m-

Denition 1.1. A C r foliation of codimension m n of M will be a max-

(1) Xi (Ui ) = Dn Dmn ;

Xj (Xi )1 (x, y) = (fi,j (x, y), gi,j (y)).

The number n is called the dimension of F . A plaque of F is a set

Dmn Dn . Thus, we have the following equivalent denition of foliation.

Denition 1.2. A C r foliation of codimension m n of M is a partition

Warning: Not every decomposition of M into immersed submanifolds

The third denition of foliation uses the notion of distinguished appli-

Denition 1.3. A C r foliation of codimension m n of M is a cover {Ui :

This last denition leads to several interesting denitions. For instance,

The above relation denes an equivalence in the space of foliations.

Fig. 1.2 Non equivalent foliations.

1.2 Examples of foliations

1.2.1 Foliations derived from submersions

Theorem 1.1. Let f : M m N n be a C r submersion. Then the connected

Fig. 1.3 Foliation and submersion.

We claim that the collection F = {(U, X)} denes a foliation of M . In

Hence 2 (X X 1 ) = (Y Y 1 ) 2 does not depend on x Dn . This

Let us present some examples to illustrate the above result.

Example 1.1. Let M and N be C r manifolds and take f : M N M as

Exercise 1.2.1. Let F be a foliation on M of codimension q. A dierentiable

Exercise 1.2.2 (Double of a foliation). Let F be a smooth foliation on

Example 1.2. Let M = R2 and f (x, y) = y x, where R. The

Example 1.3 (Reeb foliation on S 3 ). Let M = R3 and f (x, y, z) =

Fig. 1.4 Linear foliation on T 2 .

Fig. 1.5 Graph of .

The map f is a submersion since

codimension 1 of M . The leaves of this foliation, (i.e., the level curves of

f (x, y, z) = c (r2 )ez = c.

If c = 0 then (r2 ) = 0 x2 + y 2 = 1. Hence the level curve corresponding

(r2 )ez = c (r2 ) > 0.

(K = ln(c)). When c = 1 we have

The graph of the above curve in the plane y = 0 is given by

Example 1.4 (Fibrations). Let E, B, F be smooth manifolds. We say

(1) 1 (b) is dieomorphic to F , b B.

Clearly the family { 1 (b) : b B} is a C codimension dim(B) folia-

1.2.2 Reeb foliations

Reeb foliation in the cylinder Reeb foliation in the Moebious band

Fig. 1.7 Reeb foliation in the cylinder and Moebius band.

Example 1.11. Suppose B = S 1 . In this case B = R and 1 (B) = Z.

(c) if Ui Uj = then yi = fij yj for some fij G.