Hossein Movasati

Introduction to Algebraic Curves

and Foliations*
May 31, 2016

Publisher
Preface

The guiding principal in this book is to give a detailed and historical exposition of
the theory of holomorphic foliations in the projective space of dimension two. Our
emphasis is the algebraic aspects of such a theory and so, we would like to rise the
need for working with arbitrary fields instead of the field of complex numbers. This
makes our text different from the available texts in the literature such as Camacho-
Sad’s monograph [CS87] which emphasizes local aspects, Brunella’s monograph
[Bru00] which emphasizes the classification of holomorphic foliations similar to
classification of two dimensional surfaces, Lins Neto-Scárdua’s book [LNS] and
Ilyashenko-Yakovenko’s book [IY08] which both emphasize analytic and holomor-
phic aspects. We have in mind an audience with a basic knowledge of Complex
Analysis in one variable and Algebraic Geometry of curves in the two dimensional
projective space. The text is mainly written for two primary target audiences: under-
graduate students who want to have a flavor of an important class of holomorphic
foliations and algebraic geometers who want to learn how the theory of holomorphic
foliations can be written in the framework of Algebraic Geometry.

Hossein Movasati
January 2016
Rio de Janeiro, RJ, Brazil

v
Acknowledgements

The present text is written during the academic years between 2015-2017. First of
all I would like to thank my colleagues at IMPA, César Camacho, Alcides Lins
Neto, Paulo Sad and Jorge V. Pereira for many useful conversations regarding the
material of the present text. I would also like to thank the students who participated
my courses on holomorphic foliations and helped me to improve the style of the ex-
position of the present text. This includes Christian Peterson Bórquez and Yadollah
Zare.

vii
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Hilbert’s sixteen problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 Real foliations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Poincaré first return map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Hilbert 16-th problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.1 Algebraic curves invarant by foliations . . . . . . . . . . . . . . . . . . 7

3 Darboux’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1 Some algebraic notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Invariant algebraic sets and first integrals . . . . . . . . . . . . . . . . . . . . . . . 11

4 Holomorphic foliations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1 Complexification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Imagining curves and leaves in a correct way! . . . . . . . . . . . . . . . . . . . 15

5 Holonomy I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.1 Integrating Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.2 Transversal section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.3 Holonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.4 A formula for integrating Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.5 Formulas for integrating factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

6 Singularities of holomorphic foliations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

6.1 Singularities of multiplicity one . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.2 Poincaré theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3 Siegel domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.4 Separatrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.5 Singularities with resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

ix
x Contents

7 Projective spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.1 Projective spaces as complex manifolds . . . . . . . . . . . . . . . . . . . . . . . . 33
7.2 Projective spaces as schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.3 Foliations in projective spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.4 Ricatti foliations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.5 Minimal set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

8 Camacho-Sad theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
8.1 Camacho-Sad index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
8.2 Residue formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
8.3 Camacho-Sad theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

9 Baum-Bott index and Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

9.1 Baum-Bott index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
9.2 Applications of Baum-Bott and Camacho-Sad index theorems . . . . . 48

10 Blow up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
10.1 Blow-up of a point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
10.2 Blow-up and foliations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
10.3 Multiplicity along an invariant curve . . . . . . . . . . . . . . . . . . . . . . . . . . 55
10.4 Sequences of blow-ups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
10.5 Milnor number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
10.6 Seidenberg’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

11 Jouanolou foliation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

12 Fibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
12.1 Main examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
12.2 Generic conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
12.3 Ehresmann’s fibration theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
12.4 Monodromy and vanishing cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
12.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

13 Generic conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

14 Iterated Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Frequently used notations

(X, x) For a topological space X and x ∈ X we denote by (X, x) an small

neighborhood of x in X.
OC2 ,p The ring of holomorphic functions in a neighborhood of p in C2 .
k, k̄ A field of characteristic zero and its algebraic closure.
k[x, y] The ring of polynomials in x, y with coefficients in k.
k(x, y) The quotient field of k[x, y] which is the same as the field of ratio-
nal functions in x, y with coefficients in k.
Mt The transpose of a matrix M. We also write M = [Mi j ], where Mi j
is the (i, j) entry of M. The indices i and j always count the rows
and columns, respectively.
V∨ The dual of an R-module V , where R is usually the ring Z or the
field k. We always write a basis of a free R-module of rank r as a
r × 1 matrix. For a basis δ of V and α of V ∨ we denote by

[δ , α t ] := [α j (δi )]i, j

the corresponding r × r matrix.

d The differential operator or a natural number, being clear in the
text which one we mean.
X(k) or Xk The set of k-rational points of X defined over the field k. In par-
ticular for k ⊂ C, X(C) is the underlying complex manifold of X.
Sometimes, for simplicity we write X = X(C), being clear in the
context that X is a complex manifold.
FR , FC The foliation in R2 and C2 , respectively.

xi
Chapter 1
Introduction

The present text has been arisen from the lecture notes of the author during the
academic years 2015-2016 at IMPA. Its objective is to introduce the reader with a
basic knowledge in holomorphic foliations. Our approach is purely algebraic and
we avoid many transcendental arguments in the literature. For our purpose we take
foliations in the affine variety A2k with k = R or C and given by polynomial vector
fields. The most famous problem for such foliations is the centennial Hilbert 16-th
problem, H16 for short. Our aim is not to collect all the developments and theo-
rems in direction of H16 (for this see for instance [Ily02]), but to present a way of
breaking the problem in many pieces and observing the fact that even such partial
problems are extremely difficult to treat. Our point of view is algebraic and we want
to point out that the both real and complex Algebraic Geometry would be indispens-
able for a systematic approach to the H16. Here is Hilbert’s announcement of the
problem:
16. Problem of the topology of algebraic curves and surfaces
The maximum number of closed and separate branches which a plane algebraic
curve of the n-th order can have has been determined by Harnack. There arises the
further question as to the relative position of the branches in the plane. As to curves
of the 6-th order, I have satisfied myself-by a complicated process, it is true-that
of the eleven branches which they can have according to Harnack, by no means
all can lie external to one another, but that one branch must exist in whose interior
one branch and in whose exterior nine branches lie, or inversely. A thorough inves-
tigation of the relative position of the separate branches when their number is the
maximum seems to me to be of very great interest, and not less so the corresponding
investigation as to the number, form, and position of the sheets of an algebraic sur-
face in space. Till now, indeed, it is not even known what is the maximum number of
sheets which a surface of the 4-th order in three dimensional space can really have.
In connection with this purely algebraic problem, I wish to bring forward a ques-
tion which, it seems to me, may be attacked by the same method of continuous
variation of coefficients, and whose answer is of corresponding value for the topol-
ogy of families of curves defined by differential equations. This is the question as

1
2 1 Introduction

to the maximum number and position of Poincar’s boundary cycles (cycles limites)
for a differential equation of the first order and degree of the form

dy/dx = Y /X,

where X and Y are rational integral functions of the n-th degree in x and y. Written
homogeneously, this is

X(ydz/dt − zdy/dt) +Y (zdx/dt − xdz/dt) + Z(xdy/dt − ydx/dt) = 0

where X, Y, and Z are rational integral homogeneous functions of the n-th degree in
x, y, z, and the latter are to be determined as functions of the parameter t.
Chapter 2
Hilbert’s sixteen problem

In this chapter we introduce limit cycles of polynomial differential equations in R2

and state the well-known Hilbert 16-th problem. Despite the fact that this is not the
main problem of the present text, it must be considered the most important unsolved
problem related to the topic of the present text.

2.1 Real foliations

What we want to study is the following ordinary differential equation:


ẋ = P(x, y)
, (2.1)
ẏ = Q(x, y)

where P, Q are two polynomials in x and y with coefficients in R and ẋ = dx

dt . We will
assume that P and Q do not have common factors. Its solutions are the trajectories
of the vector field:
∂ ∂
X := P(x, y) + Q(x, y)
∂x ∂y
∂ ∂
(we will also write X = (P, Q)). For now, the reader may ∂x and ∂y notations defined
by
∂ ∂
:= (1, 0), := (0, 1)
∂x ∂y
This is introduced in order to distinguish between points and vectors in R2 .
Let us first recall the first basic theorem of ordinary differential equations.
Theorem 1 For A ∈ R2 if X(A) 6= 0 then there is a unique analytic function

γ : (R, 0) → R2

such that

3
4 2 Hilbert’s sixteen problem

γ(0) = A, γ̇ = X(γ(t))

Proof. Let us write formally

∞
γ = ∑ γit i , γi ∈ R2 , γ0 := A
i=0

and substitute it in γ̇ = X(γ). It turns out that γi can be written in a unique way in
terms of of γ j , j < i. This guaranties the existence of a unique formal γ. Note that if
X(A) = 0 then γi = 0 for all i ≥ 1 and so γ is the conatant map γ(t) = A.

Exercise 1 Recover the proof of convergence of γ from classical books on ordinary

differential equations.

In Theorem 1 we have even claim that γ depends on A analytically, that is, there a
small neighborhood (R2 , A) of A in R2 and an analytic function

Γ : (R2 , A) × (R, 0) → (R2 , A)

such that Γ (B, ·) for all B ∈ (R2 , A) is the solution in Theorem 1 crossing the point
B. In this way we may reformulate the following theorem:
Theorem 2 For A ∈ R2 if X(A) 6= 0 then there is a analytic isomorphism F :
(R2 , 0) → (R2 , A) such that the push-forward of ∂∂x by F is X.
∂
Proof. The push forward of the vector field ∂x by F is X. This is equivalent to

∂ F1 ∂ F1
!   
∂x ∂y 1 P(F1 , F2 )
∂ F2 ∂ F2 =
0 Q(F1 , F2 )
∂x ∂y

where F = (F1 , F2 ). By a rotation around A, we may assume that P(A) 6= 0. In a sim-

ilar way as in Theorem 1 we have a unique solution (F1 , F2 ) to the above differential
equation with
(F1 (0, y), F2 (0, y)) = A + (0, y).
We have !
∂ F1 ∂ F1
∂ x (0, 0) ∂ y (0, 0)
 
P(A) 0
∂ F2 ∂ F2 =
Q(A) 1
∂ x (0, 0) ∂ y (0, 0)

and so F = (F1 , F2 ) : (R2 , 0) → (R2 , A) is an analytic isomorphism.

Exercise 2 Describe the trajectories of the following differential equations:

  
ẋ = y ẋ = x ẋ = x
, ,
ẏ = −x ẏ = −y ẏ = y

Example 1 The trajectories of the differential equation

2.1 Real foliations 5
2.4

1.6

0.8

-2.8 -2.4 -2 -1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6 2 2.4 2.8

-0.8

-1.6

-2.4

Fig. 2.1 A limit cycle crossing (x, y) ∼ (−1.79, 0)

 2
ẋ = 2y + x2
(2.2)
ẏ = 3x2 − 3 + 0.9y

are depicted in Figure (2.1).

The collection of the images of the solutions of (2.1) gives us us an ana-
lytic singular foliation F = F (X)R = F (X) = FR in R2 . Therefore, when we
are talking about a foliation we are not interested in the parametrization of its
leaves(trajectories). It is left to the reader to verify that:
Exercise 3 For a polynomial R ∈ R[x, y] the foliation associated to X and RX in
R2 \{R = 0} are the same.
For this reason from the beginning we have assumed that P and Q have no common
factors. Being interested only on the foliation F (X), we may write (2.1) in the form

dy P(x, y)
= ,
dx Q(x, y)

ω = 0, where ω = Pdy − Qdx ∈ ΩR1 2 .

In the second case we use the notation F = F (ω)R = F (ω). In this case the folia-
tion F is characterized by the fact that ω restricted to the leaves of F is identically
zero. A systematic definition of differential 1-forms will be done in §3.
Definition 1 The singular set of the foliation F (Pdy − Qdx) is defined in the fol-
lowing way:

Sing(F ) = Sing(F )R := {(x, y) ∈ R2 | P(x, y) = Q(x, y) = 0}.

By our assumption Sing(F ) is a finite set of points. The leaves of F near a point
A ∈ Sing(F ) may be complicated.
Exercise 4 Using a software which draws the trajectories of vector fields, describe
the solutions of (2.2) near its singularities.
By Bezout theorem we have
6 2 Hilbert’s sixteen problem

#Sing(F ) ≤ deg(P) deg(Q)

The upper bound can be reached, for instance by the differential equation F (Pdy −
Qdx), where

P = (x − 1)(x − 2) · · · (x − d), Q = (y − 1)(y − 2) · · · (y − d 0 ).

2.2 Poincaré first return map

From topological point of view a leaf L of F = F (ω) is either homeomorphic to

R or to the circle S1 := {(x, y) ∈ R2 | x2 + y2 = 1}. In the second case L is called a
closed solution of F (but not yet a limit cycle).
Exercise 5 For a foliation F = F (ω)R the curve {R = 0}, where dω = Rdx ∧ dy,
intersects all closed leaves of F .
We consider a point p ∈ L and a transversal section Σ to F at p. For any point q
in Σ near enough to p, we can follow the leaf of F in the anti-clockwise direction
and since L is closed we will encounter a new point h(q) ∈ Σ . We have obtained an
analytic function
h : Σ → Σ,
which is called the Poincaré first return map. Later in the context of holomorphic
foliations we will call it the holonomy map. Usually we take a coordinate system z
in Σ with z(p) = 0 and write the power series of h at 0:
∞
h(n) (0) n
h(z) = ∑ z
i=0 n!

Definition 2 h0 (0) is called the multiplier of the closed solution L. If the multiplier
is 1 then we say that h is tangent to the identity. In this case the tangency order is n
if
h(i) (0) = 0, h(n) (0) 6= 0.
A closed solution L of F is called a limit cycle if its Poincaré first return map is
not identity. In case the Poincaré first return map is identity then the leaves of F
near L are also closed. In this case we can talk about the continuous family of cycles
δz , z ∈ Σ , where δz is the leaf of F through z.
Exercise 6 Prove that the multiplier and order of tangency do not depend on the
coordinate system z in Σ .
Proposition 1 In the above situation, we have
dω
Z
h0 (0) = exp(− ).
δ ω
2.3 Hilbert 16-th problem 7

Proof. The proof will be presented in Chapter 5

2.3 Hilbert 16-th problem

It is natural to ask whether a foliation F (Pdy − Qdx) has a finite number of limit
cycles. This is in fact the first part of Hilbert 16-th problem:
Theorem 3 Each polynomial foliation F (Pdy − Qdx) has a finite number of limit
cycles.
The above theorem was proved by Yu. Ilyashenko and J. Ecalle independently
around 80’s. We have associated to each foliations F the number N(F ) of its limit
cycles of F . It is natural to ask how N(F ) depends on the ingredient polynomial P
and Q of F .
2.1. (Hilbert 16’th problem) Fix a natural number n ∈ N. Is there some natural num-
ber N(n) ∈ N such that each foliation F (Pdx − Qdy) with deg(P), deg(Q) ≤ n has
at most N(n) limit cycles.
Of course, it would be of interest to give an explicit description of N(n) and more
strongly determine the nature of

N(n) := max{N(F (ω)) | ω = Pdy − Qdx, deg(P), deg(Q) ≤ n}.

One of the objective of the present text is to explain the fact that Hilbert 16’th
problem is a combination of many unsolved difficult problems. We note that even
the case n = 2 is open.

2.3.1 Algebraic curves invarant by foliations

Let f ∈ R[x, y]. An algebraic curve over R is defined to be

{ f = 0} := {(x, y) ∈ R2 | f (x, y) = 0}.

It can happen that such an algebraic curve is empty, for instance take f = x2 +y2 +1,
or it is a point, for instance take f = x2 + y2 . For a moment assume that { f = 0} at
at a point is really look like a smooth curve, for instance take f = x2 + y2 − 1 for
which the curve is a circle of radius 1.
Let F (X), X = P ∂∂x + Q ∂∂y be a foliation in R2 as before. We would like to
see when the smooth part of f is a part of trajectories of X (leaves of F (X)). The
gradient vector
∂f ∂ ∂f ∂
+
∂x ∂x ∂x ∂y
is perpendicular to the curve and so if we have
8 2 Hilbert’s sixteen problem

∂f ∂f
P+ Q = f · R. (2.3)
∂x ∂y

then { f = 0} in neighborhood of p is a part of a leaf of F . If the equality (2.3)

occurs then we say that the algebraic curve { f = 0} is F -invariant.
Chapter 3
Darboux’s theorem

In this chapter we will state and prove a theorem due to Darboux. It syas that if an
algebraic foliation in A2k has infinite number of algebraic leaves then it must have a
first integral. Here, k is an algebraically closed field of characteristic zero.

3.1 Some algebraic notations

The set of polynomial differential 1-forms

ΩA1 2 := {Pdy − Qdx | P, Q ∈ k[x, y]},

k

and differential two forms

ΩA2 2 = {Pdx ∧ dy | P ∈ k[x, y]}.

k

One usually defines:

ΩA0 2 := k[x, y].
k

The wedge product is defined in the following way:

(P1 dx + Q1 dy) ∧ (P2 dx + Q2 dy) = (P1 Q2 − P2 Q1 )dx ∧ dy.

Exercise 7 Verify that for all ω1 , ω2 ∈ ΩA1 2 we have ω1 ∧ ω1 = 0 and ω1 ∧ ω2 =

k
−ω2 ∧ ω1 .
We have the differential maps:

∂P ∂P
d0 : ΩA0 2 → ΩA1 2 , d0 (P) = dx + dy.
k k ∂x ∂y

9
10 3 Darboux’s theorem

d1 : ΩA1 2 → ΩA2 2 , d1 (Pdx + Qdy) = dP ∧ dx + dQ ∧ dy.

k k

Exercise 8 Show that d1 ◦ d0 = 0.

We will usually drop the sub index 0 and 1 and simply write d = d0 , d = d1 .
Exercise 9 If dω = 0 for some ω ∈ ΩA1 2 then there is a f ∈ ΩA0 2 such that ω = d f .
k k
Is this true for char(k) 6= 0? As a hint take ω = −2x p−1 ydx + x p−2 dy, where p :=
char(k). Can you classify all ω’s which do not satisfy the mentioned property.
An easier statement is that if d f = 0 for f ∈ ΩA0 2 then f is a constant, that is,
k
f ∈ k. This is false if the characteristic of k is not zero. For instance, in a field of
characteristic p we have dx p = px p−1 dx = 0 but x p is not a constant.
Stokes formula. Let δ be a closed anti-clockwise oriented path in R2 which does
not intersect itself. Let also ∆ be the region in R2 which δ encloses. Then
Z Z
ω= dω.
δ ∆

Exercise 10 Give a proof of Stokes formula using the classical books in calculus.
Let k = R or C. Let also γ = (x(t), y(t)) : (k, 0) → k2 be an analytic map and
ω = Pdx + Qdy ∈ ΩA1 2 . The pull-back of ω by γ is defined to be
k

∂ x(t) ∂ y(t)
γ ∗ ω := (P(x(t), y(t)) + Q(x(t), y(t)) )dt
∂t ∂t

Exercise 11 Show that γ ∗ ω = 0 is independent of the parametrization t, i.e if a :

(k, 0) → (k, 0) is an analytic map and γ ∗ ω = 0 then (γ ◦ a)∗ ω = 0.
If γ ∗ ω = 0 then we say that ω restricted to the image of γ is zero. We denote by
P
k(x, y) := { | P, Q ∈ k[x, y]}
Q

the field of rational (meromorphic) functions in A2k . The set of meromorphic differ-
ential i-forms is denoted by ΩAi 2 (∗) (instead of k[x, y] we have used k(x, y)).
k

Exercise 12 Show that if for ω1 , ω2 ∈ ΩA1 2 (∗) we have ω1 ∧ ω2 = 0 then ω2 = Rω1

k
for some R ∈ k(x, y). Show that if for ω1 = Pdy − Qdx, ω2 ∈ ΩA1 2 we have ω1 ∧ ω2
k
and P and Q are relatively prime then ω2 = Rω1 for some R ∈ k[x, y]. Is this exercise
true for char(k) 6= 0.
For Ω ∈ ΩA2 2 and ω ∈ ΩA1 2 we denote by Ω
ω any meromorphic differential 1-form
k k
α such that
3.2 Invariant algebraic sets and first integrals 11

Ω = ω ∧ α.

Exercise 13 Show that such an α exists and is defined up to addition by an element

in K(x, y)ω.

3.2 Invariant algebraic sets and first integrals

In this section assume that f ∈ k[x, y] is irreducible.

Definition 3 We say that a curve { f = 0} is F (ω)-invariant if

ω ∧ d f = f η, for some η ∈ ΩA1 2 . (3.1)

k

The geometric description of the equality (3.1) is as follows. Let us write ω = Pdy−
Qdy and X = P ∂∂x + Q ∂∂y as usual. We know that

∂f ∂f
ω ∧ d f = (Pdy − Qdx) ∧ ( dx + dy) = (X · ∇ f )dx ∧ dy = f Rdx ∧ dy (3.2)
∂x ∂y

where η = Rdx ∧ dy. Note that at the points where f = 0 we have that X · ∇ f = 0,
but since ∇ f is perpendicular to the level curve of f we have that X is tangent to
{ f = 0}.
Definition 4 We say that f ∈ k(x, y) is a (rational) first integral of the foliation
F (ω) if
ω ∧ d f = 0. (3.3)
In other words, there is g ∈ k(x, y) such that

ω = gd f .

If this is the case we say that F (ω) has a first integral.

F
Let us assume that f = G where F, G ∈ k[x, y]. We have d f = G.dF−F.dG
G2
and so
ω ∧ (G.dF − F.dG) = 0 and so F (G.dF − F.dG) has the first integral G .
F

Proposition 2 Let us assume that the foliation F (ω) has the first integral F
G as
above. The algebraic curves F − cG = 0, c ∈ k are F (ω)-invariant .

Proof. We have to show that ω ∧ d(F − cG) is divisible by F − cG.

Theorem 4 (Darboux) If the foliation F has infinite number of invariant algebraic

curves then F has a rational first integral.
Recall that by definition two algebraic curves { f1 = 0}, { f2 = 0} are the same if
f1 = c · f2 for some c ∈ k.
12 3 Darboux’s theorem

Proof. The proof is classical and can be found in [LNS] page 92. Let us assume
that F (ω) has infinite number of invariant algebraic curves { fi = 0}, i ∈ N. By
definition ω ∧ d fi = fi .ηi , ηi ∈ ΩA2 2 . We rewrite this
k

d fi
ω∧ = pi dx ∧ dy where pi ∈ k[x, y]
fi

We make the observation that deg{pi } is independent of the degree of fi . To see this
fact we write

∂ fi ∂ fi
(Pdx + Qdy) ∧ ( dx + dy) = fi .pi dx ∧ dy
∂x ∂y
then
∂ fi ∂ fi
P −Q = fi .pi
∂x ∂y
Let d := Max{degP, degQ}. Then

∂ fi ∂ fi
degpi + deg fi = deg( fi .pi ) = deg(P −Q ) 6 d + deg fi − 1
∂x ∂y

and so degpi 6 d − 1. The vector space k[x, y]≤n = { f ∈ k[x, y]|deg f ≤ n} is finite
dimensional and in fact  
n+2
dim k[x, y]≤n =
2
We set n = d − 1 and define an to be the dimension of the k-vector space generated
by pi ’s. We have  
n+2
an ≤
2
We choose a basis p1 , p2 , ..., pan for such a vector space. The element pan +1 is
linearly dependent with the element of such a basis, that is , there are ri ∈ k, i =
1, ..., an + 1 such that
an +1
∑ ri .pi = 0
i=1

and ran +1 6= 0. In other words

an +1
d fi an +1 d fi
ω∧ ∑ ri . = ∑ ri (ω ∧ )=0
i=1 fi i=1 fi

an +1
Let α = ∑ ri dffi i and so dα = 0 because
i=1

an +1
α= ∑ ri d(ln fi )
i=1
3.2 Invariant algebraic sets and first integrals 13

We repeat the same argument for p1 , p2 , ..., pan , pan +2

an +2
∑ r̃i .pi = 0
i=1,i6=an +1

for some r̃i ∈ k and so

an +2
d fi
ω ∧( ∑ r̃i )
i=1,i6=an +1 fi
an +2
Let β = ∑ r̃i dffi i and so we have
i=1,i6=an +1

ω ∧ α = ω ∧ β = 0.

For this we conclude that α = f β for some non-constant function f ∈ k(x, y) (see
Exercise 12 ). Since dα = 0 we conclude that d f ∧ β = 0 and so f is a first integral
of F (ω). Note that f is non-constant because in the expression of α and β we have
d f +1 d f +2
respectively the terms faan+1 and faan+2
n n

It is possible to derive refinements of the Darboux’s theorem by analyzing its

proof.
Theorem 5 If the foliation F (ω), ω = Pdy−Qdx, max(deg(P), deg(Q)) = d has
d+1

2 + 2 number of invariant algebraic curves then F has a rational first integral.
We observe that we have a new examples of foliations appearing in the proof of
Darboux’s theorem.
Definition 5 A holomorphic foliation F (ω) has a logarithmic first integral if there
are polynomials f1 , f2 , . . . , fs ∈ k[x, y] and λ1 , λ2 , . . . , λs ∈ k such that
s
d fi
ω ∧ ( ∑ λi )=0
i=1 fi

For k = R or C, the level surfaces of the multi-valued functions f1λ1 f2λ2 · · · fsλs are
tangent to the foliations F (ω). We call this a logarithmic first integral of F (ω).
Theorem
6 If the foliation F (ω), ω = Pdy − Qdx, max(deg(P), deg(Q)) = d
has d+12 + 1 number of invariant algebraic curves then F has a logarithmic first
integral.
Exercise 14 Where exactly in this chapter, we need that k is algebraically closed
and its characteristic is zero? For instance, discuss Darboux’s theorem over a field
of non-zero characteristic.
Chapter 4
Holomorphic foliations

In this chapter we will do two main things. First, we will consider the foliation
F (ω) in C2 instead of R2 . This will be the beginning of the theory of holomorphic
foliations on complex manifolds.

4.1 Complexification

Exercise 15 All the discussions in §2 is valid replacing R with C. In this way, we

replace analytic with holomorphic etc.
In particular,
Theorem 7 For A ∈ C2 if X(A) 6= 0 then there is a biholomorphism F : (C2 , 0) →
(C2 , A) such that the push-forward of ∂∂x by F is X.
The images of the complex solutions of the vector field X give us a holomorphic
foliation F = F (ω)C = FC in C2 . The leaves of FC are two dimensional real
manifolds embedded in a real four dimensional space. If P, Q ∈ R[x, y] we will de-
note by FR = F (ω)R the corresponding real foliation in R2 . Note that R2 ⊂ C2
and
FR = R2 ∩ FC
i.e. the intersection of a leaf of FC with R2 is a union of leaves of FR . Note that
FC may has more singularities.

4.2 Imagining curves and leaves in a correct way!

We consider the curves

C : x2 + y2 = 1, D : xy − 1 = 0.

15
16 4 Holomorphic foliations
2
R

A B C

Fig. 4.1 Correct intuition

The curve C(R) is the circle of radius 1 and D(R) is a hyperbola and they are not
isomorphic topological spaces because the first one has one connected component,
whereas the second one has two. However, over complex numbers these two curves
are the same and the isomorphism is given by

C(C) → D(C), (x, y) 7→ (x + iy, x − iy)

√
where i = −1. The curve D(C) is parameterrized in the polar coordinates by

x = re2πiθ , y = r−1 e−2πiθ , r ∈ R+ , θ ∈ [0, 1]. (4.1)

Since the bijection R+ → R+ , x 7→ x−1 sends 0 to ∞, both curves C(C) and D(C)
are cylinders with two infinities, let us say −∞ and +∞. A cycle δ travels from −∞
to +∞ and it covers the whole cylinder. We would like to make a correct intuition
of this travel. This is fairly easy in the case of C(C). This cycle is in the real four
dimensional space C2 . In a certain time it fully lies in the two dimensional space
R2 ⊂ C2 which is seen as a circle of radius 1 and center 0 ∈ R2 . It disappears from
the two dimensional world and continues its travel toward −∞, see Figure 4.1, A.
The case of D(C) is a little bit tricky as the first reasonable intuition turns out to be
false. First of all we have to identify two connected components of the hyperbola
D(R) inside the cylinder D(C). These are just two lines in D(C) coming from −∞
and going to +∞ without touching each other. Our cycle touches each of these lines
at exactly one point and it seems to make the intuition in Figure 4.1, B. However, a
simple check of the intuition with the parametrization (6.5) gives us the intuition in
Figure 4.1,B, that is, the cycle δ near −∞ is stretched along the y-axis and as it goes
to +∞ it becomes stretched along the x-axis.
Chapter 5
Holonomy I

5.1 Integrating Form

In this section we will work with a foliation F (ω) in the complex manifold M =
(C2 , 0) of dimension two, where ω is a holomorphic 1-form on the manifold M. We
denote by L p the leaf through p ∈ M.
Theorem 8 Assume that 0 is not a singularity of F (ω), that is, ω(0) 6= 0. There
are holomorphic functions f , g ∈ OM such that

ω = g·d f

Further, g(0) 6= 0, f (0) = 0 and f is regular at 0, that is, the derivation of f at zero
is not zero.
Proof. The proof follows from Theorem 2. Latex the proof presented in the class.
Definition 6 In Theorem 8, we call f a first integral of F (ω) and we call g an
integrating factor of F (ω).
Definition 7 In Theorem 8 we some times need to take another f˜ ∈ OC2 ,0 such that
( f˜, f ) evaluated at p is (0, 0) and the determinant of its derivation at p is non-zero.
There is an open subset U of C2 and V of 0 ∈ C2 such that ( f˜, f ) : U → V is a
biholomorphism. We call this a local chart of F around p.
Now, we would like to discuss the issue of different choices of the pair ( f , g).
Proposition 3 In Theorem 8 let us consider two pairs ( fi , gi ), i = 1, 2 such that

ω = g1 d f1 = g2 d f2 .

There is a biholomorphism h : (C, 0) → (C, 0) such that

g1
f2 = h( f1 ), g2 = .
h0 ( f1 )

17
18 5 Holonomy I

z
Lp k(z)
p
δ
Σp
Fig. 5.1 Holonomy

Proof. Write down the details for the proof presented in the class.

5.2 Transversal section

In this section we describe a transversal section to a foliation in a point to a foliation.

It has always a coordinate system given by a first integral.
Definition 8 Let F = F (ω) be a foliation in C2 and let p be a regular point of F .
A transversal section to F at p is

Σ p := {q ∈ (C2 , p) | f˜(q) = 0)}

where f˜ ∈ OC2 ,p together with a first integral f ∈ OC2 ,p form a coordinate system
around p. The transversal section Σ p has always the coordinate system given by the
image of f .
Proposition 4 Let ( f˜, f ) : U → V be a local chart for F as in Definition 7, p, q ∈ U
be two points in the same leaf and Σ p , Σq be two transversal sections to F at p and
q, respectively. There is a unique biholomorphism

h : (Σ p , p) → (Σq , q)

which is characterized by the fact that z ∈ (Σ p , p) and h(z) ∈ (Σq , q) are in the same
leaf of F in U.

The map h is called a local holonomy of F .

Proof. Latex the proof presented in the class.
5.3 Holonomy 19

5.3 Holonomy

Let δ : [0, 1] → L be a path in a leaf L of the foliation F with initial point p and
end point q. Assume that δ has a finite number of self intersecting points and take
two transversal sections Σ p and Σq at p and q, respectively. We cover the image of δ
with local charts for F and since [0, 1] is compactt we can do this by a finite number
of local charts:
Ui , i = 0, 1, 2, 3, . . . , n
Further, we can assume that Ui ∩Ui−1 6= 0./ We also take a transversal section Σi at
some point pi of the path δ in Ui−1 ∩Ui . By convention, we set

Σ0 := Σ p , Σn+1 := Σq , p0 := p, pn+1 := q.

Using Proposition 4 we get biholomorphisims

hi : (Σi , pi ) → (Σi+1 , pi+1 ), i = 0, 1, 2, . . . , n

Definition 9 The holonomy map from Σ p to Σq is defined to be

h := hh ◦ · · · ◦ h1 ◦ h0 : (Σ p , p) → (Σq , q).

The following discussion may help to have a better geometric picture of the notion
of holonomy.
There is a neighborhood Uδ of the path δ such that for every t ∈ [0, 1] and z ∈ Uδ
near δ (t), the lifting path δk(z),z of δ |[0,t] in the leaf Lz is well-defined. Roughly
speaking, the path δk(z),z , in the leaf Lz , connects k(z) ∈ Σ p to z in the direction of
the path δ |[0,t] . In Figure 6.2 we have shown that in the self intersecting points of δ ,
depending on the choice of t, we can choose non-homotop δk(z),z ’s. These paths are
depicted by dash-dot-dot lines. Let Ũδ be the set of all homotopy classes [δk(z),z ] in
an small neighborhood Uδ of δ . The reader can easily verify that Ũδ is a complex
manifold and the natural map τ : Ũδ → Uδ may not be one to one near the self
intersecting points of δ (see Figure 6.2). All functions, for example k(z), that we
define on the set Uδ are multivalued near such points and are one valued in Ũδ . For
simplicity, we will work with Uδ instead of Ũδ .
Let q = δ (t1 ), 0 ≤ t1 ≤ 1, be a point of δ and Σq be a small transverse section at
q to F . For any point z ∈ Σq , the lifting δk(z),z of δ |[0,t1 ] defines the holomorphic
function k : Σq → Σ p . The function

h = k−1 : Σ p → Σq

is called the holonomy of F along δ from Σ p to Σq . If δ is a closed path, q = δ (1)

and Σ p = Σq , we have the holomorphic germ

h : Σp → Σp
20 5 Holonomy I

h is called the holonomy of F along δ in Σ p .

5.4 A formula for integrating Factor

Let F (ω) be a holomorphic foliation in C2 and let p be a regular point of F . There

is a Zariski neighborhood U of p and a regular 1-form α defined in U such that

dω = ω ∧ α

For example, if F is given by a 1-form Pdy − Qdx, then we can define α as follows:
∂P
∂x + ∂∂Qy
α =− dx.
P
This is defined in the Zariski open set P 6= 0. For every two such 1-forms α1 and α2 ,
we have:
dω = ω ∧ α1 = ω ∧ α2 ⇒ ω ∧ (α1 − α2 ) = 0 ⇒
α1 |L = α2 |L for any leaf L of F
Therefore, the αi ’s coincide in the leaves of F . They define a holomorphic 1-form
on each leaf L. By
dω
ω
we mean the collection of all 1-Forms α. The 1-form α := dω
ω can be considered
as a multivalued holomorphic 1-form on C2 \Sing(F ) which is one valued on the
leaves of F and satisfies:
dω = ω ∧ α
By definition, if two 1-forms ω and ω 0 induce the same foliation F , then there is a
rational function f = f (x, y) such that ω 0 = f ω and therefore:

df
dω 0 = d( f ω) = d f ∧ ω + f dω = ω 0 ∧ (− + ω1 ) ⇒
f
df
α0 = α − (5.1)
f

5.5 Formulas for integrating factors

Define the integrating factor gω as follows:

gω : Uδ → C
5.5 Formulas for integrating factors 21

dω
Z
gω (z) = exp( − )
δk(z),z ω

Proposition 5 The following statements are true:

1. d( gωω ) = 0 in the definition domain Uδ of gω ;
2. Let a = δ (t1 ), b = δ (t2 ), 0 ≤ t1 ≤ t2 ≤ 1, and h : Σa → Σb be the holonomy along
the path δ 0 = δ |[t1 ,t2 ] , then
ω ω
h∗ ( |Σb ) = |Σ (5.2)
gω gω a

where h∗ a is the pull-back of the differential form a by h.

Proof. Let us prove the mentioned facts in a local coordinates system. Let q ∈ δ and
(U, (x, y)) be a foliation chart around q such that in this coordinates system the leaf
L p is given by y = 0 and

ω = f dy, ω1 = A(x, y)dx

Fix two points c = (x1 , 0) and d = (x2 , 0) in δ ∩U. For any point z in U, let r be the
intersection point of the leaf Lz and Σc (Figure 5.2). Define
Z z
g(z) = ω1
k(z)

In this coordinate we have

Z r Z z Z x
g(z) = ω1 + ω1 = s(y) + A(ξ , y)dξ
k(z) r x1

where s(y) only depends on y and so

Z x
dA
dg = s0 (y)dy + Adx + ( )dy ⇒ dg ∧ ω = dg ∧ f dy = ω1 ∧ ω
x1 dy

By definition, gω = eg and so
ω
d( ) = d(e−g ω) = e−g (−dg ∧ ω + dω) = e−g (−ω1 ∧ ω + dω) = 0
gω
and here the proof of the first statement finishes.
In the coordinate (x, y) we can write
ω
= Gdy
gω
∂G
The first part of the lemma implies that ∂x = 0 or equivalently G = G(y) does not
depend on the variable x.
22 5 Holonomy I

z
rz Lz Lp k(z)
p
c d δ
Σc Σd Σp
U
Fig. 5.2 Holonomy

ω ω
h∗ ( |Σ ) = h∗ (Gdy |Σd )) = h∗ (G |Σd )d(h∗ (y |Σd )) = Gdy |Σc = |Σ ⇒
gω d gω c
ω ω
h∗ ( |Σd ) = |Σ (5.3)
gω gω c

Let (Ui , (xi , yi )), i = 0, 1, 2, . . . , n be foliation charts which cover the path δ |[t1 ,t2 ] .
Let also Σci and Σdi be small transverse sections at ci , di ∈ Ui to F such that

c0 = p, ci = di−1 , Σci = Σdi−1 , i = 1, 2, . . . , n − 1, dn = b

Now our affirmation is obtained by the combination of (5.3)’s in each chart Ui .

Now we fix two transverse section section

Σ1 := Σ p , Σ2 := Σ p1 , p1 = δ (1).

Recall that Iω restricted to Σ1 is identically 1.

Corollary 1 We have
dω
Z
h∗ (ω |Σ2 ) = exp(− ) · ω |Σ1
δ z,h(z) ω

In particular, if we choose the coordinates system z in Σ1 and z̃ in Σ2 such that

ω |Σ1 = dz, z(p) = 0, ω |Σ2 = d z̃, z̃(p1 ) = 0,

then
dω
Z
h0 (z) = exp(− )
δz,h(z) ω

Proof. We have: Z
Iω |Σ1 ≡ 1, Iω |Σ2 ≡ exp( ω1 )
δk(z),z

we have
5.5 Formulas for integrating factors 23
R R
− δ ω1 − δ ω1
h∗ (e k(z),z )=e z,h(z) .
Putting these equalities in (5.2), our affirmation is proved.

Corollary 2 (Poincaré formula) Let δ be a closed path in a leaf L of the foliation

F , Σ be a transverse section at p ∈ δ to the foliation and h : Σ → Σ be the holonomy
along δ . Then
dω
Z
h0 (p) = exp( − ) (5.4)
δ ω
Proof. This is a direct consequence of the previous corollary.

The ideas of Theorem (5) come from the author’s first course in complex dynamical
system with S. Shahshahani in Iran.
Chapter 6
Singularities of holomorphic foliations

In this chapter we collect some local aspect of holomorphic foliations. We would

like to study

F (ω), ω = P(x, y)dy − Q(x, y)dx, P, Q ∈ OC2 ,0 .

with P(0) = Q(0) = 0. This study will be important for the algebraic aspects of
holomorphic foliations. For instance, the fact many holomorphic foliations do not
have algebraic invariant curves is closely related to the analysis of their singularities.

6.1 Singularities of multiplicity one

The following discussion can be found partially in [CS87] page 40 page 44–48. Let
ω = P(x, y)dy − Q(x, y)dx, with P, Q ∈ O(C2 ,0) , be a germ of a holomorphic foliation
at 0 ∈ C2 . We assume that 0 is a singularity of F (ω), this is, P(0) = Q(0) = 0.
Writing the Taylor series of ω at 0 we get

ω = ωm + ωm+1 + ...

with ωi = Pi (x, y)dy − Qi (x, y)dx such that Pi , Qi are homogeneous polynomials of
degree i. The number m is called the multiplicity of ω at 0 ∈ C2 . If m = 1 then we
say that ω1 is the linear part of ω.
In this section we are mainly interested in the germ of holomorphic foliations
with a non-zero linear part. We can use Jordan canonical form for a 2 × 2 matrix
with complex coefficients and get the following result.
Proposition 6 Let F (X) be a germ of holomorphic foliation at 0 and let 0 be a
singularity of F (X). Then, up to biholomorphisms h : (C2 , 0) → (C2 , 0), X can be
written in one of the following formats:
1. y ∂∂x + ...

25
26 6 Singularities of holomorphic foliations

2. (ax + y) ∂∂x + ay ∂∂y + ... with a 6= 0.

3. ax ∂∂x + by ∂∂y + ... with (a, b) 6= (0, 0)
Exercise 16 State and prove a similar proposition as Proposition 6 over the field of
real numbers. One have to use the Jordan canonical form of two times two matrices
over real numbers.
We are interested in foliations F (X) such that the linear part of X is of the form
X1 = ax ∂∂x + by ∂∂y . Let us analyze the foliation F (X1 ). The corresponding ordinary
differential equation and its solution passing through (x0 , y0 ) is

ẋ = ax x(t) = x0 eat
⇒
ẏ = by y(t) = y0 ebt

Exercise 17 For a leaf L of F (X1 ), describe the topological closure L̄ of L (Hint:

See [CS87] pages 44-46)
Let us calculate some holonomies. From the above equation we see that the x and y
axis are leaves of F (X1 ). We’ll name them L1 and L2 , respectively. Let p 6= 0 be in
L1 , that is p = (x0 , 0). Let also δ be the circle through p turning around 0 in L1 anti
clockwise and Σ = {(x, y) ∈ C2 |x = x0 }. We take a point z = (x0 , y) ∈ Σ and would
like to compute the action of holonomy on z. We know that δ (s) = (x0 e2πis , 0). The
analytic continuation of the leaf L of F (X) passing through z and along δ is of the
b
form δ̃ (s) = (x0 e2πis , ye2πi a s ).
For s = 1 we get the holonomy map

h:Σ →Σ
b
(x0 , y) 7→ (x0 , ye2πi a )

If we parametrize Σ by y this is simply

(C, 0) → (C, 0)
b (6.1)
y 7→ e2πi a y

Since d(x−bc yac ) = cx−bc−1 yac−1 (axdy − bydx) for c 6= 0, the foliation F (axdy −
bydx) has the first integral x−bc yac and the integrating factor c−1 x1+bc y1−ac .

6.2 Poincaré theorem

Definition 10 We say that de foliation F (ax ∂∂x +by ∂∂y +...) belongs to the Poincaré
domain if
1. a
b / R−
∈
6.2 Poincaré theorem 27

Fig. 6.1 Holonomy around a singularity

2. a
b / {2, 3, 4, ..., 12 , 31 , 14 , ...}
∈

Theorem 9 Let us assume that the holomorphic foliation F (X), X = ax ∂∂x +

by ∂∂y + ... is in the Poincaré domain. Then there exists a biholomorphism h :
(C2 , 0) → (C2 , 0) such that the pull-back of X by h is its linear part ax ∂∂x + by ∂∂y

Proof. The theorem can also be stated for vector fields in (Cn , 0) (see [Arnold]).
For this reason, we adopt the notation (x1 , x2 ) = (x, y) and (a, b) = (λ1 , λ2 ). Let

h = (u1 , u2 ) = (x1 + ξ1 (x1 , x2 ), x2 + ξ2 (x1 , x2 )) (6.2)

where ξ1 , ξ2 are two formal power series ξ j = ∑ ξ j,n xn , where n = (n1 , n2 ) is a

|n|≥2
multi index, |n| = n1 + n2 and xn = x1n1 x2n2 . We first prove the theorem formally, that
is, there is h as before such that

u˙j = λ j u j + φ j (u1 , u2 ) (6.3)

with X = (λ1 x1 + φ1 (x1 , x2 ) ∂∂x + (λ2 x2 + φ2 (x1 , x2 )) ∂∂x . This is the same as to say
1 2
that the pull-back of X is λ1 x1 ∂∂x + λ2 x2 ∂∂x . The equalities (6.2) and (6.3) imply
1 2
that
∂ξj ∂ξj
x˙j + x˙1 + x˙2 = λ j (x j + φ j ) + φ j (x1 + ξ1 , x2 + ξ2 )
∂ x1 ∂ x2
we have x˙j = λ j x j and so

∑ (λ j − n1 λ1 − n2 λ2 )ξ j,n xn = −φ j (x1 + ξ1 , x2 + ξ2 ). (6.4)

|n|≥2
28 6 Singularities of holomorphic foliations

The fact that F (X) is in the Poincaré domain implies that λ j − n1 λ1 − n2 λ2 6= 0

∀(n1 , n2 ) ∈ N2 with n1 +n2 ≥ 2. The equality in (6.4) is a recursion in ξ j,n . It follows
that we can determine the coefficients of ξ1 , ξ2 uniquely.
Now, let us check that these series are in fact convergent. Given two series
A(x1 , x2 ) and B(x1 , x2 ) with positive coefficients we say that A < B if An < Bn
∀n ∈ N20 . We denote by Ĉ(x1 , x2 ) the series C(x1 , x2 ) replacing its coefficients by
their norm and by Ĉ(x,ˆ x) the series Ĉ(x , x ) by taking x = x = x. We know that
1 2 1 2
Ĉ is convergent in |x1 | < R and |x2 | < R if Ĉˆ is convergent for |x| < R. Let us prove
ˆ ˆ
now that ξˆ1 + ξˆ2 is convergent. The fact that F (X) is in the Poincaré domain implies
that there exists a δ > 0 such that

δ < |λ j − n1 λ1 − n2 λ2 |, ∀|n| ≥ 2. (6.5)

From (6.4) we get

δ ξˆj < φˆj (x1 + ξˆ1 , x2 + ξˆ2 ) (6.6)
ˆ ˆ ˆ ˆ ˆ ˆ
⇒ ξˆ1 + ξˆ2 < δ −1 [φˆ1 (x + ξˆ1 + ξˆ2 ) + φˆ2 (x + ξˆ1 + ξˆ2 )] (6.7)
Our problem is reduced to the following one. Let F(x) ∈ O(C,0) be a convergent
series with positive coefficients and assume that its multiplicity at x = 0 is ≥ 2. By
implicit function theorem, there exists a holomorphic function y(x) ∈ O(C,0) such
that
y(x) = F(x + y(x)) (6.8)
and the multiplicity of y at x = 0 is ≥ 2. The coefficients of the Taylor series of y(x)
can be determined uniquely by the above equality, and moreover, one can check that
they are positive numbers. Let z(x) be another formal power series such that

z(x) ≤ F(x + z(x)) (6.9)

Then z(x) ≤ y(x). This follows by induction on the n-the coefficient of the inequality
z(x) ≤ y(x). The case n = 2 follows from (6.9). Note that the the coefficient of x2
in y(x) is the same as the coefficient of x2 in F(x). Assuming the m-th step of the
induction for all m < n, we again realize the n-th coefficient of z is ≤ a polynomial
combination with coefficients in N of the m-th coefficients of z(x) and F(x) for
m ≤ n. Replacing the coefficient of z(x) with the corresponding coefficients of y(x)
we get a bigger quantity which is the n-the coefficient of y(x).
Exercise 18 Show that the condition (6.5) is equivalent to the fact that F (X) is in
the Poincaré domain.
Theorem 10 (Dulac) let F (X), X = ax ∂∂x + by ∂∂y + · · · ) be a holomorphic folia-
tion in (C2 , 0) with a = nb, n ∈ N and n ≥ 2 . There is a unique biholomorphic
function h : (C2 , 0) → (C2 , 0) tangent to the identity such that the pull-back of X by
h is
∂ ∂
(ax + cyn ) + (by ). (6.10)
∂x ∂y
6.3 Siegel domain 29

Exercise 19 The proof is similar to the proof of the Poincaré theorem . Explain the
details?
Exercise 20 Use a computer and draw F (X) with X as in (6.10) for a = n = 2, b =
1.
Definition 11 A biholomorphism h : (C2 , 0) → (C2 , 0) is tangent to the identily if
h = (x + ζ1 (x, y) , y + ζ2 (x, y)) where the multiplicity of ζ1 , ζ2 at 0 is ≥ 2 .

6.3 Siegel domain

In the Poincaré theorem we have excluded a class of holomorphic foliations and it

is natural to ask whether they are also linearizable. Let us start with the definition of
such a class.

Definition 12 The foliation F (X), X = ax ∂∂x + by ∂∂y + · · · is in the Siegel domain

if
a
ab 6= 0, ∈ R−
b
We say that F (X) has resonance if a
b ∈ Q− .
For a holomorphic foliation F (X) in the Siegel domain and without resonance,
we have still the formal power series h conjugating X with its linear part. However,
it can happen that this formal power series is not convergent.
Definition 13 We say that F (X) is of type (c, v), c, v > 0 if

|a − n1 a − n2 b| c
> ∀ n1 , n2 ≥ 0 n1 , n2 ∈ N
|b − n1 a − n2 b| (n1 + n2 )v

Theorem 11 (Siegel) If F (X) is of type (c, v) then there exist a local biholomor-
phism h : (C2 , 0) → (C2 , 0) such that the pulled backed of X by h is the linear part
of X, that is, ax ∂∂x + by ∂∂y .

Proof. See [Arn80].

It is natural to define the sets:

SD := {(a, b) ∈ C2 | ba ∈ R− }
f := {(a, b) ∈ SD | (a, b) o f type (c, v) f or some (c, v) , c, v > 0}
SD

Exercise 21 Is SD
f dense in SD? Give examples of elements of SD
f and SD\SD.
f See
[CS87] and the references therein.
30 6 Singularities of holomorphic foliations

Fig. 6.2 Poincaré linearization theorem

6.4 Separatrix

Let F (X), X := P ∂∂x + Q ∂∂y and ω := P(x, y)dy − Q(x, y)dx be a germ of holomor-
phic foliation in (C2 , 0), and let 0 ∈ C2 be a singularity of F (ω) .
Definition 14 For f ∈ O(C2 ,0) , f (0) = 0, f = 0 is a separatrix of F (ω) ,if
ω ∧ d f = f .η for some η ∈ Ω(C2
2 ,0)

In the global (algebraic) context we say that { f = 0} is F (ω)−invariant .

Theorem 12 (Camacho-Sad) The germ of any holomorphic foliation F(ω) in (C2 , 0)
has a separatrix .
This will be proved after introducing the note of ”blow up” of singularities. We have
already proved it for the following particular cases.
1. If F (ω) is in the Poincaré domain or it’s in the Siegel domain and it is of type
(c, v). In this case it has more then two separatrix . Note that {x = 0} and {y = 0}
are separatrices of F (X1 ) and so {h1 = 0} and {h2 = 0} are two separatices of
F (X).
2. In Dulac’s theorem F (X) has at lest one separatrix. In Dulac’s theorem we have
conjugated X with (ax + cyn ) ∂∂x + by ∂∂y ), a = nb, n ≥ 2 and {y = 0} is a sepa-
ratix.

6.5 Singularities with resonance

Recall the definition of a germ of holomorphic foliation F (X) with resonance

in Definition (12). In the resonance case note that if we write ab = − mn (m, n) =
1 , n, m ∈ N then we have
6.5 Singularities with resonance 31

a − a(m + 1) − bn = 0
n+m+1 ≥ 2
b − am − b(n + 1) = 0

In this case, the coefficients ζ1,(m+1,n) , ζ2,(m+1,n) cannot be determined in the recur-
sion given in the proof of Theorem 9.
Theorem 13 Let F (X) ,X := ax ∂∂x + by ∂∂y +... be a germ of holomorphic foliation
in (C2 , 0) and assume that ab ∈ Q− (the resonance case). Then there is a biholomor-
phism h := (C2 , 0) → (C2 , 0) such that the pull-back of X by h is of the format

∂ ∂
X̃ := (ax + xyA(x, y)) + (by + xyB(x, y))
∂x ∂y

where A, B ∈ O(C2 ,0) with A(0) = B(0) = 0.

In the above theorem the foliation F (X) has at least two separatrices because F (X̃)
has two Separatrics {x = 0}and{y = 0}.
Proof. Proceeding as in theorem 9 we write X as

u˙j = λ j u j + φ j (u1 , u2 ) (6.11)

where
u j = x j + ξ j (x1 , x2 ) (6.12)
We need to find ξ j (x1 , x2 ) such that h∗ Xis of the form x˙j = λ j x j + ψ j (x1 , x2 ) where
ψ j (x1 , x2 ) ∈ (x1 · x2 ). Here, (x1 · x2 ) denotes the ideal of analytic functions generated
by x1 x2 . Making the same substitutions we get that

∂ξj ∂ξj
∑ (n1 λ1 +n2 λ2 −λ j )ξ j,n xn + ∑ ψ j,n xn = φ j (x1 +ξ1 , x2 +ξ2 )−
∂ x1
ψ1 −
∂ x2
ψ2
|n|≥2 |n|≥2
(6.13)
We define
• if xn ∈
/ (x1 · x2 ) take ψ j,n = 0
• if xn ∈ (x1 · x2 ) take ξ j,n = 0
If xn ∈
/ (x1 · x2 ) then the coefficient n1 λ1 + n2 λ2 − λ j 6= 0. It follows that we can
calculate ξ1 , ξ2 formally. To see that they are convergent we claim that if xn ∈
/ (x1 ·x2 )
then ∃δ > 0 such that |n1 λ1 + n2 λ2 − λ j | > δ . From this the calculation of ξ j,n is
done by
∑ (n1 λ1 + n2 λ2 − λ j )ξ j,n xn = φ j (x1 + ξ1 , x2 + ξ2 ) mod(x1 · x2 )
|n|≥2

and so δ ξˆj < φˆj (x1 + ξˆ1 , x2 + ξˆ2 ), which imply

ˆ ˆ ˆ ˆ ˆ ˆ
δ (ξˆ1 + ξˆ2 ) < φˆ1 (x + ξˆ1 + ξˆ2 ) + φˆ2 (x + ξˆ1 + ξˆ2 )

and so we proceed as in theorem 6.1.

Chapter 7
Projective spaces

In algebraic geometry many theorems are stated for compact/complete varieties. A

typical example is the Bezout theorem on the number of intersections of two curves.
Curves in A2k may not intersect each other at all, even if we assume that k is an
algebraically closed field. In this case there are many intersection points at infinity,
and we are going explain what means infinity in this case. Holomorphic foliations
are also best viewed in a compactification of A2k .

7.1 Projective spaces as complex manifolds

The projective space of dimension n as a complex manifold is defined as follows:

Pn = (Cn+1 − {0})/ ∼

where
a, b ∈ Cn+1 − {0}, a ∼ b ⇔ a = kb, for some k ∈ C − {0}.
For the purpose of the present text, we will mainly use P1 and P2 . The projective
space of dimension one P1 is covered by two charts x, x0 biholomorphic to C and the
transition map is given by
1
x0 = .
x
The projective space of dimension two P2 is covered by three charts (x, y), (u, v), (u0 , v0 )
biholomorphic to C2 and the transition maps are given by
y 1 x 1
v= , u= , v0 = , u0 = .
x x y y

Considering the chart (C2 , (x, y)), P2 becomes a compactification of C2 .

33
34 7 Projective spaces

7.2 Projective spaces as schemes

In this section we define the projective space of dimension two P2k over an arbitrary
field. We also explain the main idea behind the definition P2k as a scheme.
By the affine scheme A2k , we simply think of the polynomial ring k[x, y]. Open
subsets of A2k are given by the localization of k[x, y]. We will need two open subsets
of A2k given respectively by

1 1
k[x, y, ] and k[x, y, ]
y x

By the projective scheme P2k we mean three copies of A2k , namely

k[x, y], k[x, z], k[y, z]

together with the isomorphism of affine subsets:

1 1 x 1
k[x, y, ] ∼
= k[x, z, ], x 7→ , y 7→ (7.1)
y z z z
1 1 1 y
k[x, y, ] ∼
= k[y, z, ], x 7→ , y 7→
x z z z
1 1 1 z
k[x, z, ] ∼
= k[y, z, ], x →7 , z→7
x y y y
The best way to see these isomorphisms is the following. We look at an element of
k[x, y] as a function on the k-rational points k2 of the first chart and for (a, b) ∈ k2 ,
we use the identities
a 1 b 1
[a; b; 1] = [ ; 1; ] = [1; ; ].
b b a a
Let C be a curve in A2k given by the polynomial f (x, y) ∈ k[x, y]. It induces a curve
C in P2k in the following way. Let us define f1 := f and

x 1 1 y
f ( , ) = z−d f2 (x, z), f ( , ) = z−d f3 (y, z)
z z z z

We think of the the curve C in the same way as P2k , but replacing k[x, y] with
k[x, y]/h f1 i and so on. Here, h f1 i is the ideal k[x, y] generated by a single element f1 .
We can also think of C in the same way as P2k but with the following additional
relations between variables:

f1 (x, y) = 0 in k[x, y]

f2 (x, z) = 0 in k[x, z]
and
f3 (y, z) = 0 in k[y, z].
7.3 Foliations in projective spaces 35

The above discussion does not use the fact that k is a field. In fact, we can use an
arbitrary ring R instead of k. In this way, we say that we have an scheme C over the
ring R. The function field of the projective space P2k is defined to be

k(P2k ) := k(x, y) ∼
= k(x, z) ∼
= k(y, z).

where the isomorphisms are given by (7.1). The field of rational functions on the
curve C is the field of fractions of the ring k[x, y]/h f1 i. Using the isomorphism (7.1),
this definition does not depend on the chart with (x, y) coordinates. We can also think
of k(C) as k(x, y) but with the relation f1 (x, y) = 0 between the variables x, y. Any
f ∈ k(C) induces a map
C(k) → k
that we denote it by the same letter f .

7.3 Foliations in projective spaces

A foliation F (ω), ω = Pdy − Qdx extends to a holomorphic foliation in P2k . For

instance, in the chart (u, v) we have

1 v v 1 v 1 P̃(u, v)dv − Q̃(u, v)du

ω = P( , )d( ) − Q( , )d( ) = , (7.2)
u u u u u u ud+2
P̃, Q̃ ∈ k[u, v].

Definition 15 The smallest number d in the equality (7.2) is called the (projective)
degree of the foliation F (ω).
It is also natural to define the (affine) degree of F (ω):

deg(F ) := max{deg(P), deg(Q)}.

These two notions if degree are different. Working with foliations in P2k it is useful
to use the projective degree.
Proposition 7 A line in P2 which does not cross any singularity of F has a fixed
number d (counted with multiplicity) of tangency points with the foliation F . In
particular, for a generic line we have exactly d simple tangency points. This number
d is the projective degree of F .
Proposition 8 A foliation of the projective degree d in the affine coordinate A2k ⊂ P2k
is given by the differential form:

Pdx + Qdy + g(xdy − ydx)

36 7 Projective spaces

where either g is a non-zero homogeneous polynomial of degree d and deg(P), deg(Q) ≤

d or g is zero and max{deg(P), deg(Q)} = d. In the first case the line at infinity is
not invariant by F and in the second case it is invariant by F .
We may redefine F (d) to be the set of holomorphic foliations of projective de-
gree d in P2 . The spaces F (d) corresponding to two different definitions of the
degree have different aspects. For instance, a generic foliation of projective degree
d does not have an algebraic solution and a generic foliation of (affine) degree leaves
the line at infinity invariant.

7.4 Ricatti foliations

Another compactification of A2k = A1k × A1k is P1 × P1 which is useful for studying

the Riccati foliations is given by:

ω = q(x)dy − (p0 (x) + p1 (x)y + p2 (x)y2 )dx, p0 , p1 , p2 , q ∈ k[x].

1
Substituting y = y0 we have

1
ω= (−q(x)dy − (p0 (x)y2 + p1 (x)y + p2 (x))dx)
y02

and so all the projective lines {a ∈ C | q(a) 6= 0} × P1 are transversal to the foliation.
This will be later used to define the global holonomy of Ricatti foliations.

7.5 Minimal set

For a holomorphic foliation in P2k one may formulate many problems related to the
accumulation of its leaves. The most simples one which is still open is the following:

Problem 7.1. Is there a foliation F in P2 with a leaf L which does not accumulate
in the singularities of F .
For instance the above problem for Jouanolou foliation is proved numerically for
d ≤ 4 and it is still open for general d.
Let us suppose that such an F and L exist and set M := L̄, where the closure is
taken in P2 . It follows that M is a union of leaves of F . We may suppose that M
does not contain a proper F -invariant subset. In this case we call M a minimal set.
Proposition 9 A foliation in P2 with algebraic leaf has not a minimal set.
For many other useful statement on minimal sets see [CLNS88]. For local theory of
holomorphic foliations see [CS87].
Chapter 8
Camacho-Sad theorem

In this chapter we explain one of the main index theorems is holomorphic foliations,
namely the Camacho-Sad index theorem. This together with Baum-Bott index the-
orem and a a local analysis of holomorphic foliations around singularities, are our
main tools in order to study the non-existence of invariant algebraic curves for holo-
morphic foliations.

8.1 Camacho-Sad index

Let F (ω), ω := Pdy − Qdx, P, Q ∈ O(C2 ,0) be a germ of holomorphic foliation

in (C2 , 0) and assume that 0 ∈ C2 is an isolated singularity of F , that is, P(0) =
Q(0) = 0 and P and Q do not have common factors. Let also f ∈ O(C2 ,0) and { f = 0}
is a separatrix of F , that is,
2
d f ∧ ω = f .η where η ∈ Ω(C2 ,0) .

Proposition 10 There exist holomorphic functions g, h ∈ O(C2 ,0) and η ∈ Ω(C

1
2 ,0)
such that h is not divisable by f and

gω = h · d f + f η.

Proof. Since f = 0 is a separatrix, we have d f ∧ ω = f .η and so fx .P + fy .Q = f S

for some S ∈ O(C2 ,0) . Then

fy .ω = fy (Pdy − Qdx) = ( fy .P)dy − ( f .S − fx .P)dx = Pd f − f (Sdx)

The same statement is true if one replaces O(C2 ,0) with k[x, y] and ”separatrix” with
”invariant algebraic curve”.

37
38 8 Camacho-Sad theorem

Fig. 8.1 A singular separatrix

Theorem 14 (Puiseux parametrization) Let C = { f (x, y) = 0}, f ∈ O(C2 ,0) be a

germ of a curve in (C2 , 0) . There is a holomorphic map γ : (C, 0) → (C2 , 0) such
that f (γ(t)) = 0 and γ is a bijection between (C, 0) and { f (x, y) = 0}.

We will prove this theorem later when we introduce the notion of a blow-up. For
now, we only mention that the above theorem is trivial for smooth curves. If { f = 0}
is smooth at 0 , that is ( ∂∂ xf (0), ∂∂ yf (0)) 6= (0, 0) then one can find γ using implicit
function theorem. Another example is the singular curve given by f = y2 − x3 . It
has the parametrization given by γ(t) = (t 2 ,t 3 ). From now on let γ be a path in
C = { f = 0} which is the image of a path in (C, 0) turning around 0 anti-clockwise
and under the map γ. Recall the definition of dω ω from §5.4.
Definition 16 The Camacho-Sad index of (F ,C, 0) is

−1
Z
η
I(F,C, 0) := .
2πi h
γ

Note that λ = e2iπI(F,c) is the multiplier of the holonomy h of F (ω) along the path
γ.
We can reinterpret Proposition 10 in the following way. There is a meromorphic
1-form Ω in (C2 , 0) which induces the foliation F and

df
η̃ := Ω −
f
has no poles along f = 0. Actually, this 1-form η̃ is unique restricted to f = 0.
With the notation of Proposition 10 we write gω = h · d f + f · η and we have Ω =
g η
f h , η̃ = h . Note also that if we define ω̃ = f Ω = d f + f η̃ then

d ω̃
= η̃, restricted to f = 0.
ω̃
8.1 Camacho-Sad index 39

This follows from

d ω̃ d(d f + f η̃) d( f η̃) (d f ∧ η̃) + f .d(η̃)
= = =
ω̃ ω̃ ω̃ ω̃
(1) (d f ∧ η̃) (d f + f η̃) ∧ η̃
= = = η̃
ω̃ ω̃
For (1) we restrict to f = 0. Note that it makes sense to say that the restriction of
f dω̃η̃ to { f = 0} is zero, becuase η̃ has no poles along f = 0.
If we take ω = Pdy − Qdx then we have
∂P ∂Q ∂P ∂Q
dω ∂x + ∂y ∂x + ∂y
=− dx = − dy (8.1)
ω P Q

The second equality is valid when it is restricted to the leaves of F (ω). Note that
the residue of dω
ω in a separatrix differes from the Camacho-Sad index by an integer.
However, if we take the differential 1-form
1 Q
ω̂ = ω = dy − dx (8.2)
P P
then we have we have  
Q
d ω̂ ∂
P
=− dx (8.3)
ω̂ ∂y
and
Proposition 11 If the curve f = 0 is smooth and it is not tangent to the y axis at 0
then the Camacho-Sad index can be computed using dω̂ω̂ , that is,

−1 d ω̂
Z
I(F,C, 0) := .
2πi ω̂
γ

Proof. From the hypothesis it follows that fy has not zeros in (C2 , 0). From an-
other side we have ω̃ = fy ω̂ which follows from the explicit construction of η in
Proposition 10. Therefore,
d ω̃ d fy d ω̂
=− +
ω̃ fy ω̂
and the proof follows.

Exercise 22 Let ω = Pdy − yQdx and so y = 0 is a separatrix of F (ω) calculate

I(F, 0).

Sometimes we write I(F,C) = I(F,C, 0), being clear in the context which singularity
we are dealing with.
40 8 Camacho-Sad theorem

Fig. 8.2 Tangency

8.2 Residue formula

The notion of a residue is purely algebraic and we can avoid integrals in it definition,
see for instance [Tat68] and Serre’s book in this article. Therefore, the Camacho-Sad
index can be defined for foliations in P2k for arbitrary field k.
The residue formula for smooth curves.
Theorem 15 Let C ⊂ P2k be a smooth curve and let ω be a meromorphic differential
1-form in C. We have
∑ residue p (ω) = 0.
p∈C

Proof. We prove this for k = C. The curve C over C is naturally a Riemann surface.
1 H
By definition residuep (ω) = 2πi p ω. Since dω = o, by the Stokes theorem
I Z Z
∑ ω= n
S dω = 0
pi pi X\ Di
i=1

8.3 Camacho-Sad theorem

Theorem 16 Let F be a holomorphic foliation in P2k and let C be a smooth alge-

braic F -invariant curve of degree d in P2k , then

∑ I(F ,C, p) = d 2 .
p∈Sing(F )∩C

Proof. First of all note that can we choose a line P1k ⊂ P2k and we can write the
foliation F (ω), ω = P(x, y)dy − Q(x, y)dx, in the coordinates (x, y) of the affine
chart A2k = P2k \P1k such that
1. The smooth algebraic curve C ⊂ P2 intersects P1k transversely in d points.
8.3 Camacho-Sad theorem 41

Fig. 8.3 Tangency

2. In the affine chart A2k , the vertical lines x = c are either transversal or have tangen-
cis of order two with the curve C. In addition, all the tangenct points are regular
points of the foliation F . By the Bezout theorem the number of such tangency
points is d(d − 1).
Now consider the differential form (8.2) which induces the foliation F and let η
be the differential 1-form in (8.3) multiplied with −1. The poles of the 1−form
η restricted to C are divided in three groups: 1. Singularities of F in C 2. The
tangency points of the curve with vertical lines 3. The intersections of C with the
line at infinity. We compute the residue of η around all these points and use the
residue formula in Theorem 15 and we get the proof.
For a singular point p ∈ C of F , by definition we have Residue(η, p) = I(F ,C, p).
Therefore, we do not need to compute it. For tangency points, we can locally pa-
rameterized a leaf tangent to a vertical line by L : x = g(y) = t.y2 + · · · . For sim-
P(x,y)
plicity we assume that such a tangency point is at (0, 0). Since Q(x,y) = dx
dy , we have
P(g(y),y)
Q(g(y),y) = g0 (y). Therefore,

∂ ( Q(g(y),y)
P(g(y),y) ) g00
=− .
∂y (g0 )2

From this we get

∂ ( Q(g(y),y)
P(g(y),y) )
η |L = g0 (y)dy
∂y
g00
=− dy
g0

This has residue −1 at the tangecy point p and so in total we get −d(d − 1).
42 8 Camacho-Sad theorem

Now let us calculate the residue of η for a point p ∈ C in the third group. The
differential form (8.2) has a pole order −1 at infinity. Using the formula (5.1), we
conclude that the residue of dω̂ω̂ = −η at p is +1 and so in total we get −d for
residues of η for the third group. Finally, by residue formula we have

∑ I(F ,C, p) − d(d − 1) − d = 0.

p

Exercise 23 Discuss the Camacho-Sad index and theorem for arbitrary field k in-
stead of C.
∂P + ∂Q
Exercise 24 If we use dω
ω =−
∂x
P
∂y
dx in the definition of Camacho-Sad index
what would be the corresponding Camacho-Sad theorem. Repeat the same proof as
in Theorem 16.
Chapter 9
Baum-Bott index and Theorem

In this chapter we introduce the Baum-Bott index of holomorphic vector fields. The
original articles [?] deals with vector fields in arbitrary dimensions, however, in the
present chapter we focus on dimension two.

9.1 Baum-Bott index

Let F (ω), ω = Pdy − Qdx be a holomorphic foliation in U := (C2 , 0) with a sin-

gularity at 0 ∈ C2 . The differential form dω
ω that we defined in §6.1 is well-defined
in the leaves of F , however, any realization of it in (C2 , 0) might be meromorphic.
There is a way to avoid meromorphicity, allowing complex conjugate of holomor-
phic functions (real analytic functions). This is as follows.
We define a C∞ differential (1, 0)-form η in U\{0}:

P̄dx + Q̄dy
η := (Px + Qy )( ) (9.1)
|P|2 + |Q|2

Proposition 12 The differential form η satisfies the following properties:

1. η ∧ ω = dω.
2. η ∧ dη is a closed form, that is, d(η ∧ dη) = 0.
3. If ω̂ = Rω and η̂ is defined as in (9.1) using ω̂ then (η̂ ∧ d η̂ − η ∧ dη) is an
exact form.

Proof. The first and second item are easy and are left to the reader. For the third
item we proceed as follows. We have

dR
η̂ − η = + fω
R
where f is given by

43
44 9 Baum-Bott index and Theorem

Rx Q̄ − Ry P̄
f := ,
R(|P|2 + |Q|2 )
and so
 
dR dR
η̂ ∧ d η̂ − η ∧ dη = d fη ∧ω +
∧η + f ∧ω
R R
 
dR
=d ∧ η − f ω ∧ η̂ (9.2)
R

The second and third item in Proposition 12 as above are equivalent to say that η
induces an element in HdR3 (U\{0}) which depends only on the foliation F and not

the differential 1-form ω. where α = f ω

Definition 17 The Baum-Bott index of a foliation F defined in (C2 , 0) is defined
to be
1
Z
BB(F , 0) := Residue(η ∧ dη, 0) = 2 η ∧ dη
4π S
Here, S = S3 (0, r) is the sphere of dimension three, radius r and the center 0 ∈ C2 ,
for a small positive number r.
It is an easy exercise to show that if ηi , i = 1, 2 are two C∞ differential (1, 0)-
forms in U\{0} such that they satisfy the item 1 and 2 of Proposition 12 then η2 ∧
dη2 − η1 ∧ dη1 is exact and so in the definition of the Baum-Bott index we can use
any C∞ differential (1, 0)-form in U\{0} satisfying item 1 and 2 of Proposition 12.
In particular, this implies that the Baum-Bott index is invariant under coordinates
change. This fact is going to be used in the next proposition.
Proposition 13 Let F be a foliation in an open subset U of C2 and let A be an
open sub set of U such that the topological closure of A in C2 is inside U and the
boundary of A is compact and smooth. Further, assume that there is no singularity
of F in the boundary of A. Then
1
Z
∑ BB(F, p) = η ∧ dη
T
p∈A Sing(F )
4π 2 ∂A

Proof. This follows from Stokes’ theorem and the fact that η ∧ dη is closed:
Z Z Z
η ∧ dη − ∑ η ∧ dη = d(η ∧ dη) = 0
∂A ∂ Si A

where S3i are small spheres around the singularity pi of F in A.

Proposition 14 Let X = P ∂∂x +Q ∂∂y be a vector field in U = (C2 , 0) with an isolated

singular point at 0 ∈ C2 and let F = F (X) be the induced foliation. Assume that
9.1 Baum-Bott index 45

2.png

Fig. 9.1 U

 
Px (0) Py (0)
A = DX0 = .
Qx (0) Qy (0)

has a non-zero determinant. We have

1 trace(A)2
Z
BB(F, p) = η ∧ dη = ,
4π 2 S3 det(A)

Proof. An explicit calculation shows that

(Px + Qy )2
η ∧ dη = (Q̄d P̄ − P̄d Q̄) ∧ dx ∧ dy
|P|2 + |Q|2

Let D := det(A) and T := trace(A). We consider two cases.

1-X is a linear vector field. By our hypothesis, ϕ(x, y) = (P, Q) = (u, v) is a bi-
homeomorphism (a coordinate change). We have

T2 1
θ = ϕ ∗ (η ∧ dη) = (−ūd v̄ + v̄d ū) ∧ ( (du ∧ dv))
(|u|2 + |v|2 ) D

We integrate θ over S3 = (|U|2 + |V |2 = 1) and use the stokes theorem. We have

T2 T2
Z Z Z
θ= (ūd v̄ − v̄d ū) ∧ du ∧ dv = 2du ∧ d ū ∧ dv ∧ d v̄
D D
S3 S3 B

where B = B(0, 1) is the unit ball with the center 0. Since du ∧ d ū ∧ dv ∧ d v̄ = 4dV ,
where dV is the Euclidean volume form of C2 , we get

1 T2
Z
BB(F , 0) = θ=
4π 2 D
S3

2-The general case. Let us write P = P1 + R and Q = Q1 + S, where P1 , Q1 are

linear and the vanishing order of R and S at 0 is ≥ 2. Consider the function
46 9 Baum-Bott index and Theorem

Ht : C2 → C2 , Ht (p) = t p

For 0 < t ≤ 1 we have S3 ⊆ Ht−1 (B(0, 2)) = B(0, 2t ) where θt = Ht∗ (η ∧ dη) and so

1
Z
BB(F, 0) = θt
4π t S3

We have

(∆t)2
Ht∗ (η ∧dη) = [(Q̄1 + S̄t )d(P̄1 + R̄t )−(P̄1 + R̄t )d(Q̄1 + S̄t )]∧dx∧dy
(|P1 + Rt |2 + |Q1 + St |2 )2

where ∆t := T + Rt ◦ Ht + Sy ◦ Ht and Rt := t −1 (S ◦ Ht ) , St = t −1 (S ◦ Ht ). We take

the limit t → 0 and we see that ∆t uniformly converges to T and Rt , St uniformly
converges to zero. Therefore, θt converges uniformely to θ0 which is derived from
the linear part of X. Using the first case we get the result.

Theorem 9.1. (Baum-Bott in P2C ). Let F be a foliation of degree k in P2C with iso-
lated singularities. Then

∑ BB(F , p) = (k + 2)2 (9.3)

p∈Sing(F )

Proof. After taking a proper affine chart E0 = {[1; x; y]|x, y ∈ C} ⊂ P2C and a multi-
lication of X with a constant, we can assume that Sing(F ) ⊂ P0 , where

P0 := {[1, x, y]| |x| < 1, |y| < 1}.

Let us consider the other affine charts E1 = {[u; 1; v]|u, v ∈ C} and E2 = {[z; w; 1]|z, w ∈
C} and the corresponding polydisces

P1 = {[u, 1, v]| |u| < 1, |v| < 1}, P2 = {[u0 , v0 , 1]| |z| < 1, |w| < 1}.

We notice that [ [
P2C = P̄0 P̄1 P̄2 , ∂ Pi = ∪ (P̄i ∩ P̄j ).
j6=i

Let Xi be a polynomial vector field which induces the foliation F in Ei , i = 0, 1, 2

and let ωi be the polynomial 1-form such that ωi (Xi ) = 0. Let also φi j : Ei → E j be
the change of coordinates between Ei and Ei . We have
∗
φ10 (ω0 ) = u−(k+2) ω1 , φ20
∗
(ω0 ) = (u0 )−(k+2) ω2
1
In other words in the intersection Ei ∩ E j we have ωi = fi j ω j , where fi j = f ji and

yk+2
f01 |E0 = xk+2 , f02 |E0 = yk+2 , f12 |E0 = .
xk+2
We have fi j f jk fki = 1 which implies that
9.1 Baum-Bott index 47

d fi j d f jk d fki
+ + = 0, ∀i, j, k ∈ {1, 2, 3}.
fi j f jk fki

Now let us consider η j such that dω j = η j ∧ ω j . We have

 
d fi j
ηi ∧ ωi = dωi = d( fi j ω j ) = d fi j ω j + fi j dω j = + η j ∧ ωi (9.4)
fi j

and in a similar way as in the proof of the third part of Proposition 12, there are C∞
functions gi j in Ei ∩ E j \ Sing(F ) such that

d fi j
ηi = η j + + gi j ωi
fi j

Let αi j := gi j ωi and so αi j + α jk + αki = 0. Using (9.2) we get

 
d fi j
Θi −Θ j = d ∧ η j + ηi ∧ αi j (9.5)
fi j

where Θi := ηi ∧ dηi . Using these and the fact that Sing(F ) ⊂ P0 we have that
Z Z Z Z
4π 2 ∑ BB(F , p) = Θ0 = Θi + Θ1 + Θ3
p∈Sing(F ) ∂ P0 ∂ P0 ∂ P1 ∂ P3
Z Z Z Z Z Z
= Θ0 + Θ0 + Θ1 + Θ1 + Θ2 + Θ2
P01 P02 P10 P12 P20 P21
Z Z Z
= (Θ0 −Θ1 ) + (Θ1 −Θ2 ) + (Θ2 −Θ0 )
P01 P12 P20
Z
= α
T

where
d f01 d f12 d f20
α = −η1 ∧ + η0 ∧ α01 − η2 ∧ + η1 ∧ α12 − η0 ∧ + η2 ∧ α20 (9.6)
f01 f12 f20

and Pi j = P̄i ∩ P̄j and T is the two dimensional torus which is the boundary of all
Pi j ’s. For the last equality we have used the Stokes’s theorem. We substitute ηi =
η0 + dffi0i0 + αi0 in the expression of α and we have

d f10 d f01 d f20 d f12

α = −(η0 + + α10 ) ∧ + η0 ∧ α10 − (η0 + + α20 ) ∧ +
f10 f01 f20 f12
d f10 d f20 d f20
(η0 + + α10 ) ∧ α12 − η0 ∧ + (η0 + + α20 ) ∧ α20
f10 f20 f20
d f20 d f12 d f01 d f12 d f10 d f20
=− ∧ − α01 ∧ − α20 ∧ + ∧ α12 + ∧ α20
f20 f12 f01 f12 f10 f20
48 9 Baum-Bott index and Theorem

Now, we just note that

d f01 d f12 d f10 d f20

α01 ∧ + α20 ∧ − ∧ α12 − ∧ α20
f01 f12 f10 f20
 
d f12 d f20 d f01 d f10
= α20 ∧ + + α10 ∧ − ∧ α12
f12 f20 f01 f10
 
d f01 d f01 d f10
= α20 ∧ − + α10 ∧ − ∧ α12
f01 f01 f10
d f01
= ∧ (α01 + α12 + α20 ) = 0
f01
Hence α = − dff2020 ∧ dff1212 = (k + 2)2 dx dy
x ∧ y . Using the parametrization (x, y) =
(eiθ , eiψ ) with θ , ψ ∈ (0, 2π) we obtain that

dx dy
Z Z
4π 2 ∑ BB(F , p) = α = (k + 2)2 ∧ = 4π 2 (k + 2)2
p∈Sing(F ) T T x y

9.2 Applications of Baum-Bott and Camacho-Sad index

theorems

Let F be a foliation of degree k in P2 with finite singularities , let X be a vector field

respect to F and for each singularity Pi of F then DX pi has non zero eigenvalues
a
a j , b j and b jj 6∈ Q+ . We say that F ∈ Ak .
Let X be a vector field and define a foliation F and pi is singularity of F such
a
that DX pi has two non zero eigenvalues such that b jj 6∈ Q+ then at this point there is
exactly two separatrix .?

Proposition 15 Let F be a non degenerate foliation of degree k over P2 then

|Sing(F )| = 1 + k + k2

this means the number of singularities of F is k2 + k + 1 .

Suppose that α j , β j are separatrix of F at p j , let X be a vector field respect to F
and at neighborhood of p j ,DX p j has two eigenvalues a j , b j such that

bj aj
I(F , α j ) = and I(F , β j ) =
aj bj

Definition 18 A configuration associated to F is a subset of all separatrix of F

sep(F ) = {a j , b j | j = 1, ..., N}
9.2 Applications of Baum-Bott and Camacho-Sad index theorems 49

Fig. 9.2 separatrix

We say a configuration C is proper if C 6= sep(F )

Given a configuration C sep(F ) then we use the notation

I(F ,C) = ∑ I(F , δ ).

δ ∈C

Observe that I(F ,C) is a

I(F ,C) = ∑ I(F , α j ) + ∑ I(F , β j )

α j ∈C β j ∈C

If V = { f = 0} is a F -invariant , we can define a configuration associated to F and

V
C(F ,V ) = {δ ∈ sep(F ) : S ⊂ V }

Proposition 16 Let F ∈ Ak , k ≥ 2 . suppose that I(F ,C) ,for all proper configu-
ration C $ sep(F ) is not positive integer then F has not F -invariant (algebraic
solution ).

Proof. Proof by contradiction , let V be F -invariant by the Comacho-Sad in-

dex we have I(F ,V ) = I(F ,C(F,V )) is a positive integer so by the hypothesis
C(F , S) = sep(F ). let us compute I(F , sep(F )) use the Baum-Bott theorem on
P2 , according to this theorem we have :
N
∑ BB(F , p) = ∑ BB(F , p j ) = (k + 2)2
p∈Sing(F ) j=1

T j2
since F at each its singularity is non-degenerate then BB(F , p j ) = D j (by the ex-
ample)

(a j + b j )2 aj bj
BB(F , p j ) = = + + 2 = I(F , β j ) + I(F , α j ) + 2
a j .b j bj aj
50 9 Baum-Bott index and Theorem

then
N
(k + 2)2 = ∑ (I(F , β j ) + I(F , α j ) + 2) = I(F , sep(F )) + 2(k2 + k + 1)
j=1
I(F , sep(F )) = −k2 + k + 1

If k ≥ 3 then (−k2 + k + 2 < 0) ,sep(F ) can not be a configuration of a algebraic

curve(F -invariant) . for k = 2 then (−k2 + k + 2 = 2) then so by Comacho-Sad
index sep(F ) can not be a configuration of a algebraic curve .

Proposition 17 For k ≥ 2 the Jouanulou J(2,k) foliation has not algebraic solution
( J − invariant) curve .

Proof. Suppose that X = (yk −xk+1 ) ∂∂x +(1−xk y) ∂∂y and J(2, k) = F (X). For every
singularities p j of F , F at p j is non degenerate (?) and
 
−(k + 1) k
DX(1, 1) =
−k −1

Therefor, the quotients of eigenvalue of DX at p j are the roots of the equation

T2 (k+1)2
Z + Z1 = D = N where N = k2 + k + 1 then the roots are
√
−k2 +2k+2+k(k+2) 3i
z1 = 2N √
−k2 +2k+2−k(k+2) 3i
z2 = 2N

In particular α1 , β1 are separattrixes of F at (1, 1) , we have I(J(2, k), α1 ) =

z1 ,I(J(2, k), β1 ) = z2 if C a proper configuration then
√
−k2 + 2k + 2 k(k + 2) 3i
I(J(2, k),C) = m.z1 + n.z2 = (m + n)( ) + (m − n)( )
2N 2N
? where 0 < m + n < 2N, note that I(J(2, k),C) is real , so m=n then I(J(2, k),C) =
2
(m)( −k +2k+2
N ) . if k ≥ 3 then I(J(2, k),C) 6∈ R or I(J(2, k),C) ≤ 0, for k = 2
,I(J(2, k),C) 6∈ R or I(J(2, k),C) = 2m 7 can not be positive integer while m < 7.
by the last proposition J(2, k) has not algebraic solution .
Chapter 10
Blow up

Similar to the case of singularities of algebraic varieties, the notion of blow-up is

essential in order to understand how complicated is the singularities of holomorphic
foliations. The final result is a Theorem of Seidenberg which says that after a finite
number of blow-ups any singularity of a holomorphic foliation, we get the so-called
reduced singularities. As in the case of previous chapters, we work in the algebraic
context of foliations in A2k , however, the whole discussion can be done for germs of
foliations in (C2 , 0).

10.1 Blow-up of a point

Let us consider the affine variety A2k with coordinates (x, y). Let also fix a point
p ∈ A2k . For simplicity we take p = (0, 0). The blow-up of A2k at p is the variety Ã2k
obtained by glueing

U0 = Spec(k[x,t]), U1 := Spec(k[u, y])

up 1.png

Fig. 10.1 blowing up

51
52 10 Blow up

via
ut = 1, y = tx.
we have the following well-defined map

ϕ : Ã2k → A2k ,

which is obtained by gluing the maps

ϕ0 : U0 → A2k , (x,t) 7→ (x, xt)

ϕ1 : U1 → A2k , (u, y) 7→ (uy, y)

We will frequently use Figure 10 to visualize a blow-up.

10.2 Blow-up and foliations

Let consider a foliation F (ω), ω = Pdx + Qdy in A2k . Let p be an isolated singular
point of F . For simplicity we assume that p = (0, 0) ∈ A2k . We write the homoge-
neous decomposition of ω

ω = ∑ ωi , ωi = Pi (x, y)dx + Qi (x, y)dy, (Pk , Qk ) 6= (0, 0)

i=k

where Pi , Qi are homogeneous polynomials of degree i and the sum is finite. In the
local context of holomorphic foliations in (C2 , 0) the above sum can be inifinte.
Definition 19 The natural number m p (F ) := k is called the algebraic multiplicity
of F at p.
Let us now analyze the foliation F after a blow-up at p. In the chart U0 with (x,t)
coordinates we have

ϕ0∗ ω = ∑ Pi (x,tx)dx + Qi (x,tx)(xdt + tdx)

i=k
= ∑ x j (Pj (1,t) + tQ j (1,t))dx + x j+1 Q j (1.t)dt
j=k

In a similar way in the chart U1 with the coordinates (u, y) we have

ϕ1∗ ω = ∑ Pi (uy, y)(udy + ydu) + Qi (uy, y)(dy)

i=k
= ∑ y j+1 Pj (u, 1)du + y j (Q j (u, 1) + uQ j (u, 1))dy
j=k

Let
Ci := xPi (x, y) + yQi (x, y), C := Ck
10.2 Blow-up and foliations 53

We consider two cases:

1-dicritical case C(x, y) ≡ 0. In this case ϕ0∗ ω and ϕ1∗ ω are divisible by xk+1
and yk+1 , respectively. In other words, ϕ ∗ ω has a zero divisor of order k + 1 along
ϕ −1 (0). Let us define
1
ω̂ = k+1 ϕ0∗ ω
x
1
ω̌ = k+1 ϕ1∗ ω
y
We write
ω̂ = [(Pk+1 (1,t) + tQk+1 (1,t))dx + Qk (1,t)dt] + xα
(10.1)
ω̌ = [Pk (u, 1)du + (Qk+1 (u, 1) + uPk+1 (u, 1)dy] + yβ
where α and β are 1-forms in A2k , and so

ω̌ = uk+1 ω̂ (10.2)

Now, we analyze the foliation ϕ ∗ F near ϕ −1 (0). For simplicity, we work in the
chart U0 with (x,t) coordinates. Since Pk (1,t) + tQk (1,t) = 0, the polynomial Qk
is not identically zero. For points q := (0,t) with Qk (0,t) 6= 0 the leaf of F pass-
ing through q is transversal to x = 0. For a point q = (0,t) with Q(1,t) = 0 and
Ck+1 (1,t) 6= 0, we still have a regular point of F , however, the leaf of F through q
is tangent to x = 0. The singularities of F in x = 0 are given by (0,t)’s where t is a
solution of
Qk (1,t) = 0, Ck+1 (1,t) = 0
The point (u, y) = (0, 0) in the chart U1 must be treated separately. The foliation
F is transversal to ϕ −1 (0) at this point if Pk (0, 1) 6= 0, that is, the homogeneous
polynomial Pk (x, y) has yk term. In a similar way, F is tangent to ϕ −1 (0) at (u, y) =
(0, 0) if Pk (0, 1) = 0 and Ck+1 (0, 1) = Qk+1 (0, 1) 6= 0. If both Pk (0, 1) and Ck+1(0,1)
vanish then we have a singularity of F at the point (u, y) = (0, 0).
2-Non-dicritical case C(x, y) 6= 0. In this case ϕ0∗ ω and ϕ0∗ ω are divisible by xk
and yk , respectively. We define
1 ∗
ω̂ = ϕ ω = [(Pk (1,t) + tQk (1,t))dx + xQk (1,t)dt] + xα
xk 0
1 ∗
ω̌ = ϕ ω = [(Qk (u, 1) + uPk (u, 1)dy + yPk (u, 1)du] + yβ
yk 1
where α are 1-forms. These are related by the equality ω̌ = uk ω̂. In this case the
Projective line ϕ −1 (0) is invariant by the foliation F . The singularities of F in
{x = 0} = ϕ −1 (0) ∩ U0 are given by (0,t) with C(1,t) = 0. It has a singularity at
(u, y) = (0, 0) in the chart U1 if Qk (0, 1) = 0.
Now, we consider some examples.
Example 2 Let
ω = (yk + 2yx2k−2 )dx − x2k−1 dy, k ≥ 2.
54 10 Blow up

up 2.png

Fig. 10.2 blowing up

We have m0 (F ) = k and C(x, y) = xyk 6= 0. In the chart U0 the foliation ϕ ∗ F is

given by ω̂ = (t k + txk−1 )dx − xk dt, and so, it has a singularity of multiplicity k at
(x,t) = (0, 0). In U1 we have

ω̌ = (u + u2k−1 yk−1 )dy + (yk + 2yk u2k−2 )du

and so ϕ ∗ F has a singularity of multilicity 1 at (u, y) = (0, 0). We do one more

blow-up at (x,t) = 0. For simplicity we redefine (x, y) := (x,t), reuse t, u as before
and redefine
ω = (yk + yxk−1 )dx − xk dy
Given y = tx then we have η̂(x,t) = t k dx − xdt , it shows that at x = 0 and t = 0
has algebraic multiplicity 1. Finally , when x = uy , then η̂(u, y) = udy − uk−1 ydu
again we found the algebraic multiplicity at u = 0 , y = 0 is 1 .

Example 3 Determination of typical leaf ω = d(x3 − y2 = 3x2 dx − 2ydy = 0) . The

point p = (0, 0) is singularity of F (ω) with algebraic multiplicity one(m p (F ) =
1). Since C1 = −2y2 6= 0 ,ϕ −1 (0) is F -invariant and ω̂ = (3x − 2t 2 )dx − (2xt)dt
has singularity at (x,t) = (0, 0).We redefine (x, y) := (x,t) so will be ω = (3x −
2y2 )dx − 2xydy and F = F (ω),it has algebraic multiplicity one at the new point
p = (0, 0) ,i.e,m p (F ) = 1 and again since c1 = 3x2 6= 0 so ϕ −1 (0) is F -invariant and
ω̌ = (3u2 − 4yu)dy + (3uy − 2y2 )du, it has singularity at p = (0, 0) , so by again we
redefine (x, y) := (u, y) then ω = (3xy + 2y)dx + (3x2 + 4x)dy and given F = F (ω)
it has a singularity at p = (0, 0) with m p (F ) = 2. note that C1 = 6x2 y − 6xy2 6= 0 so
ϕ −1 (0) is F -invariant and after the pulled back of ω we have that

ω̂ = (6t − 6t 2 )dx + (3x − 4xt)dt

ω̌ = (6u2 − 6u)dy + (3uy − 2y)du
10.3 Multiplicity along an invariant curve 55

up 3.png

Fig. 10.3 blowing up

10.3 Multiplicity along an invariant curve

Let F be a foliation in A2k with an isolated singularity at 0 ∈ A2k . Let us assume that
S := {y = 0} is F -invariant, and hence,

ω = yα(x, y)dx + β (x, y)dy, β (0, 0) = 0

The differential 1-form ω evaluated at (x, 0) is of β (x, 0)dy.

Definition 20 The multiplicity of F along S at the singular point p = (0, 0) ∈ S is
defined to be the multiplicity of β (x, 0) at x = 0. It will be denoted by the m p (F , S).

Proposition 18 m p (F ) ≤ m p (F , S)

Proof. This follows immediately form the definitions of m p (F ) and m p (F , S).

We would like to know how the number m p (F , S) is bahaves after a blow-up. Recall
the notations of §10.1. We have a transform S0 of S by the blow-up map which is
uniquely determined by the fact that it is irreducible and

ϕ −1 (S) = ϕ −1 (0) + S0

The curve S0 intersects the exceptional divisor ϕ −1 (0) in a unique point q. Let F1
be the pull-back of the foliation F by the blow-up map ϕ.
Proposition 19 We have

mq (F1 , S0 ) = m p (F , S) − (m p (F ) − 1)

Proof. We have
56 10 Blow up

1 ∗
ω̂ := ϕ ω
xk 0
1
= k [txα(x,tx)dx + β (x,tx)(tdx + xdt)]
x
1
= k [(txα(x,tx) + tβ (x,tx))dx + xβ (x,tx)dt]
x
where k = m p (F). The proposition folows immediately. Note that in the chart U0 , S0
is given by t = 0.

10.4 Sequences of blow-ups

We are going to apply a sequence of blow-ups in a point p = (0, 0) ∈ A2k . We fix the
notations as follows. In the (m − 1)-th step we have a surface Mm−1 with a divisor
m−1 (m−1) (m−1)
D(m−1) = ∪ Pj such that each Pj is isomorphic to the projective line P1k .
j=1
We have also a point pm−1 ∈ D(m−1) . The variety Mm is obtained by a blow-up at
pm−1 . Let ϕm−1 : Mm → Mm−1 be the blow-up map. For the new divisor D(m) we
have
(m) −1
Pm = ϕm−1 (pm−1 ),
(m) −1 (m−1)
Pj = ϕm−1 (Pj ), j = 1, 2, . . . , m − 1.

(m)
Definition 21 We define the weight ρ(Pj ) in the following way. By definition

(m) (m) ( j)
ρ(P1 ) = 1, ρ(Pj ) = ρ(Pj )

and
( j) ( j)
ρ(Pj ) = ∑ ρ(Pi ).
i< j
( j) ( j)
Pi ∩Pj 6=0/

(m)
In other words, the weight of Pm is the sum of the weights of the projective lines
containing pm−1 .
Now, let us consider a foliation F in M0 = A2k . We denote by Fm the pull-back
of F by the the blow-up map ϕ0 ◦ ϕ1 ◦ · · · ◦ ϕm−1 . We will choose the point pm from
the singular set of the foliation Fm .
(m)
In the discussion below we will assume that all Pj are Fm -invariant.
(m) T
Definition 22 Let q ∈ Pj Sing(Fm ). If q is a smooth point of D(m) then we
define
(m)
m∗q (Fm , Pjm ) = mq (Fm , Pj )
10.4 Sequences of blow-ups 57

otherwise
(m)
m∗q (Fm , Pjm ) = mq (Fm , Pj ) − 1.
Proposition 20 We have
(m) (m)
m p (F ) + 1 = ∑ ρ(Pj )m∗q (Fm , Pj ). (10.3)
q∈Sing(Fm )∩D(m)
j=1,...,m

Proof. We prove the proposition by induction on m. For m = 1 we have to show that

(1)
m p (F ) + 1 = ∑ mq (F1 , P1 )
(1)
q∈P1 ∩SingF1

We know that the foliation F1 in U0 is given by

ω̂(x,t) = [(Pk (1,t) + tQk (1,t))dx + xQk (1,t)dt] + xα.

We can assume that Qk (0, 1) 6= 0, or equivalently Qk (1,t) is of degree k = m p (F ).

Therefore, all the singularities of F1 are in the U0 chart and so
(1)
∑ mq (F , P1 ) = deg(Pk (1,t) + tQk (1,t)) = m p (F ) + 1
q∈Sing(F )

Now, assume that (10.3) is true for after m blow-ups. We consider two cases:
(m)
1. pm ∈ Pk is a regular point of D(m) . By definition
(m+1) (m) (m) (m)
ρ(Pm+1 ) = ρ(Pk ), m∗pm (Fm , Pk ) = m pm (Fm , Pk )

(m+1) T (m+1)
Let {q} = Pk Pm+1 . We have

(m) (m) (m+1) (m+1)

ρ(Pk )m pm (Fm , Pk ) = ρ(Pk )[mq (Fm+1 , Pk ) + m pm (Fm ) − 1]+

(m+1) (m+1) (m+1) (m+1)

ρ(Pk )(mq (Fm+1 , Pk ) − 1) + ρ(Pm+1 )( ∑ mr (Fm+1 , Pm+1 )) =
(m+1)
r∈Pm+1
r∈Sing(Fm+1 )

(m+1) (m+1) (m+1) (m+1)

ρ(Pk )m∗q (Fm+1 , Pk ) + ρ(Pm+1 )(mq (Fm+1 , Pm+1 ) − 1)+

(m+1) (m+1)
ρ(Pm+1 )( ∑ mr (Fm+1 , Pm+1 ))
(m+1)
r∈ Pm+1
r∈Sing(Fm+1 )\{q}

(m) (m)
Now, in (10.3) for m we replace ρ(Pk )m pm (Fm , Pk ) with the sum obtained in
the above equality and we get (10.3) for m + 1.
58 10 Blow up

up 4.png

Fig. 10.4 blowing up

up 5.png

Fig. 10.5 blowing up

(m) (m)
2- pm ∈ Pk1 ∩ Pk2 and so it is not a smooth point of D(m) . In this case we will
replace the terms
(m) (m) (m) (m)
ρ(Pk1 )m∗pm (Fm , Pk1 ) + ρ(Pk2 )m∗pm (Fm , Pk2 )

of (10.3) for m with apropriate sums and we will get (10.3) for m + 1. Let {qi } =
(m+1) T (m+1)
Pki Pm+1 , i = 1, 2. We have
10.5 Milnor number 59

(m) (m) (m) (m)

ρ(Pk1 )m∗pm (Fm , Pk1 ) + ρ(Pk2 )m∗pm (Fm , Pk2 ) =

(m+1) (m+1)
ρ(Pk1 )(m∗q1 (Fm+1 , Pk1 ) + m pm (Fm ) − 1)+

(m+1) (m+1)
ρ(Pk2 )(m∗q2 (Fm+1 , Pk2 ) + m pm (Fm ) − 1) =

(m+1) (m+1) (m+1) (m+1)

ρ(Pk1 )m∗q1 (Fm+1 , Pk1 ) + ρ(Pk2 )m∗q2 (Fm+1 , Pk2 )+

(m+1)
ρ(Pm+1 )(m pm (Fm ) − 1).

We have also
(m+1)
m pm (Fm ) − 1 = ∑ mr (Fm+1 , Pm+1 ) − 2
(m+1)
r∈Pm+1
r∈Sing(Fm+1 )
(m+1) (m+1) (m+1)
= ∑ mr (Fm+1 , Pm+1 ) + mq1 (Fm+1 , Pm+1 ) + mq2 (Fm+1 , Pm+1 ) − 2
(m+1)
r∈Pm+1
r∈Sing(Fm+1 )
r6=q1 ,q2
(m+1)
= ∑ m∗r (Fm+1 , Pm+1 ).
(m+1)
r∈Pm+1
r∈Sing(Fm+1 )

10.5 Milnor number

For a holomorphic foliation F (ω), ω = Pdy − Qdx in A2k with an isolated singu-
larity at p = 0 we define he Milnor number

OA2 ,p
k
µ p (F ) := dimk (10.4)
hP, Qi

we also define l p (F ) to be the zero order of ϕ ∗ (ω) along the exceptional divisor
ϕ −1 (0). According to our discussion in §10.3, l p (F ) = m p (F ) if ϕ −1 (0) in F1 -
invariant and l p (F ) = m p (F ) + 1 otherwise.
Theorem 17 If p is a dicritical singular point of F then

µ p (F ) = l p (F )2 + l p (F ) − 1 + ∑ µq (F1 ) (10.5)
q∈ϕ −1 (p)

and if p is a non-dicritical point of F then

60 10 Blow up

µ p (F ) = l p (F )2 − l p (F ) − 1 + ∑ µq (F1 ) (10.6)
q∈ϕ −1 (p)

For a proof see Mattei-Moussu or Soares-Mol

10.6 Seidenberg’s theorem

Let F be a foliation in A2k .

Definition 23 We say that p is a reduced singularity of F (X) if its linear part is not
zero and it is of the form λ1 x ∂∂x + λ2 y ∂∂y , where

1. λ1 6= 0 and λ2 6= 0 and λλ1 6∈ Q+ or

2
2. One of the λi ’s is zero and the other is not.
In this section we prove the following.
Theorem 18 There is finite sequence of blow-ups such that pull-back foliation has
only reduced singularities.

Proof. We first prove that after a sequence of blow-ups all the singularities of the
pull-back foliation have multiplicity m p equal to 1. We use the fact that if l p (resp.
µ p ) equals to 1 then m p equals to 1. If l p (F ) = 1 then m p (F ) = 1 and we are done.
Otherwise, we perform a blow-up at p and use (10.5) and (10.6) and we conlude
that µq (F1 ) < µ p (F ). This means that after a finite number of blow-ups we have
singularities with either l p or µ p equal to 1.
Now, assume that p is a singularity of F (X) with a non-zero linear part. Using
Proposition 6 we can consider only the following cases:
1. y ∂∂x + ...
2. (ax + y) ∂∂x + ay ∂∂y + ... with a 6= 0.
3. ax ∂∂x + by ∂∂y + ... with (a, b) 6= (0, 0)
Chapter 11
Jouanolou foliation

The holomorphic foliation Fd defined in C2 by the 1-form

ω := (yd − xd+1 )dy − (1 − xd y)dx

is called the Jouanolou foliation of degree d. Consider the group

2 +d+1
G := {ε ∈ C | ε d = 1}.

It acts on C2 discontinuously in the following way:

(ε, (x, y)) → (ε d+1 x, εy) ε ∈ G, (x, y) ∈ C2

It has a fixed point p1 = (0, 0) at C2 (and two other fixed points p2 = [0 : 1 : 0], p3 =
[1 : 0 : 0] at infinity). For each ε ∈ G we have ε ∗ (ω) = ε d+1 ω and so G leaves Fd
invariant. We have
Sing(Fd )C = {(ε, ε −d ) | ε ∈ G}
(there is no singularity at infinity) and G acts on Sing(Fd ) transitively. For pictures of Jouanolou foliation see
[MV09]

61
Chapter 12
Fibrations

The most simple foliations are fibrations. These foliations lack dynamics, however,
they enjoy a beautifull topological theory. This is known under the name Picard-
Lefschetz theory. In this chapter we deal with fibrations by curves in P2 , however,
the theory can be developed for fibrations on projective varieties, all of them of
arbitrary dimension.

12.1 Main examples

Fibrations with multiple fibers

Fp
Let us consider the fibration in P2 given by the rational function Gq , where F and G
deg(F)
are two relatively prime irreducible polynomials in an affine chart C2 of P2 , deg(G) =
q
p and g.c.d.(p, q) = 1.

Pull-back fibrations

Product of lines

Tame polynomials

Our main example for Picard-Lefschetz theory is the fibration of tame polynomials
which has been extensively studied in [Mov11], Chapter 6.
Definition 24 We say that a polynomial f of degree d is tame if there are positive
integers α1 and α2 such that the last homogeneous part g of f in the weighted ring
C[x, y], weight(x) = α1 and weight(y) = α2 , has an isolated singularity at the origin.
Note that For α1 = α2 = 1, g has an isolate singularity at the origin if and only if
g = (x − a1 y)(x − a2 y) · · · (x − ad y) and ai ’s are distict.

63
64 12 Fibrations

12.2 Generic conditions

Theorem 19 The projective variety Pn is complete, that is, for all algebraic vari-
eties V , the projection map π : Pn × V → V is closed. This means that any closed
subset W of Pn ×V , π(W ) is a closed subset of V . In particular, any closed subver-
iety of Pn is complete
For a proof of the above theorem see for instance [Mil]. Since V \π(W ) is a Zariski
open set, if we prove that it is non-empty then a generic point of V is not in the image
of π. Sometimes, it is hard to find points in V \π(W ) despite the fact that we are sure
that is is a non-empty Zariski open set. In these sitations, we take an arbitrary point
of x ∈ Pn × V and compute the map induced in the tangent spaces (T Pn × V )x →
(TV )π(x) and prove that it is not surjective. For instance, if dim(W ) < dim(V ) this
is always the case. This also prove that V \π(W ) is non-empty. In some sitations
W = ∪si=1Wi is a union of some closed varieties, and it is easier to find points in
each V \π(Wi ). This implies that V \π(W ) is non-empty and so we do not need to
give explicit examples of its elements.
p
In the space of fibration FGq the following conditions are generic:
1. {F = 0} and {G = 0} are smooth varieties;
2. {F = 0} and {G = 0} intersect each other transversally;
3. The restriction of f to Pn \({F = 0} ∪ {G = 0}) has nondegenerate critical points,
namely p1 , p2 , . . . , pr .
4. The images c1 = f (p1 ), c2 = f (p2 ), . . . , cr = f (pr ) are distincts.
p
For the proof we first take V the parameter space of FGq . The set Wi ⊂ Pn ×V, i = 1, 2
p
is the algebraic closure of the set of ( FGq , p) such that for i = 1 (1) fails, and for i = 2
p
either (2) or (3) fails. Let also W3 ⊂ Pn × Pn ×V the closure of the set of ( FGq , p, q)
such that either (4) fails, that is, f (p) = f (q). In this case we use the fact that Pn ×Pn
is complete.
p
It remains to find an explicit example of FGq which satisfies the above generic
conditions. This might get hard. That is why we have introduced many Wi ’s. Finding
a point in V \π(W1 ) is trivial. For instance, we take F = 0 and G = 0 Fermat varieties.
An example of a point in both V \π(Wi ) i = 2, 3 is as follows. We take F and G a
product of lines in general position. This example satisfies the condition (2), (3) or
(4), see Exercise (12.1).

12.1. Show that the rational function

Fp (x(x − 1)(x − 2) · · · (x − a)) p
= , ap = bq
Gq (y(y − 1)(y − 2) · · · (y − b))q

has (a − 1)(b − 1) non-degenerated critical points with distinct images. Show also
that
f = xd+1 + yd+1 − (d + 1)x − (d + 1)y : C2 → C.
has d 2 non-degenerated critical points with distinct images.
12.3 Ehresmann’s fibration theorem 65

12.3 Ehresmann’s fibration theorem

Throughout the text when we do not write the coefficients used in the homology
we mean homology with coefficients in Z. Let f be a rational function on a smooth
algebraic variety X of dimension two. The whole discussion is valid for meromor-
phic functions on complex manifolds, however, we prefer to work in the algebraic
framework. The indeterminacy set R of f contains the points of X in which f has
the form 00 .
Theorem 20 There is a smooth algebraic variety X̃, regular maps f˜ : X̃ → P1 and
π : X̃ → X such that
π /
X̃ X (12.1)
f
f˜  
P1
commutes, that is, f ◦ π = f˜.
Proof. We use desingularization theorem for the holomorphic foliation F = F (d f )
in X. The indeterminacy points of f are singular points of F . We perform a se-
quence of blow-ups at R and obtain π : X̃ → X such that the singularities of
F˜ := π −1 (F ) over R are reduced singularities in the sense of Definition 6.1. We
have rational function f˜ : X̃ → P1 such that the diagram (12.1) is commutative and
we have to prove that f˜ is regular, that is, it has no indeterminacy points. We have
F˜ := F (d f˜) and if such a point exists then we have a singularity of F˜ with infin-
intely many separatrix, and so it is not reduced.
The indeterminacy set R is discrete and the following holomorphic function is
well-defined:
f : X − R → P1
We use the following notations

LK = f −1 (K), XK = LK , K ⊂ P1

For any point c ∈ P1 by Lc and Xc we mean the set L{c} and X{c} , respectively.
Throughout the text by a compact f -fiber we mean Xt and by a f -fiber only we
mean Lt .
Corollary 3 There exists a finite subset C = {c1 , c2 , . . . , cr } of P1 such that f fibers
X − R locally trivially over B = P1 −C i.e., for every point b ∈ B there is a neigh-
borhood U of b and a C∞ -diffeomorphism φ : U × f −1 (c) → f −1 (U) such that

DD

commutes, that is, f ◦ φ = π1 = the first projection

Proof. The main ingredient of the proof is Ehresmann’s Fibration Theorem. we use
Theorem 20 and we have a commutative diagram (12.1). The surface X̃ is compact
66 12 Fibrations

Fig. 12.1 Atypical fibers due to their behaviour at infinity

and f˜ is regular, therefore f˜ is is a proper map. Let P1 , P2 , . . . , Ps , all isomorphic to

P1 , be the set of all blow-up divisors such that f˜ restricted to Pi is not a constant
map. The set C is the union of the crtical values of f˜ and f˜|Pi , i = 1, 2, . . . , s.
Let f be the tame polynomial in §12.1.
Proposition 21 Let C := {c1 , c2 , . . . , cr } be the set of critical values of f . Then f is
a C∞ fibration over C\C
Proof. A proof can be found in [Mov11]. Theorem 6.1. We have to verify that there
is no atypical fibers due to their behaviour at the indeterminacy points. If f is a tame
polynomial with weight(x) = α1 and weight(y) = α2 then the same is true replacing
αi with (α α,αi ) , and so, we can assume that (α1 , α2 ) = 1.
1 2

In this case we only need to do just one blow-up at each indeterminacy points.

12.4 Monodromy and vanishing cycles

For this section we use the notation of the book [Mov16].

Theorem 21 Suppose that H1 (X − X∞ , Q) = 0. Then a distinguished set of vanish-
ing 1-cycles related to the critical points in the set C\{∞} = {c1 , c2 , . . . , cr } gener-
ates H1 (Lb , Q).
Note that in the above theorem ∞ can be a critical value of f .
Definition 25 The cycle δ in a regular fiber Lb is called simple if the action of
π1 (B, b) on δ generates H1 (Lb , Q).
Note that in the above definition we have considered the homology group with ra-
tional coefficients. Of course, not all cycles are simple. For instance if the mero-
morphic function in a local coordinate (x, y) around q ∈ R has the form xy , then
the cycle around q in each leaf has this property that it is fixed under the action of
monodromy, therefore it cannot be simple.
12.5 Exercises 67

12.5 Exercises
Chapter 13
Generic conditions

69
Chapter 14
Iterated Integrals

71
72 14 Iterated Integrals

References

Arn80. V. Arnol0 d. Chapitres supplémentaires de la théorie des équations différentielles ordi-

naires. “Mir”, Moscow, 1980. Translated from the Russian by Djilali Embarek.
Bru00. Marco Brunella. Birational geometry of foliations. Monografı́as de Matemática. Insti-
tuto de Matemática Pura e Aplicada (IMPA), Rio de Janeiro, 2000. Available electron-
ically at http://www.impa.br/Publicacoes/Monografias/Abstracts/brunella.ps.
CLNS88. C. Camacho, A. Lins Neto, and P. Sad. Minimal sets of foliations on complex projective
spaces. Inst. Hautes Études Sci. Publ. Math., (68):187–203 (1989), 1988.
CS87. César Camacho and Paulo Sad. Pontos singulares de equações diferenciais analı́ticas.
16So Colóquio Brasileiro de Matemática. [16th Brazilian Mathematics Colloquium].
Instituto de Matemática Pura e Aplicada (IMPA), Rio de Janeiro, 1987.
Ily02. Yu. Ilyashenko. Centennial history of Hilbert’s 16th problem. Bull. Amer. Math. Soc.
(N.S.), 39(3):301–354 (electronic), 2002.
IY08. Yulij Ilyashenko and Sergei Yakovenko. Lectures on analytic differential equations,
volume 86 of Graduate Studies in Mathematics. American Mathematical Society, Prov-
idence, RI, 2008.
LNS. Alcides Lins Neto and Bruno Scárdua. Introdução á Teoria das Folhações Algébricas
Complexas. Available online at IMPA’s website.
Mil. James Milne. Algebraic Geometry. Lecture notes available at www.jmilne.org.
Mov11. Hossein Movasati. Multiple Integrals and Modular Differential Equations. 28th Brazil-
ian Mathematics Colloquium. Instituto de Matemática Pura e Aplicada, IMPA, 2011.
Mov16. Hossein Movasati. A course in Hodge theory:
with emphasis on multiple integrals. Available at:
http://w3.impa.br/∼hossein/myarticles/hodgetheory.pdf .
2016.
MV09. Hossein Movasati and Evilson Vieira. Projective limit cycles. Moscow Mathematical
Journal, 9(4):855–866, 2009.
Tat68. John Tate. Residues of differentials on curves. Ann. Sci. École Norm. Sup. (4), 1:149–
159, 1968.