Topology of Polynomials and Low Dimensional Algebraic

Geometry

Enrique Artal Bartolo

 Contents

Lecture 1. Germs of curves and Puiseux expansions 5

1.1. Germs of curves 5
1.2. Some classical invariants in singularity theory 7
1.3. Unramified coverings 8
1.4. Picard-Lefschetz 9
1.5. Puiseux expansions 12

Lecture 2. Braid monodromy of germs 15

2.1. Newton construction 15
2.2. Braid group 17
2.3. Local braid monodromy 19

Lecture 3. Global braid monodromy 25

3.1. Applications to the local case. 25
3.2. Braid monodromy of affine plane curves 26
3.3. Zariski-van Kampen Theorem 28

Lecture 4. Topological applications 31

4.1. Puiseux monodromy and topology 31
4.2. Discriminant of polynomial mappings 32

3
 LECTURE 1

Germs of curves and Puiseux expansions

The main object of study in this series of lectures is the topology of polynomials f : C2 → C.
The strategy to study these polynomials have several steps:
• Find the critical points of f and study the behaviour of f around them.
• Find the critical values at infinity and study f around them.
• Study the embedded topology of the fibers.
• Study the global behaviour of f .

1.1. Germs of curves

Example 1.1.1. Let us consider the polynomials f, g : C2 → C given by f (x, y) := x2 − y 2

and g(x, y) := x2 − y 2 + x3 . They have different global properties but near the origin they share
the same analytical properties: note that g(x, y) = (xu(x) − y)(x(u(x) + y)) where u(x) is a local
√
determination of 1 + x around 0.
There exist small neighbourhoods U, V of 0 in C2 such that ψ : U → V , ψ(x, y) := (xu(x), y),
is a well-defined analytic isomorphism and g|U = f|V ◦ ψ.

Let f : U → C, g : V → C two analytical mappings such that U, V are neighbourhoods of

0 in C2 . We say that f ∼ g if there exists W ⊂ U ∩ V neighbourhood of 0 in C2 such that
f|W = g|W . The equivalence class of such an f is called the germ of f at 0. The study of such
germs is the first ingredient in the comprehension of the global topology of polynomials.
The usual operations respect the equivalence relation and the germs form a C-local algebra
O2 := O(C2 ,0) which is naturally isomorphic with the ring of convergent power series C{x, y}.
The maximal ideal M2 is the set of germs of functions vanishing at 0. In the same way we define
the ring O1 := O(C,0) ∼= C{x}. The order of a non-zero element f of O2 (or O1 ) is the minimal
n ≥ 0 such that f ∈ Mjn (j = 1, 2).

Example 1.1.2. The maximal ideal M1 is generated by x. The only proper ideals of O1 are
M1m ,m ∈ N, i.e., O1 is principal (and hence neotherian, factorial).

Let us fix a germ f : (C2 , 0) → C. We will usually identify f with a representative for
a suitable neighbourhood V of 0 in C2 . After a change of coordinates, we may suppose that
f (0, y) 6≡ 0 and is of order n.

Definition 1.1.3. A non-zero element of O2 is a y-regular of order n if f (0, y) 6≡ 0 and of

order n.
5
6 1. GERMS OF CURVES AND PUISEUX EXPANSIONS

Using the Principle of Analytic Continuation, we can fix ε > 0 such that f (0, y) 6= 0 if
0 < |y| ≤ ε and {0} × ∆ ¯ ε ⊂ V , where ∆ε (resp. ∆
¯ ε ) is the open (resp. closed) disk of radius ε
in C centered at 0.
Using standard compactness arguments, ∃δ > 0 such that f (x, y) 6= 0 if |y| = ε, x ∈ ∆ ¯ δ and
∆¯δ × ∆
¯ ε ⊂ V . Let us consider the mapping
Z ∂f (x0 ,y)
¯ δ → C, Vf (x0 ) := 1
Vf : ∆
∂y
dy.
2iπ |y|=ε f (x0 , y)
This mapping is well-defined and continuous by the above construction.
Let us recall the following classical theorem in complex analysis.

Theorem 1.1.4. Let f be a homolomorphic function in a neighbourhood of ∆ ¯ ε , ε > 0 such

that f does not vanish on its boundary. Let us suppose that z1 , . . . , zk are the zeroes of f in ∆ε ,
with multiplicities n1 , . . . , nk . Then,
k
f 0 (t)dt X
Z
1
= nj .
2iπ |t|=ε f (t) j=1

Moreover, if g is also holomorphic in a neighbourhood of ∆ ¯ ε , then

k
g(t)f 0 (t)dt X
Z
1
= nj g(zj ).
2iπ |t|=ε f (t) j=1

Hence Vf (∆ ¯ δ ) ⊂ Z, i.e. it is a constant map and its value equals Vf (0) = n. Given x ∈ ∆
¯ δ,
n = Vf (x) is the number of zeroes of y 7→ f (x, y) in ∆ε (counted with multiplicities); we call
them y1 (x), . . . , yn (x) (we do not bother about the ordering). Let us consider now
Z ∂f (x0 ,y)
¯ δ → C, Vf (x0 ) := 1
Vf,m : ∆ ym
∂y
dy.
2iπ |y|=ε f (x0 , y)
¯ δ and analytic on ∆δ . Moreover
These mappings are well-defined, continuous on ∆
n
X
Vf,m (x) = yj (x)m .
j=1

Remark 1.1.5. Given j = 1, . . . , n, let σj : ∆ ¯ δ → C the j th elementary symmetric function

in y1 , . . . , yn . Recall that σj is the sum of all the possible products of j different terms in
{y1 , . . . , yn }. Following standard invariant theory, σj is a polynomial function (with rational
¯ δ and analytic on ∆δ .
coefficients) of Vf,1 , . . . , Vf,j , and, in particular, it is continuous on ∆

Let us consider f˜ : ∆
¯δ × ∆ ¯ ε → C given by
n
X n
Y
f˜(x, y) := y n + (−1)j σj (x)y n−j = (y − yj (x)).
j=1 j=1

f
which is holomorphic in Wδ,ε := ∆δ × ∆ε . By construction the quotient u := is well-defined,
f˜
continuous on W̄δ,ε , holomorphic on Wδ,ε and vanishes nowhere. Note that σj (0) =, ∀j =
1, . . . , n.
 1.2. SOME CLASSICAL INVARIANTS IN SINGULARITY THEORY 7

Definition 1.1.6. A non-zero element of O1 [y] is a Weierstrass polynomial if it is monic

and the non-leading coefficients are in M1 .

The above arguments prove the following classic result.

Theorem 1.1.7 (Weierstrass Preparation Theorem). Let f ∈ O2 regular of order n. There

exist a unique Weierstrass polynomial f˜ (of degree n) and a unique unit u ∈ O2 \ M2 such that
f = uf˜.

Corollary 1.1.8 (Implicit Function Theorem). Let f ∈ O2 such that f (0) = 0 and ∂f
∂y (0) 6=
0. Then, there exists g ∈ M1 such and neighbourhoods U, V of 0 in C and C such that
2

Z := {(x, y) ∈ U | f (x, y) = 0} = {(x, g(x)) | x ∈ V };

in particular, Z and V are analytically isomorphic.

Proof. The conditions f (0) = 0 and ∂f∂y (0) 6= 0 imply that f is y-regular of order 1.
Then f (x, y) = u(x, y)(y − g(x)). Choose a neighbourhood U where u does not vanish and
f (x, y) = 0 ⇐⇒ y = g(x). 

Remark 1.1.9. The Weierstrass Preparation and Division Theorems (the second one is
proved with similar techniques) allow to prove that O2 is noetherian and factorial. Moreover, if
p ∈ O2 is a Weierstrass polynomial, its irreducible decompositions in O2 and O1 [y] coincide and
all the factors are Weierstrass polynomials.

Passing to germs, f˜ ∈ O1 [y] and we can consider its decomposition f˜ = f1r1 · . . . · flrl , where
fj are irreducible Weierstrass polynomials.

1.2. Some classical invariants in singularity theory

There are several analytic-topological invariants associated to f .

Definition 1.2.1. The Milnor number of f is µ(f ) := dimC O2 /Jf , where Jf is the
Jacobian ideal of f , i.e., the one generated by ∂f ∂f
∂x and ∂y .

Remark 1.2.2. The following fact is classical: r1 = · · · = rl = 1 if and only if µ(f ) < ∞.
We assume for the rest of the section the finiteness of µ(f ).

Proposition 1.2.3. There exists 0 < η  δ  ε such that if 0 < |t| ≤ η then Ft :=
−1
f (t) ∩ (W̄δ,ε ) verifies χ(Ft ) = 1 − µ(f ).

Let us denote
• Bη,δ,ε := f −1 (∆¯ η ) ∩ W̄δ,ε ,
∗ −1 ¯ ∗
• Bη,δ,ε := f (∆η ) ∩ W̄δ,ε ,
• Sη,δ,ε := f −1 (∆¯ η ) ∩ ∂ W̄δ,ε = f −1 (∆¯ η ) ∩ (∂ ∆
¯δ × ∆¯ ε ) and
• ¯ ∗η ) ∩ W̄δ,ε = f −1 (∆
S∗η,δ,ε := f −1 (∆ ¯ ∗η ) ∩ (∂ ∆
¯δ × ∆
¯ ε ).

Proposition 1.2.4.
8 1. GERMS OF CURVES AND PUISEUX EXPANSIONS

(1) ¯ ∗η is a locally trivial fibration.

The restriction f| : B∗η,δ,ε → ∆
(2) The restriction f| : Sη,δ,ε → ∆¯ η is a trivial fibration.
(3) Both fibrations are compatible on S∗η,δ,ε .
(4) Bη,δ,ε is homeomorphic to a 4-dimensional ball.
Moreover, these fibrations do not depend on the choice of suitable η, δ, ε.

The proof of Proposition 1.2.4(1) is based on the fact that it is a proper submersion. For
(2), we use also that any fibration over a contractible space is trivial.
A first consequence of (2) is that, up to isotopy, there is a unique homeomorphism Φ :
∆η × ∂Fη → Sη,δ,ε such that f| ◦ Φ is the first projection. The other fibration is determined
up to isotopy by the so-called geometric monodromy φ : Fη → Fη , whose construction will be
explained in §1.3; the compatibility implies that φ is the identity over ∂Fη .
¯δ ×∆
Note that ∂ W̄δ,ε is a 3-dimensional sphere obtained as the union of two solid tori ∂ ∆ ¯ε
and ∆¯ δ × ∂∆¯ ε.
We fix some notation:
• Kf,δ,ε := ∂ W̄δ,ε ∩ f −1 (0).
• B4ε := {(x, y) ∈ C2 | |x|2 + |y|2 ≤ ε2 }
• S3ε := {(x, y) ∈ C2 | |x|2 + |y|2 = ε2 } which is also a 3-dimensional sphere.
• Kf,ε := S3ε ∩ f −1 (0).

Theorem 1.2.5 (Milnor). The homeomorphism type of (∂ W̄δ,ε , Kf,δ,ε ) and (S3 , Kf,ε ) coin-
cide for δ and ε small enough. Moreover the homeomorphism types of their cones coincide with
the ones of (W̄δ,ε , f −1 (0)) and (B4ε , f −1 (0)).

Definition 1.2.6. The link of f is the homeomorphism type of (∂ W̄δ,ε , Kf,δ,ε ).

Remark 1.2.7. All these invariants coincide for f and f˜.

1.3. Unramified coverings

We recall some definitions and results on coverings which will be used in these lectures. All
topological spaces will be locally path-connected.

Definition 1.3.1. A continuous mapping π : X → Y is an unramified covering if ∀y ∈ Y ,

there exists an open neighbourhood V of y such that π −1 (Vy ) is a disjoint union i∈Iy Ui of
`

open sets such that π| : Ui → Vy is a homeomorphism ∀i ∈ Iy .

There are some easy criteria to deduce if a mapping is a covering.

Lemma 1.3.2. Let π : X → Y be a surjective local homeomorphism such that ∀y ∈ Y ,

−1
#π (y) is finite and constant. Then, π is a covering.

There is a close relationship between covering theory and fundamental groups via the lifting
path and homotopy properties. Let us recall the definition of classical monodromy of a covering.
Let us suppose that we have a covering π : X → Y where Y connected. Fix y ∈ Y and let
G := π1 (Y ; y); let F := π −1 (y) and let ΣF be the group of permutations of F (with exponent
 1.4. PICARD-LEFSCHETZ 9

notation). The main invariant of π is the monodromy ρ : G → ΣF : for each x ∈ F and each
loop γ based at y there is a unique lifting of the loop starting at x and xρ(γ) is the ending point
of this lifting. The following properties are important:
• If x ∈ F , then π1 (X; x) is identified by π∗ with {g ∈ G | xρ(g) = x}.
• X is connected if and only if ρ is transitive, i.e., if ∀x1 , x2 ∈ F , there exists g ∈ G such
ρ(g)
that x1 = x2 .
• Moreover, each connected component of X is associated with the orbits of F by the
action of G and the restriction of π to each connected component is a covering.
There are two main theorems associated with this construction.

Theorem 1.3.3. Let π : X → Y be a covering, X, Y . Let h : Z → Y be a continuous

mapping, Z connected. Fix x ∈ X and z ∈ Z such that π(x) = h(z) =: y ∈ Y . Then,
there exists h̃ : Z → X continuous mapping such that π ◦ h̃ = h and h̃(z) = x if and only if
h∗ (π1 (Z; z)) ⊂ π∗ (π1 (X; x)).
It it exists, h̃ is unique.

Theorem 1.3.4. Let Y be a connected space, y ∈ Y . For any subgroup H of π1 (Y ; y) there

exists a covering π : X → Y , X connected, and x ∈ π −1 (y) such that π∗ (π1 (X; x)) = H. The
covering π is essentially unique.

Example 1.3.5. Let us recall that π1 (C∗ ; 1) ∼ = Z, generated by the class of γ : [0, 1] → C∗
where γ(t) := exp(2iπt). The subgroups of Z are {0} and {mZ}m∈N . The associated coverings
are C → C∗ , t 7→ exp(2iπt) and C∗ → C∗ , t 7→ tm .
If we replace C∗ by ∆¯ ∗η , η > 0, we have the same result for Hη := {z ∈ C | 2π=z ≥ log η} →
¯ ∗ and ∆
∆ ¯∗ 1 → ∆ ¯ ∗η .
η
ηm
In particular, if we have a continuous mapping h : X → C∗ , X connected, let h∗ : π1 (X; x) →
π1 (C∗ ; h(x)) ≡ Z. Then, h has an mth -root if and only if h∗ (π1 (X; x)) ⊂ mZ.

Remark 1.3.6. We can explain the construction of the monodromy of the fibration of
¯ ∗η and the covering
Proposition 1.2.4. We consider the locally trivial fibration f| : B∗η,δ,ε → ∆
Hη → ∆¯ ∗η . Let us define
e η,δ,ε := {((x, y), t) ∈ B∗ × Hη | f (x, y) = exp(2iπt)}.
B η,δ,ε

The restriction of the second projection is a locally trivial fibration with fiber Fη . Since Hη is
contractible, this fibration is a trivial one and there is a homeomorphism Φ̃ : Hη × Fη → B e η,δ,ε
which identifies the fibrations.
The mapping B e η,δ,ε → B
e η,δ,ε given by ((x, y), t) 7→ ((x, y), t + 1) is an automorphism of the
projection and induces via Φ̃ the geometric monodromy φ : Fη → Fη as follows.
Let (x, y) ∈ Fη and let tη := log η
2iπ ∈ iR. Hence ((x, y), tη ) ∈ Bη,δ,ε and it is also the case for
e
−1
((x, y), tη + 1). Then, Φ̃ ((x, y), tη + 1)) = (tη + 1, φ(x, y)).

1.4. Picard-Lefschetz

Let f (x, y) := x2 + y 2 ; we can choose ε = 10, δ = 2 and η = 1.

10 1. GERMS OF CURVES AND PUISEUX EXPANSIONS

• W̄δ,ε ∩ f −1 (0) = {(x, ix) | |x| ≤ 2} ∩ {(x, −ix) | |x| ≤ 2}, i.e., a union of two disks
intersecting at their centers.
• F1 = {(x, y) ∈ W̄δ,ε | x2 + y 2 = 1}. This condition can be given by:

<x2 + <y 2 =1 + =x2 + =y 2

<x=x + <y=y =0.

Let us recall that the tangent space T S1 is the space of points (u, v, w, t) ∈ R4 such
that:

u2 + v 2 =1
uw + vt =0.

Let us consider the homeomorphism C2 → R2 × {(w, t) ∈ R2 | w2 + t2 < 1} given by

1
(x, y) 7→ p (<x, <y, =x, =y).
1 + =x2 + =y 2
We deduce that F1 is homeomorphic to a cylinder, i.e., to a closed neighbourhood of
S1 in T S1 which is a disk fibration over S1 .
• Sη,δ,ε = {(x, y) ∈ C2 | |x2 + y 2 | ≤ 1, |x| = 2}.
¯ 1 , and then z lives in the
Let us denote z = y 2 and ω := x2 + z. Note that ω ∈ ∆
2
closed disk of radius 1 centered at −x .

Lemma 1.4.1. The function

√ √
¯ 1, x ∈
ρ̂ : T := {(x, ω) ∈ C2 | ω ∈ ∆ / [− ω, ω]} → C∗ , x 7→ ω − x2 ,

has two square roots.

Proof. Let G := π1 (T ; (2, 1)) and H := π1 (C∗ ; −3). It is easily seen that H ∼
=Z
at is generated by γ : t 7→ −3 exp(2iπt). In the same way, G ∼ = Z and is generated by
the path β : t 7→ (2 exp(2iπt), 1).
It is not hard to check that ρ̂∗ ([β]) = [γ 2 ]. Since ρ̂(G) ⊂ 2Z, we conclude the
existence of the square roots. 

Let ρ the square root of ρ̂ such that ρ(1, 0) = i; note that ρ(x, 0) = ix. Hence,

¯ 1 , |x| = 2} ∪ {(x, −ρ(x, ω)) ∈ C2 | ω ∈ ∆

Sη,δ,ε = {(x, ρ(x, ω)) ∈ C2 | ω ∈ ∆ ¯ 1 , |x| = 2}.

In particular,

∂F1 = {(x, ρ(x, 1)) ∈ C2 | |x| = 2} ∪ {(x, −ρ(x, 1)) ∈ C2 | |x| = 2}

and we obtain a homeomorphism Sη,δ,ε ∼ =∆¯ η × ∂F1 .

• Hη = {t ∈ C | =t ≥ 0}.
¯ ∗, x ∈ ∆
• B∗η,δ,ε := {(x, y) ∈ C2 | x2 + y 2 ∈ ∆ ¯ 2 }.
1
 1.4. PICARD-LEFSCHETZ 11

¯ 2 , x2 + y 2 = exp(2iπt)}.
e η,δ,ε := {(x, y, t) ∈ C3 | t ≥ 0, x ∈ ∆
• B
There is a trivial way of identifying Hη × F1 with B e η,δ,ε :

(t, (x, y)) 7→ (x exp(iπt), y exp(iπt), t)

which produces a monodromy F1 → F1 given by (x, y) 7→ (−x, −y), but unfortunately
it does not respect the identifications in the boundary. Following the ideas of Arnol’d,
Gussein-Zade and Varchenko, we can see the Fw as the Riemann surface of the function
√
w − x2 .

¯ 2 × {1})
ρ(∆ −1 1 −1 1 ¯ 2 × {1})
−ρ(∆

√
Figure 1.1. Riemann surface of 1 − x2

With this language, the natural trivialization is:

(t, (x, ±ρ(x, 1))) 7→ (t, (xt , ±ρ(xt , exp(2iπt))), xt := exp(iπt)x.
The trivialization in the boundary is:
(t, (x, ±ρ(x, 1))) 7→ (t, (x, ±ρ(x, exp(2iπt))).
¯ 2 × {1})
ρ(∆ ¯ 2 × {exp(i π )})
ρ(∆ ¯ 2 × {i})
ρ(∆
4

Figure 1.2. Monodromy

Let ξ : [0, 2] × (0, 1] → R a differentiable function such that ξ(•, v) is decreasing

∀v ≥ 0, ξ(u, v) = 1 if u ∈ [0, v] and ξ(2v, v) = 1. The good trivialization is:
(t, (x, ±ρ(x, 1))) 7→ (t, (xt,ξ , ±ρ(t,ξ , exp(2iπt))) , xt,ξ := exp (iξ (|x|, exp(−2π|t|)) πt) x.
12 1. GERMS OF CURVES AND PUISEUX EXPANSIONS

The monodromy of F1 is given by

(x, ±ρ(x, 1)) 7→ (xξ(|x|, 1), ±ρ(xξ(|x|, 1), 1))

1.5. Puiseux expansions

In this section we will fix the variables x, y. Fix g ∈ O2 an irreducible Weierstrass polynomial
of degree m and consider ε and δ as above. Let discy (g) ∈ O1 be the discriminant of f˜ with
respect to y, which has a representative ∆ ¯ δ → C.

Lemma 1.5.1. discy (g) 6≡ 0.

Proof. Since O1 is factorial, if discy (g) ≡ 0, then g and ∂f

∂y have a common factor, which
is impossible since g is irreducible and degy ∂f
∂y < degy g. 

Remark 1.5.2. As a consequence, we can suppose that discy (g)(x0 ) 6= 0 if 0 < |x0 | ≤ δ, i.e.,
g(x0 , y) ∈ C[y] has m distinct roots.

¯ δ onto the first

Let Z := {(x, y) ∈ W̄ | g(x, y) = 0}. We consider the projection πx : Z → ∆
variable.

Notation 1.5.3. If X is a subset of C or C2 containing the origin, X ∗ := X \ {0}.

¯ ∗ is an unramified covering of degree m.

Lemma 1.5.4. The restriction πx| : Z ∗ → ∆ δ

Proof. We use Lemma 1.3.2. The statement about cardinality comes from Remark 1.5.2.
The assumption about local homeomorphism comes from the Implicit Function Theorem. 

Applying Example 1.3.5, we obtain that Z ∗ is a disjoint union of punctured disks. We

are going to prove that Z ∗ is connected. Let Z1∗ be a connected component of Z ∗ and let
m1 be the degree of π|Z1∗ . Given x ∈ ∆ ¯ ∗ , we can reorder the roots of g(x, y) such that
δ
∗
(x, y1 (x)), . . . , (x, ym1 (x)) ∈ Z1 .

¯δ
Lemma 1.5.5. The symmetric functions of y1 , . . . , ym1 define continuous functions on ∆
which are holomorphic on ∆δ .

Proof. First, we can proof the statement for ∆ ¯ ∗ . Given x ∈ ∆ ¯ ∗ and fixing yj (x), j =
δ δ
1, . . . , m1 , the Implicit Function Theorem allows us to find a local homeomorphism (analytic
isomorphism if x ∈ ∆ ¯ δ ) of a connected neighbourhood of x in ∆
¯ ∗ onto a neighbourhood of yj (x)
δ
in Z ∗ which by connectedness is contained in Z1∗ . This local homeomorphism may not respect
the indices and this is why the statement is reduced to the symmetric functions.
Since all the roots tend to zero, these functions are locally bounded at the origin, and by
Riemann’s removal singularity Theorem, the symmetric functions extend to ∆ ¯ δ. 

Proposition 1.5.6. Z1∗ = Z ∗ .

 1.5. PUISEUX EXPANSIONS 13

Proof. Let σ1,1 , . . . , σ1,m1 be the elementary symmetric functions of y1 , . . . , ym1 . We have
proved that
m1
X m1
Y
m1 j m1 −j
h(x, y) := y + (−1) σ1,j (x)y = (y − yj (x)).
j=1 j=1
is a Weierstrass polynomial and it divides g; since g is irreducible hence g = h, m = m1 and
Z1∗ = Z ∗ . 

Let us consider the following diagram:

Z∗
↓π
¯∗ 1
∆ −→ ¯∗
∆ δ
δm
t 7−→ tm .
This diagram induces a corresponding one for fundamental groups:
π1 (Z ∗ ; (δ, y1 (δ)))
↓ π∗
Z∼ ¯ ∗ 1 ; δ m1 ) −→
= π1 (∆ π1 (∆∗δ ; δ) ∼
¯ =Z
δm
1 7−→ m.
The subgroup corresponding to the vertical arrow is mZ and we have proved the following result.

Theorem 1.5.7 (Puiseux expansions). There exists h ∈ O1 admiting a representative in a

neighbourhood of ∆ ¯ 1 such that Z = {(tm , h(t)) | t ∈ ∆
¯ 1 }. The set of germs satisfying this
δm δm
property is {h(ζt) | ζ m = 1}.

Let us consider the expansion of h:

∞
X
h(t) = ak tk .
k=1

Lemma 1.5.8. gcd{k | ak 6= 0} = 1.

Notation 1.5.9. Since the parametrization of Z provides the equality x = tm , it is classical

1 P∞ k
to write down h(x m ) = k=1 ak x m .

Let us consider the following construction:

• If m = 1 the construction does not start.
• If m > 1, using Lemma 1.5.8, there exists n1 := min{k | ak 6= 0, m - k}. The rational nm1
is the first Puiseux exponent of g with respect to x. Let nm1 =: pq11 , where gcd(p1 , q1 ) = 1.
1
If we see h as a series in x m , it is the first term with non-integer exponent.
• If p1 = m, we are done.
• If not, let n2 := min{k | ak 6= 0, pm1 - k}. Then, nm2 is the second Puiseux exponent of g
with respect to x. Let nm2 =: p1q2p2 , where gcd(p2 , q2 ) = 1. It is the first term such that
its exponent is not in p11 Z.
• If p1 p2 = m, we are done.
14 1. GERMS OF CURVES AND PUISEUX EXPANSIONS

n or
• If not, we continue the procedure constructing a sequence { nmi }ri=1 = Qi qi p ,
j=1 j i=1
Qr
where gcd(pi , qi ) = 1, pi > 1, j=1 pj = m.
n or
Definition 1.5.10. The sequence { nmi }ri=1 = Qi qi p is the set of Puiseux exponents
j=1 j
i=1
of g.

Let g1 , g2 be two irreducible Weierstrass polynomials of degrees m1 , m2 . Let h1 , h2 ∈ O1 be

their Puiseux expansions. For convenience we denote them by
X
hj := ak,j tkmi .
k∈ m1 N
i

Definition 1.5.11. The coincidence exponent of g1 , g2 is

min{k | ∀ζ1 , ζ2 s.t ζ1m1 = 1, ζ2m2 = 1 one has ζ1k ak,1 6= ζ2k ak,2 }.

Remark 1.5.12. Let f = f1 · . . . · fr be a reduced Weierstrass polynomial, fj of degree mj .

1
For each j we consider a Puiseux expansion hj (x mj . Let us denote
r
1 1 1
mj
[
{h̃1 (x m ), . . . , h̃1 (x m )} = {hj (ζj x mj ) | ζj = 1},
j=1

where m is the least common multiple of m1 , . . . , mr and n = degy f . Then, we have that
Qn 1 1 1
f (x, y) = j=1 (y − h̃j (x m )) and the set of orders of h̃j (x m ) − h̃k (x m ), j 6= k, is the set of
Puiseux exponents of the roots and the set of coincide exponents.
 LECTURE 2

Braid monodromy of germs

2.1. Newton construction

The computation of the Weierstrass polynomial f˜ of f , the factorizacion of f˜ and the con-
struction of Puiseux exponents for its irreducible components in §1.5 are not constructible.
There is method which comes from Newton to obtain them. Let us fix f ∈ O2 , where we
consider coordinates such that f is y-regular of order n,
X
f (x, y) := an xnx y ny .
n:=(nx ,ny )∈Z2≥0

{n + R2≥0 | an 6= 0} in R2≥0 .
S
Let ∆f be the convex closure of

Definition 2.1.1. The Newton polygon Γf of f is the union of the compact faces of ∂∆f .

x + 2y = 4

Figure 2.1. Newton polygon of y 2 + x2 y + x4 + x3 y 3

Let γ be and edge of Γf . We have:

γ ⊂ {(u, v) | qu + pv = N }, gcd(p, q) = 1, p, q, N ∈ N.

We consider a weighted-degree in O2 where deg x = q and deg y = p; the order of f with respect
to this degree equals N and the minimal form is fγ := n∈γ an xnx y ny , which is a weighted
P

homogeneous polynomial and can be decomposed in a unique way as

rγ
Y
fγ (x, y) = xNx y Ny (y q − wγ,j xp )mγ,j ,
j=1
∗
where Nx , Ny ≥ 0, rγ , mγ,j ∈ N and wγ,j ∈ C . The following properties hold:
15
16 2. BRAID MONODROMY OF GERMS

Prγ
• Nx q + Ny p + pq j=1 mγ,j = N .
• rγ > 0.
Prγ
• Ny + q j=1 mγ,j ≤ n.

Example 2.1.2. In the germ of Figure 2.1 we have only one edge γ and in this case rγ = 2.

Definition 2.1.3. We say that x, y are good coordinates for f if no edge γ of Γf satisfies
q = 1 and rγ = 1.

Proposition 2.1.4. There is a change of variables of the form y 7→ y + α(x), α ∈ O1 , such

that x, y are good coordinates for f .

Fix j = 1, . . . , rγ . To the pair (γ, wγ,j ) we associate the following change of variables:
x = xq1 , y = xp1 (y1 + w̃γ,j ),
where w̃γ,j is a q th -root of wγ,j .

Definition 2.1.5. The above change is the Newton transformation associated to γ and wγ,j .

Remark 2.1.6. It is not a change of variables since the mapping defined is not injective. It
is generically q : 1.

Note that f (x, y) = xN 1 fγ,wγ,j (x1 , y1 ), where

 
Y
fγ,wγ,j (x1 , y1 ) =(y1 + w̃γ,j )Ny  ((y1 + w̃γ,j )q − wγ,k )mγ,k  ((y1 + w̃γ,j )q − wγ,j )mγ,j + · · · =
k6=j
m
=Cy1 γ,j + ...
The new series fγ,wγ,j (x1 , y1 ) ∈ C{x1 , y1 } satisfies that the order of fγ,wγ,j (0, y1 ) equals mγ,j .

Lemma 2.1.7. mγ,j < n.

Proof. It is clear that mγ,j ≤ n. The equality holds if Ny = 0, rγ = 1 and p = 1, but this
is impossible if we have good coordinates. 

In this way we construct a tree of Newton transformations:

P
• We start with a root; the number of branches issued from this root is γ rγ .
• The other vertex of these branches correspond to fγ,wγ,j .
• For each vertex, if mγ,j = 1 we stop the process for this branch.
• If mγ,j > 1, we repeat the above construction for fγ,wγ,j . First, we perform a change of
variables as in Proposition 2.1.4. We attach again branches to this vertex in function
of the new Newton polygon.
Let us consider a vertex which is at and end of the tree; we have series fk in xk , yk such
that:
1
• xk = x l , for a suitable l.
P∞ j c
• y = j=1 aj x l + bx l yk , where these terms come from the changes of variables in order
to obtain good coordinates and the Newton transformations.
 2.2. BRAID GROUP 17

• Since fk is of order 1 in yk , using the Implicit Function Theorem we obtain that

fk (xk , yk ) = 0 is equivalent to y = hk (xk ), h ∈ C{xk }.
1 1
The final result of this process is a series h̃k (x l ) such that f (x, h̃k (x l )) ≡ 0.

Remark 2.1.8. At the Newton transformation there are q choices for w̃γ,j which are related
with the indetermination of h. Taking into account all the possibles indeterminations, we obtain
1
h̃k (ζx l ), ζ l = 1.

Remark 2.1.9. The branch of the tree associated with an edge γ and a root wγ,j such that
mγ,j = 1 stops after the Newton transformation.
P
Proposition 2.1.10. The number of irreducible components of f is at least γ rγ .

Example 2.1.11. Let us consider f (x, y) := (y 2 − x)2 − 4x2 y − x3 ; we have n = 4. We have

Figure 2.2. Newton polygon of (y 2 − x)2 − 4x2 y − x3

only one edge γ with one root 1. The Newton transformation is

x = x21 , y = x1 (y1 + 1).
We obtain
f (x, y) = x41 x1 + 4y1 + 4 + y12 y12 − x1



1 1 3
Then, up to a unit, fγ,1 = y12 − x1 . We have y = x12 , and then y = x1 + x1 y1 = x 2 + x 4 .

2.2. Braid group

Artin’s braid group Bn has a presentation

* +
σ1 , . . . , σn−1 [σi , σj ] , σi σi+1 σi = σi+1 σi σi+1 , .
1≤i<j<n 1≤i<n−1

There is a canonical epimorphism Ψn : Bn → Σn , σi 7→ (i, i + 1).

Let us denote by X the configuration space of n points in C: it is defined as the quotient of
Cn \ {(x1 , . . . , xn ) ∈ Cn | ∃i < j s.t. xi = xy } by the coordinate-permutation action of Σn . It
can be also interpreted as the set of subsets of C with cardinality n or else as
{p(t) ∈ C[t] | p monic , deg p = n, disc(p) 6= 0}.
18 2. BRAID MONODROMY OF GERMS

Paths in X are identified with sets of n paths γi : [0, 1] → C, i = 1, . . . , n, such that ∀t ∈ [0, 1],
#{γ1 (t), . . . , γn (t)} = n.
Let n := (−1, . . . , −n). It is possible to identify Bn with π1 (X; n), where σj is identified
with the homotopy class of the set of n paths [0, 1] → C given by:
t 7→ −k, k 6= j, j + 1,
2j + 1 exp(iπt)
t 7→ − + ,
2 2
2j + 1 exp(iπt)
t 7→ − − .
2 2
In particular, for any x ∈ X, Bn is isomorphic to π1 (X; x). We can construct isomorphisms
between Bn and π1 (X; x) in two essentially equivalent ways.
The first way is the standard one for identifying the fundamental group of a space at two
different points. Choose a braid τ starting at x and ending at n; the braid σj is sent to τ σj τ −1 .
The other way is as follows. Let us consider n − 1 paths δ1 , . . . , δn−1 such that δj :=
[−j, −(j + 1)]. Note that σj ∈ Bn ≡ π1 (X; n) consists on n − 2 straight segments and two
counter-clockwise half-circles having δj as the common diameter.

Figure 2.3. Basic segments

Let us order x := (x1 , . . . , xn ) and let us consider simple paths δjx , starting at xj and ending
at xj+1 , j = 1, . . . , n − 1, which intersect only at the extremities. We define an isomorphism
which sends σj to a set of paths starting and ending at x such that:
• The paths are constant at xk , k 6= j, j + 1.
• Consider a small tubular neighbourhood of δjx and consider two paths that run coun-
terclockwise the two semicircles in its boundary cut by δjx .

Definition 2.2.1. The ordered set of paths δjx , j = 1, . . . , n − 1, is a basic segment of x.

Remark 2.2.2. Note that a braid τ (seen as element of π1 (X; x)) is represented by a set of
paths starting and ending at x; the path starting at xj ends at xj σ , where Ψ(τ ) = σ ∈ Σn . It is
usual to represent braids as the union of the graphs of these paths in ∆ × [0, 1](⊂ R3 ) where ∆
is a disk which contains the paths in its interior.
There is other useful representation of a braid if one identifies ∆ × {0} and ∆ × {1}. The
braid produces a link (called the closed braid ) contained in a solid torus ∆ × S1 ⊂ S3 . Note that
conjugate braids produce the same link.
 2.3. LOCAL BRAID MONODROMY 19

Definition 2.2.3. A braid is positive if can be written as word in σ1 , . . . , σn−1 having only
positive exponents. A braid is quasipositive if it is conjugated to a positive braid.

Definition 2.2.4. The degree map of Bn is the homomorphism deg : Bn → Z such that
deg(σj ) := 1.

2.3. Local braid monodromy

Let f be a Weierstrass polynomial of degree n; for convenience, we will suppose that f is

reduced (there is no multiple irreducible component). We fix 0 < δ  ε  1 as in §1.1 and let
¯δ × ∆
Z := {(x, y) ∈ ∆ ¯ ε | f (x, y) = 0}.

¯δ × ∆
Definition 2.3.1. The discriminant locus of f relative to ∆ ¯ ε = W̄δ,ε is discf,δ :=
¯
{x0 ∈ ∆δ | discy (f )(x0 ) = 0}.

The main property of the discriminant locus is that the projection of Z onto ∆ ¯ δ is an
unramified covering outside discf,δ . In our case, discf,δ = {0}.
The covering π : Z ∗ → ∆¯ ∗ is classified by the classical monodromy at δ; let F := π −1 (δ).
δ
Since #F = n, ordering F we can identify ΣF with Σn . The classical monodromy is
¯ ∗δ ; δ) ≡ Z → Σn ,
ρ : π1 (∆

where ρ(1) is the product of r disjoint cycles corresponding to the r irreducible components of f .
In fact, we have much more; any loop based at δ lifts in a set of n non intersecting-paths
starting and ending at F , and (upon the choice of a basic segment of x) we can define the braid
monodromy of f as a morphism
¯ ∗ ; δ) ≡ Z → π1 (X; F ) ≡ Bn ,
∇ : π1 (∆ δ

such that Ψn ◦ ∇ = ρ.

Example 2.3.2. If f is irreducible and h is a Puiseux expansion, then the image by ∇ of

t 7→ δ exp(2iπt) is the set of paths {t 7→ h(ζδ exp(2iπt)) | ζ n = 1}. In the general case, we
consider the union of these paths for each irreducible component.

¯ ∗ ; δ) is denoted τf,π ,
Notation 2.3.3. The image by of the standard generator of π1 (∆ δ
because it depends both on f and the projection.

Theorem 2.3.4. The closed braid associated to τf is a model for the link of f . In particular,
τf is only well-defined up to conjugation.

We construct τf in several examples.

Example 2.3.5. Let us consider f (x, y) := y 2 − x. We can choose δ = 1 and hence F :=

{±1}. The set of paths is formed by t 7→ exp(iπt) and t 7→ − exp(iπt). If the basic segment is
formed by the segment [1, −1], the braid is σ1 .
20 2. BRAID MONODROMY OF GERMS

Figure 2.4. Braid for an ordinary tangent

k + 1 times

Figure 2.5. Braid for an Ak -singularity

Example 2.3.6. Let us consider f (x, y) := y 2 − xk+1 . We can choose again δ = 1 and hence
F := {±1}. The set of paths is formed by t 7→ exp((k + 1)iπt) and t 7→ − exp((k + 1)iπt). If the
braid in Example 2.3.5 is a single half-twist, in this case we have k + 1 half-twists. If the basic
segment is formed by the segment [1, −1], the braid is σ1k+1 .

Example 2.3.7. Let us consider f (x, y) := y k − x. We can choose again δ = 1 and hence
F := µk = {ζ ∈ C | ζ k = 1}. The set of paths is formed by t 7→ exp(2iπ t+2j
k ), j = 0, 1 . . . , k − 1.

k times
Figure 2.6. Braid for a flex

We choose as basic segment the arcs in the unit circle joining consecutively and counterclockwise
the k-roots of unity and starting at 1. The braid is σk−1 · . . . · σ1 .
 2.3. LOCAL BRAID MONODROMY 21

We can proceed in the same way for f (x, y) := y k −xl ; in this case the braid is (σk−1 ·. . .·σ1 )l .

Example 2.3.8. Let us consider f (x, y) := x(y k − xl ). We can proceed as in Example 2.3.7;
in this case we have a braid in k + 1 strings. There is another way of drawing this braid. We
consider a cylinder with k vertical lines in the boundary and its core and we glue the bottom
and the top faces by a rotation of angle − 2πl
k , see Figure 2.7.

− 2πl
k

Figure 2.7. Braid with core

Example 2.3.9. All the above examples have nice equations. In fact, their Puiseux ex-
1 3
pansions are minimal. What happens if we have a Puiseux expansion like y = x 2 + x 2 . The
main idea is that the last term only produces some noise which can be homotopically avoided.
Choosing δ small enough, we proceed as follows:
1
• Consider the braid associated to y = x 2 .
1 3
• We can construct a tube around this braid such that the braid of y = x 2 + x 2 is
3 1
contained in this tube, since the term x 2 is very small with respect to x 2 .
• Since the exponent of the second term has the same denominator, we can perform a
homotopy of the actual braid to the first one.

Example 2.3.10. Let us consider f (x, y) := (y 2 − x)2 − 4x2 y − x3 as in Example 2.1.11.

Note that n = 4 and the Puiseux expansion is h(t) = t2 + t3 . In the classical notation, we
1 3
have y = x 2 + x 4 , i.e., we have two Puiseux exponents. We proceed in this example as in
1
Example 2.3.9. We start with a curve having as parametrization y = x 2 ; the braid has two
strings, producing the braid σ1 . If we choose δ small enough, we have a braid with 4 strings
1
pairwise distributed in the two tubes. The core of these tubes is parametrized by y = x 2 . We
are drawing the braids in C × [0, 1] as usual, but they should be drawn in in C × S1 , where
22 2. BRAID MONODROMY OF GERMS

Figure 2.8. Perturbed braid

Figure 2.9. Two-pair braid

1
S1 := {t ∈ C | |t| = 1}. The tube becomes a solid torus and the braid associated to y = x 2 is its
core.
We can consider that the two parts of the tube glue in a tube ∆η × [0, 2], where the variables
1 1 3
are x1 := x 2 and y1 := y − x 2 . The new parametrization is y1 = x12 which is represented in the
left-hand cylinder of Figure 2.9. Cutting this cylinder in two pieces give the right-hand braid.
The basic segment is constructed as follows. First, we construct a basic segment in each small
disk which is used for the construction of the tube. Second, we join these basic segments using
the basic segment of the first Puiseux pair. The basic segments are those of Example 2.3.7.

We explain the inductive process to construct a braid associated to a germ f . First, consider
good coordinates for f and let Γf be its Newton polygon. Consider the highest edge γ of Γf .
Let pq be the anti-slope of γ, gcd(p, q) = 1. If γ is the unique edge of Γf we consider the braid
associated to y prγ −xqrγ ; if Γf has more edges, we consider the braid associated to x(y prγ −xqrγ ).
We replace this braid by tubes; when closing the braids, we obtain rγ tubes in the first case and
rγ + 1 in the second one.
Each tube different from the core is associated to a root wj,γ . The germ fγ,wγ,j is simpler
than the one of f and we may suppose constructed its braid. We cut it in the p sub-tubes.
 2.3. LOCAL BRAID MONODROMY 23

We forget now the first edge and we continue the process in the tube associated to the core.

Theorem 2.3.11. The braids associated to algebraic germs of curves are quasipositive.
 LECTURE 3

Global braid monodromy

3.1. Applications to the local case.

The construction of the local braid monodromy can be applied in a more general setting.

Proposition 3.1.1. Let g be a monic polynomial in y of degree n, whose coefficients are

¯ δ and holomorphic on ∆δ . Let us suppose that:
continuous functions on ∆
(1) ∀x0 ∈ ∆ ¯ δ , the roots of g(x0 , y) = 0 are in ∆ε (in particular, g −1 (0) ∩ (∆
¯ δ ∩ ∂∆
¯ ε ) = ∅).
¯
(2) discf,δ ∩∂ ∆δ = ∅.
Then, there exists a braid monodromy

¯ \ discg,δ ; δ) → Bn .
∇ : π1 (∆

We apply this proposition to the local braid monodromy of §2.3.

Theorem 3.1.2. Let τ := τf,π be a braid associated to a reduce germ f , regular of order n.
Then, deg τ = n + µ(f ) − 1.

Proof. For |t| small enough, we can apply Proposition 3.1.1 to f − t. We may suppose that
f −1 (t) is smooth. For a generic choice of t, for each x0 ∈ discf −t,δ there exists a unique y0 ∈ ∆ ¯ε
−1
(in fact in ∆ε ) such that f (x0 , y0 ) = t and the tangent line to f (t) at (x0 , y0 ) is vertical. Let

¯ δ \ discf −t,δ ; δ) → Bn
∇t : π1 (∆

be the braid monodromy of f − t. Let rt := # discf −t,δ ; the group π1 (∆ ¯ δ \ discf −t,δ ; δ) is free of
rank rt and a base γ1 , . . . , γrt is obtained as in Figure 3.1.
Note that ∇t (γi ) is decomposed as βi αi βi−1 , where βi is the braid associated to the straight
part and αi is the braid of the boundary of the small disk. The braid αi is formed by n−2 unlinked
straight braids and the two remaining strings behave as in Example 2.3.5, i.e., αi = σj for some j.
Q rt ¯ δ ; note that deg ∇t (γi ) = 1 and
Let ϕ := j=1 γrt −j+1 , i.e., the counterclockwise boundary of ∆
then deg ∇t (ϕ) = rt .
It is easily seen that ∇t (ϕ) = ∇(ϕ) = τ . We conclude that deg τ = rt .
If we compose ∇ with Ψn we obtain the classical monodromy of the unramified covering
f (t) \ π −1 (discf −t,δ ) → ∆
−1 ¯ δ \ discf −t,δ which is composed of rt transpositions. Since we have
an n-sheeted covering,

¯ δ \ discf −t,δ ) = n(1 − rt ) = n(1 − deg τ ).

χ(f −1 (t) \ π −1 (discf −t,δ )) = nχ(∆
25
26 3. GLOBAL BRAID MONODROMY

γ1

γ2
ϕ

γrt −1

γrt

¯ δ \ discf −t,δ ; δ)
Figure 3.1. Base of π1 (∆

Over each point of discf −t,δ we have n − 1 points:

µ = 1−χ(f −1 (t)\π −1 (discf −t,δ ))−(n−1) deg τ = 1−n(1−deg τ )−(n−1) deg τ = 1−n−deg τ. 

In general, it is also possible to make a change of variables such as n coincides with the
multiplicity m of f . Let us suppose that we have f a germ y-regular of order n, which is
bigger than the multiplicity of f . We construct ε and η as above, and we consider the braid
τf,π . Then, ft (x, y) := f (x + ty, y), for |t| small enough,is y-regular of order m. We can apply
Proposition 3.1.1 to ft , δ and ε and we denote ∇t : π1 (∆ ¯ δ \ discf ,δ ) → Bn . For a generic choice
t

of t the set discft contains 0 and n − m other points.

Let us choose a base γ0 , γ1 , . . . , γn−m as in Figure 3.1 such that γ0 runs around 0. Then
• ∇t (γ0 ) is conjugated to a braid obtained by adding constant maps to a braid of the
type βft ,π .
• ∇t (γj ), j > 0, are conjugated to σ1 .
• ∇t (γn−m · . . . · γ1 · γ0 ) = τf,π .

3.2. Braid monodromy of affine plane curves

Let us consider a reduced affine plane curve C with no vertical asymptotes and no vertical
lines. This means that C has an equation f (x, y) = 0, where f has no multiple factor and f is
monic as an element of C[x, y]; such a curve will be called horizontal. Let n := degy f (which
may be smaller than deg f ).
Let discf := {x ∈ C | discy f (x) 6= 0}. Since f has no multiple factor, then discy f is
a non-zero element of C[x], and then it is a finite set {x1 , . . . , xr }. Let M  0 such that
discf ⊂ ∆M . Since the roots of f (x0 , y) vary continuously with x0 then there exists N > 0,
¯ M and f (x, y) = 0, then y ∈ ∆ε . In particular, f, M, N are in the hypothesis
such that if x ∈ ∆
of Proposition 3.1.1.
 3.2. BRAID MONODROMY OF AFFINE PLANE CURVES 27

Definition 3.2.1. The homomorphism ∇ : π1 (∆M \discf ; M ) → Bn is the braid monodromy

morphism of C.

The homomorphism ∇ is only well-defined up to conjugation in Bn .

Example 3.2.2. Let us consider the affine curves y k − xl = 0, where discf = {0}. In this
case, braid monodromy is defined as in the local case ∇ : Z → Bk , where the image of the
generator of Z equals (σk−1 · . . . · σ1 )l .

Example 3.2.3. Since braid monodromy is well-defined up to conjugation, the particular

choice of a base point is irrelevant. Let us consider now f (x, y) := y d − x + 1; it is clear that
discf := {1}. One can choose M big and π1 (∆ ¯ M \ {1}; 0) is generated by γ : [0, 1] → ∆ ¯ M \ {1},
given by γ(t) := 1 − exp(2iπt). Following Example 3.2.2, we know that braid monodromy equals
σd−1 · . . . · σ1 .
Let us consider now f (x, y) := y d − xd + 1; in this case, discf := {ζ ∈ C | ζ d = 1}. Let us
consider the mapping x 7→ xd . Taking preimages, γ produces d loops based at 0, as in Figure 3.2.
¯ M \ discf ; 0) as in Figure 3.2, and it is easily seen that
We consider a basis γ1 , . . . , γd of π1 (∆

γ2
2iπ
e 3

0 1
γ3
1
0
2iπ
e− 3

γ1

Figure 3.2. Basis for d = 3

∇(γj ) = σd−1 · . . . · σ1 . Note that the product γd · . . . · γ1 is homotopic to the boundary of a

(topological) disk containing discf .

Definition 3.2.4. Let M > 0 and let A ⊂ ∆M , #A = d < ∞. We say that an ordered
¯ M \ A) is seudogeometric if γd · . . . · γ1 is homotopic to the
basis γ1 , . . . , γd of the free group π1 (∆
(positive) boundary of a (topological) disk containing A and it is constructed as in Figures 3.1
and 3.2. If the base point is in ∂ ∆ ¯ M and the above product is homotopic to ∂ ∆ ¯ M then the base
is called geometric.

Remark 3.2.5. Basic segments are closely related to geometric basis. Let us order A as
x1 , . . . , xd and let us consider a basic segment δ1 , . . . , δd−1 associated respecting this ordering.
28 3. GLOBAL BRAID MONODROMY

We choose small disks around x1 , . . . , xd and we choose a path δ0 from the base point in the
boundary to the first disk. The base γ1 , . . . , γd is obtained as follows:
• Start from δ0 , run counterclockwise the boundary of the first disk and come back
through δ0 : this is the construction of γ1 .
• Start again from δ0 , run counterclockwise the boundary of the first disk until encoun-
tering the path δ1 ; run δ1 until the boundary of the second disk. Then, run counter-
clockwise the boundary of the second disk and come back to the base point: this is the
construction of γ2 .
• The other elements are constructed in the same way.
Note that all bases we have used up to now are seudogeometric.
Definition 3.2.6. Let C be a horizontal curve of degree n and equation f = 0, r := # discf .
We say that an element (τ1 , . . . , τr ) ∈ (Bn )r represents the braid monodromy of C if there exists
a geometric basis γ1 , . . . , γr such that τi = ∇(γi ).
Example 3.2.7. Let us consider f (x, y) = y d − (x + ay)d + 1, |a|  1. It is possible to find
a geometric basis (with d(d − 1) elements) such that (σ1 , . . . , σd−1 , . . . , σ1 , . . . , σd−1 ) represents
the braid monodromy.
Definition 3.2.8. Let G be a group; the group Br acts naturally on Gr by Hurwitz moves.
If σ1 , . . . , σr−1 is the standard basis of Br , then the result of the action of σj on (g1 , . . . , gr ) is
−1
(g1 , . . . , gj−1 , gj+1 , gj+1 gj gj+1 , gj+2 , . . . , gr ).
This action commutes with the action of G on Gr by conjugation and it defines an action
of Br × G on Gr .
Theorem 3.2.9 (Artin). Two seudogeometric basis are related by an inner automorphism
and a Hurwitz move.
Corollary 3.2.10. Let C be a horizontal curve of degree n and equation f = 0, r := # discf ;
fix a representative (τ1 , . . . , τr ) ∈ (Bn )r of the braid monodromy of C. Then, an element of (Bn )r
represents the braid monodromy of C if and only if is in the orbit of (τ1 , . . . , τr ) by the action of
Br × Bn ; this orbit will be called the braid monodromy of C.

3.3. Zariski-van Kampen Theorem

The topology of a curve C is the homeomorphism type of the pair (C2 , C). Given M, N as
¯M × ∆
in §3.2, let us denote CM,N := C ∩ (∆ ¯ N ) and
¯M × ∆
KM,N := C ∩ ∂(∆ ¯ N ) = C ∩ (∂ ∆
¯M × ∆
¯ N ).

Proposition 3.3.1. The pair (∆ ¯M ×∆ ¯ N , CM,N ) is a strong deformation retract of (C2 , C).
Moreover,
¯M × ∆
(∆ ¯ N , CM,N ) ∪ (∂(∆¯M × ∆ ¯ N ), KM,N ) × [0, 1) ∼
= (C2 , C).
The topological type of (∂(∆¯M × ∆ ¯ N ), KM,N ) does not depend on M, N big enough and it is
called the link at infinity of C.
 3.3. ZARISKI-VAN KAMPEN THEOREM 29

We can interpret (∆ ¯ M ×∆
¯ N , CM,N ) as a compact model for (C2 , C). We are going to recover
the classical Zariski-van Kampen theorem from braid monodromy. The goal is to compute the
group:
¯M × ∆
G := π1 (C2 \ C) = π1 (∆ ¯ N \ CM,N ).
We denote X := ∆ ¯M × ∆¯ N \ CM,N and
¯ N },
X̌ := X \ {(x, y) | x ∈ discf , y ∈ ∆ Ǧ := π1 (X̌).

Theorem 3.3.2. The first projection defines a locally trivial fibration X̌ → ∆M \discf whose
fiber is a closed disk with n punctures.

Corollary 3.3.3. Let us denote Fm the free group with m generators. There is a short
exact sequence
(3.1) 1 7→ Fn → Ǧ → Fr → 1.

Proof. We consider the long exact sequence of homotopy of the fibration of Theorem 3.3.2.
The result follows since π1 (∆M \ discf ) is free of rank r, π2 (∆M \ discf ) is trivial and the
fundamental group of the fiber is free of rank n. 

In order to obtain a presentation of Ǧ from (3.1) we proceed as follows:

(P1) Choose a geometric basis γ1 , . . . , γn of of π1 ({M } × (∆ ¯ N \ FM ; (M, N )), where FM :=
{y | f (M, y) = 0}, the basis is associated to a basic segment.
(P2) Choose a geometric basis α1 , . . . , αr of π1 (∆ ¯ M \ discf ; M ).
¯
(P3) Lift this geometric basis to ∆M \ discf ×{N }. The lifted elements are denoted by
β1 , . . . , βr .
(P4) Compute βi−1 γj βi as a word in γ1 , . . . , γn .
Braid monodromy appears in (P4). We consider a braid monodromy morphism ∇ : π1 (∆ ¯M \
discf ; M ) → Bn which uses the basic segment of γ1 , . . . , γn . The representative of the braid
monodromy we will consider is (∇(α1 ), . . . , ∇(αr )).
There is a natural action of Fn × Bn → Fm given by:

γi

 if i 6= j, j + 1,
σ
γi j := γ if i = j,
 j+1
−1

γj+1 γj γj+1 if i = j + 1.


σj
Note that (γn · . . . · γ1 ) = γn · . . . · γ1 .
This action can be explained geometrically as follows:
• Identify π1 ({M } × (∆ ¯ N \ FM ; (M, N )) with Fn via γ1 , . . . , γn .
• A braid τ where we identify Bn with π1 (x; FM ) via the basic segment.
It is easily seen that
¯ N × [0, 1]) \ τ ∼
Xτ := (∆ ¯ \ FM ) × [0, 1].
= (∆
For t ∈ {0, 1}, the mappings jt (x) := (x, t) induce isomorphisms (jt )∗ : Fn → π1 (Xτ ; (N, t)). If
γ ∈ Fn then γ τ is constructed as follows:
30 3. GLOBAL BRAID MONODROMY

• Consider (j0 )∗ (γ) ∈ π1 (Xτ ; (N, 1)).

• If v : [0, 1] → Xτ is defined by v(t) := (N, t), then γv := v −1 (j0 )∗ (γ)v ∈ π1 (Xτ ; (N, 0)).
• Finally γ τ := (j1 )∗ )−1 (γv ).
∇(αi )
Lemma 3.3.4. βi−1 γj βi = γj .
¯ M \ discf by the mapping αi ;
Proof. The space Xτ is the pull-back of the fibration X̌ → ∆
by this pull-back, βi is sent to v and the result follows. 

Proposition 3.3.5 (Fibered version of Zariski-van Kampen). The group Ǧ is generated by

∇(α )
γ1 , . . . , γn , β1 , . . . , βr with the relations βi−1 γj βi = γj i .

As a consequence of the standard Seifert-van Kampen Theorem, we have the following result.

Theorem 3.3.6 (Zariski-van Kampen). The group G is generated by γ1 , . . . , γn with the

∇(α )
relations γj = γj i , j = 1, . . . , n − 1, i = 1, . . . , r.
 LECTURE 4

Topological applications

4.1. Puiseux monodromy and topology

The usual way to construct braid monodromy is as follows. We decomposed αi = ζi−1 α̃i ζi ,
where α̃i is the counterclockwise boundary of a small disk Di and ζi joins α̃i with the base point,
as it is shown in Figure 3.1. A suitable choice of basic segments determines a decomposition
∇(αi ) = τi−1 ρi τi , where ρi is an unlinked union of the braids associated with the Puiseux
expansions of the points of C with first variable xi , see §2.3. The braids ρi have an algebraic
meaning, but it is usually difficult to compute the braids τi .

Definition 4.1.1. A decomposition ∇(αi ) = τi−1 ρi τi as above, j = 1, . . . , r, is called a

Puiseux monodromy of C.

Proposition 4.1.2. Let us consider a Puiseux monodromy of C and let us assume that αi
is an unlinked union of ri local braids αi,k , k = 1, . . . , ri . Let mi,k be the number of strings of
∇(α )
αi,k (n = k mi,k ). Then, the relations γj = γj i can be replaced by γjρi τi = γjτi . Moreover,
P

for each bunch of mi,k relations one can drop one of them.

Theorem 4.1.3 (Libgober). If we replace the relations of Theorem 3.3.6 by the relations of
Proposition 4.1.2, the resulting presentation has the homotopy type of X.

From a Puiseux monodromy one can reconstruct the topology of (C2 , C).

• Above the base point M , we consider the pair (C, n).

• Over each path ζi we consider the pair (ζi × C, τi ).
• Over each loop α̃i , we consider α̃i × C with the closed braid ρi .
• Over each disk Di we consider Di × C with ri cones based at different vertices and
based on the ri braids giving α̃i
¯ M we consider the braid ∇(γr · . . . · γ1 .
• Above ∂ ∆
• We complete the rest of ∆ ¯ M using the homotopy between the braids.
• We complete to C using a product structure.

This construction depends on the Puiseux monodromy but a stronger result is true.

Theorem 4.1.4 (Carmona). Two curves having the same braid monodromy are topologically
equivalent.
31
32 4. TOPOLOGICAL APPLICATIONS

4.2. Discriminant of polynomial mappings

Definition 4.2.1. A polynomial f (x, y) is good at infinity if all the curves Ct : f (x, y) = t
have the same link at infinity.
Let f (x, y) = fd (x, y) + fd−1 (x, y) + · · · + f1 (x, y) + f0 be the homogeneous decomposition
of f . Consider the standard compactification C2 ,→ P2 given by (x, y) 7→ [x : y : 1] in the
homogeneous coordinates [x : y : z]; the line at infinity L∞ := P2 \ C2 has equation z = 0 and
the compactifications of the curves Ct have the homogeneous equations:
fd (x, y) + fd−1 (x, y)z + · · · + f1 (x, y)z d−1 + f0 z d = tz d .
We can decompose
r
Y
fd (x, y) = (aj y − bj x)mj , (aj , bj ) 6= (0, 0).
j=1

Lemma 4.2.2. ∀t ∈ C, C̄t ∩ L∞ = {[aj : bj : 0] | j = 1, . . . , r} and these points are called the
points at infinity of f .
Fix a point Pj := [aj : bj : 0]; for simplicity we suppose Pj = [0 : 1 : 0]. We consider the
germs {(C̄t , Pj )}t∈C ; in the coordinates x, z they are defined by
fd (x, 1) + fd−1 (x, 1)z + · · · + f1 (x, 1)z d−1 + f0 z d = tz d .
Proposition 4.2.3. The polynomial f is good at infinity if and only if ∀j = 1, . . . , r, the
Milnor number µPj ,t := µ(C̄t , Pj ) is independent of t ∈ C (such a polynomial has isolated critical
points).
Example 4.2.4. If mj = 1 for all j (in particular r = d), then f is good at infinity. In fact
in that case the link at infinity is the Hopf link with d components.
Example 4.2.5 (Broughton). Let f (x, y) := x(xy + 1). We have two points at infinity
P1 := [1 : 0 : 0] and P2 := [0 : 1 : 0]. At P1 we can consider coordinates y, z and the family is
given by y + z 2 − tz 3 = 0 and µP1 ,t = 0. At P2 we can consider coordinates x, z and the family
is given by x2 + xz 2 − tz 3 = 0 and µP2 ,t = 2 if t 6= 0, µP2 ,0 = 3. In particular, f is not good at
infinity.
Theorem 4.2.6 (Thom). Let f : C2 → C a polynomial. There exists a minimal finite set
B(f ) (the bifurcation set of f ) such that f| : C2 \ f −1 (B(f )) → C \ B(f ) is a locally trivial
fibration. If f is good at infinity, then B(f ) coincides with the set Crit(f ) of critical values.
Example 4.2.7. In Example 4.2.5, B(f ) = {0} and there is no critical value.
A nice invariant of the topology of f is the monodromy action associated to the fibration
f| : C2 \ f −1 (B(f )) → C \ B(f ). In the homological level, if we choose t0 ∈ C \ B(f ) we have a
morphism
ρ : π1 (C \ B(f ); t0 ) → Aut(H1 (Ct0 ; Z)).
In the case of polynomials which are good at infinity, we can understand it with Picard-Lefscthez
techniques. For technical reasons, we restrict ourselves to a narrower class of polynomials.
 4.2. DISCRIMINANT OF POLYNOMIAL MAPPINGS 33

Definition 4.2.8. A polynomial f is tame if the polynomials f (x, y)+a` are good at infinity
and have the same link at infinity, where ` is a generic linear form and |a| is small enough.

Example 4.2.9 (Cassou-Noguès). Of course, any tame polynomial is good at infinity. The
polynomial f (x, y) = (x2 + y 3 )2 + y is good at infinity but not tame.

From now on we fix a tame polynomial f and we keep the previous notations. Let us suppose
that x is generic for Definition 4.2.8.
S
We fix a geometric basis γ1 , . . . , γr of π1 (C \ Crit(f ); t0 ) = π1 (C \ t∈Crit(f ) Dt ; t0 ) where
D(t) is a small disk centered at t, for each t ∈ Crit(f ).

Lemma 4.2.10. If |a| is small enough the mapping fa := f − ax : C2 → C is a Morse

function, i.e., the critical points of fa are ordinary double points (of local equation u2 + v 2 = 0)
and have pairwise distinct critical values. Moreover:
(M1) The generic fibers of f and fa are isotopic.
(M2) Let p ∈ f −1 (t1 ) be a critical point of f ; there are µ(f −1 (t1 ), p) Morse points of fa close
to p (whose values are in D(t1 )).
S
(M3) The mappings f and fa define isotopic locally trivial fibrations over C \ t∈Crit(f ) D(t).

Picard-Lefscthez theory allows us to compute the main invariants of these fibrations: fiber,
intersection form and monodromy. We consider a morsification fa .

Notation 4.2.11. For t ∈ Crit(f ), we call Sa,t the set of values of the Morse points coming
from points in f −1 (t).

We choose a geometric basis of π1 (C \ Crit(fa ); t0 ) which is compatible with the one of

π1 (C \ Crit(f ); t0 ) as in Figure 4.1; namely the reversed product of loops associated with the
values in Sa,t equals the corresponding loop for t.

Figure 4.1. Geometric basis of the morsification

In this way, we obtain a distinguished basis of vanishing cycles for H1 (f −1 (t0 ), Z) (identified
with H1 (fa−1 (t0 ), Z) together with the intersection form h•, •i in terms of this basis. Following
34 4. TOPOLOGICAL APPLICATIONS

the results of §1.4, given a vanishing cycle a coming from an loop γ around the corresponding
singular value of fa , then the monodromy around γ is given by:
b ∈ H1 (f −1 (t0 ), Z) 7→ b + ha, bib.
As in the local case, we obtain the monodromy action for the elements of the geometric basis
π1 (C \ B(f ); t0 ) and the morphism ρ is computed.
In order to effectively compute the distinguished basis and the intersection form we will
consider the polar mapping. Let ψ := (f, x) : C2 → C2 (we will denote t the first variable of the
second C2 ). This mapping is a submersion outside the polar locus Γ defined by ∂f ∂y (x, y) = 0.
Let ∆ := ψ(Γ) be the discriminant locus of ψ. We have that ψ| : C2 \ ψ −1 (∆) → C2 \ ∆ is an
unramified covering of degree, say n. Moreover the equation of ∆ is given by discy (f (x, y)−t) = 0
(equation in x, t).

Remark 4.2.12. We do not need to compute discy (fa (x, y) − t) since it is obtained from
discy (f (x, y) − t) by the change t 7→ t + ax. We denote ∆a the curve discy (fa (x, y) − t) = 0.

Lemma 4.2.13. The polynomial f is good at infinity if and only if ∆ is horizontal and it is
tame if and only if deg(discy (f (x, y) − t)) = degx (discy (f (x, y) − t)).

Let us consider the vertical line Lt0 of equation t = t0 . We have a ramified covering
φ| : f −1 (t0 ) → Lt0 which is used as multiparametrization in Riemann’s sense. The associated
unramified covering is φ| : f −1 (t0 ) \ φ−1 (∆) → Lt0 \ ∆.

Lemma 4.2.14. A value t1 ∈ C is in Crit(f ) if and only if there is a point (x, t1 ) ∈ ∆ such
that the vertical line t = t1 belongs to the tangent cone of ∆ at (x, t1 ).
In particular, singular values of fa correspond to ordinary vertical tangencies of ∆a .

Let us fix a vertical tangency of ∆a at (x, t1 ). At the line t = t1 + ε, 0 < ε  1, we

can consider a vanishing path δ, whose extremities are two points in ∆ which are close to the
tangency.

Figure 4.2. Vanishing paths

The preimage of this vanishing path in fa−1 (t1 − ε) (|ε| small) is composed by n edges; two
of these edges share the extremities and its union define a vanishing cycle (up to orientation) in
fa−1 (t1 − ε).
 4.2. DISCRIMINANT OF POLYNOMIAL MAPPINGS 35

This vertical tangency is related to a loop γ in the geometric basis of π1 (C \ B(fa ); t0 ); let
τ be the braid monodromy of ∆a over γ. We can decompose τ = βσ1 β −1 , where β is a braid
which geometrically corresponds to a path from t0 to t1 − ε. This braid β allows us to transport
the edge γ to a vanishing path δ̃ in fa−1 (t0 ) ∼
= f −1 (t0 ).

Remark 4.2.15. It is possible to prove that the above construction does not depend on a
particular decomposition τ = βσ1 β −1 .

Theorem 4.2.16 (Escario). The classical monodromy of φ| : f −1 (t0 ) \ φ−1 (∆) → Lt0 \ ∆
allows us to construct the corresponding vanishing cycle from δ̃.