Complete intersections with isolated singularities (ICIS)

Algebraic methods and singularities

T. Gaffney, J. J. Nuño-Ballesteros, B. Oréfice-Okamoto,

M. A. S. Ruas & R. Wik-Atique

D4 singularity, Oliver Labs, Imaginary

CIMPA RESEARCH SCHOOL

Singularities and its applications, 11–22 July 2022

ICMC–USP São Carlos, Brazil.

2
Contents

I Lecture 1 3
1 Preliminary results 5
1.1 Necessary algebraic results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 The local ring On . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Germ of Complex Analytic Sets . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1 The local algebra of a germ of analytic space . . . . . . . . . . . . . . 12
1.3.2 Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.3 Finite germs of maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.4 The singular locus and the Jacobian Criterion . . . . . . . . . . . . . . 16
1.4 Isolated Complete intersection Sigularities . . . . . . . . . . . . . . . . . . . . 17

2 The contact group 19

2.1 The K group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

II Lecture 2 21
3 K -Determinacy 23
3.0.1 Infinitesimal criteria of finite K -determinacy . . . . . . . . . . . . . . 23
3.0.2 Geometric criterion of finite K -determinacy . . . . . . . . . . . . . . 24
3.1 Classification of stable germs . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Complete transversal method . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4 K -Versal unfoldings of map germs (real and complex) 27

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Basic definitions and examples . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3 Characterizations of versality . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 Relation between K -equivalence and A -equivalence . . . . . . . . . . . . . . 31

III Lecture 3 37
5 Deformations of ICIS 39
5.1 Basic invariants of ICIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Versal deformations of ICIS . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3 Some analytic properties of versal deformations . . . . . . . . . . . . . . . . . 46

i
ii CONTENTS

6 Topology of the generic fibres 49

6.1 The link of an isolated singularity . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 The Milnor fibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.3 The homotopy type of the Milnor fibre . . . . . . . . . . . . . . . . . . . . . . 52
6.4 The Gaffney-Hauser theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

IV Lecture 4 59
7 Equisingularity and ICIS 61
7.1 Introduction and some basic examples . . . . . . . . . . . . . . . . . . . . . . 61
7.2 Rugose vectorfields and Verdier’s condition W . . . . . . . . . . . . . . . . . . 63
7.3 The Theory of Integral Closure of Ideals and Modules . . . . . . . . . . . . . . 66
7.4 Multiplicities and Integral closure . . . . . . . . . . . . . . . . . . . . . . . . 80
Introduction

This is the second half of the course on ”Singularities and Algebraic Methods". The first half of
the course was about hypersurfaces and was taught in 2021 in the Part 1 of the CIMPA Research
School "Singularities and Applications" (online version). We will assume the reader is already
familiar with the contents of that part.
In this second half we will focus on complete intersections with isolated singularities (ICIS).
The complete intersections are a natural generalisation of hypersurfaces in the following sense:
a hypersurface singularity (X, x) is set germ in Cn+1 defined as the zero locus of a single
non-constant holomorphic function f : (Cn+1, x) → (C, 0). This forces that (X, x) must have
dimension n. If we consider now k holomorphic functions f = ( f 1, . . . , f k ) : (Cn+k , x) →
(C k , 0) then in general it is not true that the zero locus (X, x) of f has also dimension n. But
when this happens we say that (X, x) is a complete intersection.
In the hypersurface case, the algebraic methods to study the classification and the main
invariants are provided by the R-equivalence of smooth functions f : (Kn+1, 0) → (K, 0) where
K = R or C and R is the Mather’s group of right equivalences. We saw in the first half of
the course how the infinitesimal methods can be applied to describe the tangent space to the
R-orbit and hence, to obtain algebraic criteria for finite determinacy, which is a crucial step in
the classification process.
In the case of ICIS, we introduce the Mather’s contact group K which acts on the space
of smooth map germs f : (Kn, 0) → (K p, 0). When K = C and n ≥ p, we will see that there
exists a one-to-one correspondence between the isomorphism classes of ICIS (X, x) and the
K -equivalence classes of map germs f : (Cn, 0) → (C p, 0). Again, the infinitesimal machinery
will enter into action to provide algebraic methods for the classification and the invariants of
these singularities.
The course is organised into 4 lectures. In the first lecture, we will review some basic results
about commutative algebra and local analytic geometry that will be needed for the course. We
will give the precise definition of ICIS. We also introduce the contact group K and the notion of
K -equivalence of map germs f : (Kn, 0) → (K p, 0). The main result will be that two map germs
f and g are K -equivalent if and only if their local algebras Q( f ) and Q(g) are isomorphic.
The second lecture is dedicated to finite determinacy and versality of unfoldings of maps
for the contact group K and for map germs f : (Kn, 0) → (K p, 0). We will present first the
infinitesimal criterion of finite K -determinacy, which says that f is finitely K -determined if and
only if its Ke -codimension is finite. In the complex case, finite K -determinacy is equivalent
to that f is finite-to-one on its critical locus, which is known as the geometric criterion of
Mather-Gaffney. Then, we will introduce the complete transversal method, which is very useful
to obtain classifications under K -equivalence. The last part of this lecture is about K -versality
of unfoldings of maps. The main result is that the r-parameter unfolding F is versal if and only
if it is transversal to the K -orbit, that is, the residue classes of the partial derivatives of F with

1
2 CONTENTS

respect to the parameters generate the quotient θ( f )/TKe f over K. Then, some consequences
are obtained about the relationship between K -versality and A -stability.
In the third lecture we will study versal deformations of ICIS and also the topology of the
generic fibres. Here we restrict ourselves to the complex case K = C and consider an ICIS
(X, x) given as the fibre of a K -finite map germ f : (Cn+k , x) → (C k , 0). We will see how the
notion of versal deformation of the ICIS (X, x) is strongly related to the K -versal unfoldings of
f . Then, we will deduce some interesting properties of the discriminant of a versal unfolding.
We will introduce the link of an ICIS (X, x) and also the Milnor fibration. As in the case of
an hypersurface with isolated singularity, the Milnor fibre has the homotopy type of a wedge
of n-spheres and the number of such spheres is called the Milnor number, denoted by µ(X, x).
We will prove a theorem of Gaffney and Hauser that gives a criterion to prove that two ICIS are
isomorphic in terms of the modules of infinitesimal deformations.
Finally, in the fourth lecture we will present an introduction to equisingularity, with applica-
tions to the case of ICIS. We will present Whitney’s conditions (a) and (b) as well as Verdier’s
W condition. Next, we will give basic definitions and properties about integral closure of ideals
and modules and their connection with equisingularity. In the last section, we will give an appli-
cation to characterise the Whitney equisingularity of families of ICIS in terms of multiplicities
and also the constancy of the µ∗ -sequence.

São Carlos, July 2022.

 Part I

Lecture 1

3
Chapter 1

Preliminary results

1.1 Necessary algebraic results

We will assume that all the rings have identity and thta if ϕ : R → S is a homomorphism of
rings then ϕ(1 R ) = 1S .

Definition 1.1.1. A ring R is called Noetherian if every ideal in R is finitely generated.

Lemma 1.1.2. Let R be a ring. Then the following conditions are equivalent:

1. R is Noetherian.

2. Any chain of ideals in R

I1 ⊂ I2 ⊂ · · · ⊂ Ik ⊂ · · ·
becomes stationary, that is, there exists an n > 0 such that Ik = In for all k ≥ n.

3. Every nonempty set of ideals in R has a maximal element with respect to inclusion.

Proof. See, for instance, [6, Lemma 1.1.2]. 

Definition 1.1.3. A ring R is called local if it has an unique maximal ideal, m. One often says
that (R, m) is a local ring to indicate that m is its unique maximal ideal.

Lemma 1.1.4. Let R be a ring and m ⊂ R be an ideal. Then R is local, with maximal ideal m,
if and only if R\m is the set of the units of R.

Proof. Assume that R is local with maximal ideal m. Take an element a ∈ R. If (a) , R, then
a belongs to a maximal ideal, therefore a ∈ m. That is, if a < m then a is a unit. Moreover, if
a ∈ m thana is not a unit, as otherwise (a) = R.
Suppose, on the other hand, that R\m is the set of units. Take an ideal I , R. Then the ideal
I does not contain units, and therefore must be contained in m. This shows that m is the unique
maximal ideal of R. 

Theorem 1.1.5. [6, Theorem 1.3.4] (Nakayama’s Lemma). Let (R, m) be a local ring and M
be a finitely generated R-module with m · M = M. Then M = 0.

5
6 CHAPTER 1. PRELIMINARY RESULTS

Proof. We assume that M , 0. Let t be the minimal number of generators of M and

m1, m2, . . . , mt be a set of generators of M. Since mM = M, there exist a1, . . . , at ∈ m
such that
mt = a1 m1 + a2 m2 + · · · + at mt .
Therefore
(1 − at )mt = a1 m1 + a2 m2 + · · · + at−1 mt−1 .
But, since 1 − at < m and R is local then 1 − at is a unit and, then, mt can be generate
by m1, m2, . . . , mt−1 . Hence, M is generated by m1, m2, . . . , mt−1 , in contradiction with the
minimality of generators of M. 
Corollary 1.1.6. [6, Corollary 1.3.5] (Krull’s Intersection Theorem). Let (R, m) be a Noetherian
local ring, and M be a finitely generated R-module. Then

∩ k∈N m k M = (0).

Proof. We will assume that M = R. The general case is analogous. We write I = ∩ k∈N m k . By
the Nakayama’s Lemma, we just need to show that mI = I. We consider the set

A = { J ideal in R : J ∩ I = mI}.

Clearly mI ∈ A and, since R is Noetherian then A has a maximal element, which we call J.
We claim that there exist a k ∈ N such that m k ⊂ J. In fact, since m is finitely generated, we
just need to show that for each f ∈ m, there exists α ∈ N such that f α ∈ J. For each f ∈ m we
consider the chain of ideals
J : f ⊂ J : f2 ⊂ ....
Since the ring is Noetherian, this chains stabilizes, that is, there exists α such that J : f α ⊂ J :
f α+1 . This is the α we were looking for:

x ∈ (J + ( f α )) ∩ I ⇒ x = y + a f α ∈ I, with y ∈ J, a ∈ R ⇒ a f α+1 = f x − f y ∈ mI + J = J,

hence a ∈ J : f α+1 = J : f α and then x ∈ I ∩ J = mI. Therefore, since the other inclusion is
trivial, (J + ( f α )) ∩ I = mI. Since J is maximal in A, (J + ( f α )) = J, which completes the
proof of the claim.
To conclude the proof we observe that I ⊂ m k ⊂ J and then I ⊂ J ∩ I = mI. 
Another important corollary of the Nakayama’s Lemma is the following lemma on the
number of generators of a module over a local ring.
Corollary 1.1.7. [6, Corollary 1.3.6] Let M be a finitely generated module over a local ring
(R, m).
1. Let f 1, . . . , f s ∈ M such that the classes of the f i generate M/mM as R/m-vector space.
Then f 1, . . . , f s generate M.

2. The minimal number of generators of M is equal to dim R/m (M/mM).

In particular, if (R, m) be a Noetherian local ring. The embedding dimension, edim(R), of
R is defined by
edim(R) := dim R/m (m/m2 ).
1.1. NECESSARY ALGEBRAIC RESULTS 7

Definition 1.1.8. An analytic algebra (also called analytic C - algebra) is a C - algebra of type
C{x 1, . . . , x n }/I, where I is an ideal in C{x 1, . . . , x n }.

Lemma 1.1.9. Let (R, m R ), (S, m S ) be analytic C-algebras and ϕ : R → S a morphism of

C-algebras. Then ϕ(m R ) ⊂ m S .

Proof. In fact, if not then there exist f ∈ m R such that ϕ( f )(0) = c , 0. Hence ϕ( f )(0)−c ∈ m R
is not a unit. Since it is a ring homomorphism, ϕ sends units in R to units in S, therefore
ϕ( f )(0) − c is not a unit. However, ϕ is also a C-vector space homomorphism and then
ϕ( f ) − c = ϕ( f − c) which is a unit. 

Definition 1.1.10. The height of a prime ideal P of a ring R, ht(P), is the maximum of the
length, n, of a chain of strict inclusions

P0 ⊂ P1 ⊂ P2 · · · ⊂ Pn = P.

where all of the Pi are prime ideals of R. The Krull dimension of a ring R is the maximum of
the heights of prime ideals in R. Usually this is simply called the dimension of R. The height of
a not necessarily prime ideal I is defined to be the minimum of the heights of the prime ideals
containing I. The (Krull) dimension of a module, M, dim M, is the Krull dimension of the ring
R/Ann R (M), where Ann R (M) = {r ∈ R : rm = 0, ∀m ∈ M }.

We say that a Noetherian local ring , R, is regular if edimR = dim R.

Definition 1.1.11. [6, Definition 6.5.1] Let (R, m) be a local ring and M be an R-module.

1. A sequence f 1, . . . , f r of elements in m is called a regular sequence of M if f 1 is not a

zerodivisor of M, and f i is not a zerodivisor of M/( f 1, ..., f i−1 )M for i = 2, . . . , r.

2. Let I ⊂ I R be an ideal with I M , M. Then the I-depth of M, depth(I, M) is the maximal

length of a regular sequence of M in I. If I M = M we define depth(I, M) = ∞.

3. The depth of M, depth(M) is the maximal length of a regular sequence of M, that is,
depth(M) = depth(m, M). If we want to emphasize the ring R, we will write depth R (M).

Over a Noetherian local ring, it is always true that

depth(M) ≤ dim M.

When the equality holds, we say that the module M is Cohen-Macaulay. The ring R is called
Cohen-Macaulay, if R is a Cohen-Macaulay R-module.
Let R be a ring. An element x ∈ R is called nilpotent if there exists an n ∈ N with x n = 0.
A ring is called reduced if it has no nonzero nilpotent elements.
One can show that a ring R is reduced if and only if it satisfies the Serre’s conditions R0 and
S1 ,
R0 : Rp is regular for every prime ideal p of R with height equals to zero.
S1 : depth Rp ≥ min{ht(P), 1}, for every prime ideal p of R.
Here Rp denotes the localization of R on R − p.
8 CHAPTER 1. PRELIMINARY RESULTS

1.2 The local ring On

Let N and P be real or complex manifolds of dimensions n and p respectively, and x ∈ N.
In the set of smooth (C ∞ in real case or holomorphic in complex case) mappings defined in a
neighborhood of x in N into P we introduce the following equivalence relation:

Definition 1.2.1. Two mappings f 1 : U1 → P and f 2 : U2 → P are equivalent if there exists a

neighborhood U ⊂ U1 ∩ U2 of x in N such that f 1 |U and f 2 |U coincide.
An equivalence class under this relation is called germ of mapping or map-germ at x and
is denoted by f : (N, x) → (P, y), y = f (x). An element of an equivalence class is called
representative of the germ.

Let f : (N, x) → (P, y) be a map-germ at x. Its derivative d f x : Tx N → Ty P is defined

as the derivative at x of any representative of the germ. We say that f : (N, x) → (P, y) is a
germ of a diffeomorphism if one of its representative (and so any) is a local diffeomorphism. It
follows from the Inverse Function Theorem that a map-germ at x is a germ of a diffeomorphism
if and only if its derivative at x is an isomorphism.
Let f : (N, x) → (P, y) and g : (P, y) → (M, z) be two map-germs where M is a manifold
and z ∈ M. We define the composition g ◦ f : (N, x) → (M, z) as: take representatives
f˜ : U → P and g̃ : V → M, f˜(U) ⊂ V , of f and g respectively, the map-germ g ◦ f is the
equivalence class of g̃ ◦ f˜.
The rank of a map-germ at x is defined as the rank of its derivative at x. When the rank is
n the map-germ is immersive and when the rank is p the map-germ is submersive. We say that
the map-germ is singular when it is neither immersive nor submersive.

Definition 1.2.2. Two map-germs f 1 : (N1, x 1 ) → (P1, y1 ) and f 2 : (N2, x 2 ) → (P2, y2 ) are A -
equivalent if there exist germs of diffeomorphisms h : (N2, x 2 ) → (N1, x 1 ) and k : (P2, y2 ) →
(P1, y1 ) such that the following diagram commutes:
f1
(N1, x 1 ) −→ (P1, y1 )
h↑ k↑
f2
(N2, x 2 ) −→ (P2, y2 )
that is, f 1 = k ◦ f 2 ◦ h−1 (or f 1 ◦ h = k ◦ f 2 ).

Note that since any map-germ f : (N, x) → (P, y) is A -equivalent to some germ g :
(Kn, 0)→ (K p, 0), where K = R or C, we consider only smooth map-germs (Kn, 0) → (K p, 0).

We denote by On,p the set of smooth map-germs f : (Kn, 0) → K p . When p = 1 we denote

it by On .
On is a local ring whose maximal ideal is mn = { f ∈ On : f (0) = 0}.

1.3 Germ of Complex Analytic Sets

We present here a small introduction about germs of analytic spaces. The text is based on the
Chapter 3 of [6]. See also [19].
1.3. GERM OF COMPLEX ANALYTIC SETS 9

Definition 1.3.1. 1. A set X ⊂ Cn is called locally analytic if, for each p ∈ X, there exists
an open subset V of Cn with p ∈ V and finitely many holomorphic functions f 1, . . . , f s
such that
X ∩ V = {x ∈ V : f 1 (x) = · · · = f s (x) = 0}.

2. Let U be an open subset of Cn . A subset X ⊂ U is called an analytic subset of U, if X is

locally analytic and closed in U.

Given holomorphic functions f 1, . . . , f s : V ⊂ Cn → C, with V open, we denote the analytic

subset of V defined as the zeroset of f 1, . . . , f s by V ( f 1, . . . , f s ). That is

V ( f 1, . . . , f k ) = {x ∈ V : f 1 (x) = · · · = f s (x) = 0}.

Example 1.3.2. 1. Cn = V (0) and ∅ = V (1) are analytic subsets of Cn .

2. The set
X := {(x, y) ∈ C2 : y = 0, Im(x) ≥ 0}
is not locally analytic. In fact, assume that there exist a connected open subset V of C2
with (0, 0) ∈ V and holomorphic functions f 1, . . . , f s defined on V such that

X ∩ V = {(x, y) ∈ V : f 1 (x, y) = · · · = f s (x, y) = 0}.

Therefore X ∩ V ⊂ {(x, y) ∈ V : f 1 (x, y) = 0}. We denote gi (y) = f i (x, 0), i = 1, . . . , s.

By the Identity Theorem ([6, Theorem 3.1.9]) in one variable, gi |V ∩{y=0} is identically
equal to 0. Hence {(x, y) ∈ V : y = 0} ⊂ X ∩ V .

Definition 1.3.3. 1. Let f 1, . . . , f n be holomorphic on an open subset U ⊂ Cn . Let p ∈ U,

and suppose f 1 (p) = · · · =  f n (p) = 0. The set { f 1, ..., f n } is called a set of coordinate
∂ fj
functions at p if det ∂ xi (p) , 0.

2. A subset X ⊂ Cn is called a complex submanifold of Cn if for every x ∈ X there exists an

open subset U in Cn and coordinate functions w1, ..., wn of x such that

X ∩ U = {y ∈ U : w1 (y) = · · · = wm (y) = 0}

for some m ≤ n.

Definition 1.3.4. Let U ⊂ Cn be an open set and X ⊂ U an analytic subset. A point x ∈ X is

called regular, or X is called smooth at x, if there exists an open subset V in Cn with x ∈ V such
that X ∩ V is a complex submanifold of Cn . If x ∈ X is not regular, then x is called singular, or
a singularity. The set of singular points of X is denoted by Sing(X ).

In order to study local properties, we introduce the notion of germs of set and of analytic
space.

Definition 1.3.5. 1. Let X be a topological space and p ∈ X. Two subsets A and B of X are
called equivalent at p if there exist an open neighborhood U of p such that A ∩ U = B ∩ U.
The equivalence class of a subset A is called germ of A at p and denoted by ( A, p).
10 CHAPTER 1. PRELIMINARY RESULTS

2. Let ( A, x) and (B, x) be germs. We define ( A, x) ⊂ (B, x) if there are representatives A

of ( A, x) and B of (B, x) such that A ⊂ B.

3. Let ( A, x) ⊂ (Y, x) and (B, x) ⊂ (Y, x) be germs. Then we define ( A, x) ∩ (B, x) to be

the germ of A ∩ B at x for any representative A of ( A, x) and B of (B, x). We define
( A, x) ∪ (B, x) to be the germ of A ∪ B at x for any representative A of ( A, x) and B of
(B, x).

4. A germ of an complex space (X, x) is a germ at x of a locally analytic subset of Cn .

In particular, if f ∈ On,x then we define the germ of analytic hypersurface (V ( f ), x) as
follows. Consider an open neighborhood U of x on which f converges and the analytic subset

V ( f ) = {p ∈ U : f (p) = 0}

of U. The germ (V ( f ), x) is called the zero set of f .

Moreover, if I = ( f 1, . . . , f s ) ⊂ On,x is an ideal then the germ of complex space of I is

(V (I), x) = ∩i=1
s
(V ( f i ), x).

This definition is independent of the generators of the ideal I.

Since the ring On,x is Noetherian, any germ of analytic space is a germ of type (V (I), x) for
some ideal I of On,x .
Let (X, x) be a germ of a complex space. The ideal of (X, x)

I (X, x) = { f ∈ On,x : (X, x) ⊂ (V ( f ), x)}.

The inclusion here is an inclusion of germs. So f ∈ I (X, x) if there exists a representative X of

(X, x) and an open neighborhood U of x such that X is an analytic subset of U, f converges on
U, and its restriction to X is the zero function.
Remark 1.3.6. When we write germ of analytic space, we are usually referring to germ of set
(X, x) together with its local algebra, On,x /I, where I is the ideal which defines X.
A germ of analytic space is a germ of complex space (X, 0) defined by a radical ideal. That
is, the local algebra On /I is reduced, where I is the ideal which defines (X, 0)
For instance, let I = (x 2 ) ⊂ C{x}. Then (V (I), 0) is the point 0. In this case, I defines a
germ of a complex space O X,x = C{x}/I.
Associated to such a germ of a complex space (X, x) is its reduction (Xred, x), with local
ring O Xr ed,x the reduction of O X,x , obtained by dividing out the nilpotent elements.
Example 1.3.7. Let X = V (xy, y 2 ) ⊂ C2 . Then (X, 0) is the analytic complex space with
ring O2 /(xy, y 2 ). Here X is the union of the x-axis with the fat point with support {0} (see
[19, Example 1.38.2]. We associate to it the (reduced) analytic variety (X, 0) with local ring
(O2 /(x y, y 2 ))/(y) = O2 /(y)
Next, we see how the definitions the ideal of an analytic space and the zero set of an ideal
relate.
Proposition 1.3.8. Let (V, x) and (W, x) be germ of analytic spaces in (Cn, x), ( A, x) be a germ
of set in (Cn, x) and let I and J be ideals of On,x .
1.3. GERM OF COMPLEX ANALYTIC SETS 11

1. I ( A, x) is a radical ideal.
2. (V, x) ⊂ (W, x) ⇒ I (W, x) ⊂ I (W, x).
3. I ⊂ J ⇒ (V (J), x) ⊂ (V (I), x).
4. (V (I J), x) = (V (I ∩ J), x) = (V (I), x) ∪ (V (J), x). In particular, finite unions of analytic
spaces are analytic spaces.
5. (V (I + J), x) = (V (I), x) ∩ (V (J), x). In particular, finite intersections of analytic spaces
are analytic spaces.
6. I ((V, x) ∪ (W, x)) = I (V, x) ∩ I (W, x).
It is not difficult to see that for any ideal I ⊂ On , I ⊂ I (V (I)). The other inclusion is not
true. However, there
√ is a very important result called the Nullstellensatz theorem which says
that I (V (I)) = I. The prove can be found at [6, Theorem 3.4.4] and [19, Theorem 1.72], for
instance.
Definition 1.3.9. Let (X, x) be a germ of an analytic space. We say that (X, x) is irreducible if
from (X, x) = (X1, x) ∪ (X2, x), with (X1, x) and (X2, x) germs of analytic spaces it follows that
either (X, x) = (X1, x) or (X, x) = (X2, x).
Proposition 1.3.10. Let (X, x) be a germ of an analytic space.
1. (X, x) is irreducible if and only if I (X, x) is a prime ideal.
2. There is a, up to permutation, unique decomposition (X, x) = (X1, x) ∪ÂůÂůÂů∪ (Xr , x),
with (Xi, x) irreducible and (X j , x) 1 (Xi, x) for i , j. This is called the irreducible
decomposition of (X, x). The (Xi, x) are called irreducible components of (X, x).
Proof. 1. We assume that (X, x) is irreducible. Let f , g ∈ On,x such that f g ∈ I (X ). Then
(X, x) = (V (I (X )), x) ⊃ (V ( f g), x) = (V ( f ), x) ∪ (V (g), x).
Hence, by the hypothesis, (X, x) = (X, x) ∩ (V ( f ), x) or (X, x) = (X, x) ∩ (V (g), x). So
either (X, x) ⊂ (V ( f ), x) or (X, x) =⊂ (V (g), x), which means that either f ∈ I (X ) or
g ∈ I (X ).
On the other hand, we assume that I (X, x) is a prime ideal. If there exist germs of analytic
sets (X1, x) and (X2, x) such that (X, x) = (X1, x) ∪ (X2, x) then, by the Nullstellensatz
theorem
I (X ) = I (V (I (X ))) = I (X1, x) ∩ I (X2, x).
Since I (X, x) is a prime ideal, either I (X, x) = I (X1, x) or I (X, x) = I (X2, x).
Therefore, either (X, x) = (X1, x) or (X, x) = (X2, x).
2. Let I (X, x) = I1 ∩ · · · ∩ Ir be an irredundant primary decomposition of I (X, x). Then
(X, x) = V (I (X, x), x) = (V (I1 ), x) ∪ · · · ∪ (V (Ir ), x)
and, for i = 1, . . . , r, (Xi, x) := (V (Ii, x), x) is irreducible by item 1. Moreover this
decomposition is unique up to permutation by the uniqueness of the primary decomposition
of ideals and (X j , x) 1 (Xi, x) for i , j because the primary decomposition of the ideal
was no redundant.

12 CHAPTER 1. PRELIMINARY RESULTS

The following table summarizes the relationships between radical ideas and germs of analytic
spaces.
Ideals analytic spaces
radical ideals analytic spaces
I −→ V (I)
I (V ) ←− V
inclusion of ideals inclusion of analytic spaces
I⊂J −→ V (I) ⊃ V (J)
I (V ) ⊃ I (W ) ←− V ⊂W
addition of ideals intersection of analytic spaces
I+J −→ V (I) ∩ V (J)
product of ideals union of analytic spaces
√ IJ −→ V (I) ∪ V (J)
I (V )I (W ) ←− V ∪W
intersection of ideals union of analytic spaces
I∩J −→ V (I) ∪ V (J)
I (V ) ∩ I (W ) ←− V ∪W
prime ideals ←→ irreducible analytic spaces
minimal decomposition minimal decomposition
I = P1 ∩ P2 · · · ∩ Pm −→ V (I) = V (P1 ) ∪ V (P2 ) · · · ∪ V (Pm )
I (V ) = I (V1 ) ∩ I (V2 ) ∩ · · · ∩ I (Vm ) ←− V = V1 ∪ V2 ∪ · · · ∪ Vm

Exercises
1. Prove Proposition 1.3.8.

1.3.1 The local algebra of a germ of analytic space

Definition 1.3.11. Let (X, p) and (Y, q) be two germs of topological spaces. A germ of a
continuous map f : (X, p) → (Y, q) is defined as an equivalence class of maps f : U → W ,
with f (p) = q, and where U and W are representatives of (X, p) and (Y, q) respectively. Two
such maps f 1 : U1 → W and f 2 : U2 → W are called equivalent if they agree on an open
neighborhood of p contained in U1 ∩ U2 .
Let (X, x) ⊂ (Cn, x) and (Y, y) ⊂ (Cm, y) be germs of an analytic space. A germ of an
analytic map ϕ : (X, x) → (Y, y) is a germ of a map ϕ : (X, x) → (Y, y) such that some
representative is the restriction to X of an analytic function on an open neighborhood of x in
Cn . When Y = C, we say that ϕ is a germ of analytic function.
We denote by O X,x the set of the germ of analytic functions f : (X, x) → C. Germs of
analytic functions can be added and multiplied, so O X,x has the structure of a commutative
C-algebra. In fact, it is a local algebra whose maximal ideal, m X,x , is the subset of germs
h ∈ O X,x such that h(x) = 0. This is called the local algebra of (X, x).
Lemma 1.3.12. 1. Let (X, x) ⊂ (Cn, x) be a germ of an analytic space, and I (X, x) be the
ideal of X. Then O X,x = On,x /I (X, x).
2. Let (X, x) ⊂ (Kn, x) be a germ of a submanifold. Then O X,x ≈ K{x 1, ..., x k } for some k.
1.3. GERM OF COMPLEX ANALYTIC SETS 13

3. Let ϕ : (X, x) → (Y, y) be a germ of an analytic mapping. Then by composition ϕ induces

a map of K-algebras
ϕ∗ : OY,y → O X,x
f 7→ f ◦ ϕ
We say that a germ of an analytic map ϕ : (X, x) → (Y, y)) is a isomorphism if it has a
two-side inverse which is also a germ of a analytic map. If there exists such a isomorphism,
we say that the germs of analytic spaces (X, x) and (Y, y) are isomorphic. Our goal now is
to show that two germs of analytic spaces are isomorphic if and only if their local algebra are
isomorphic.
Theorem 1.3.13. Let (X, x) ⊂ (Cn, x) and (Y, y) ⊂ (Cm, y) be germs of analytic spaces. Let
α : O(Y,y) → O(X,x) be a C-algebra homomorphism. Then there exists a unique germ of an
analytic mapping ϕ : (X, x) → (Y, y) with ϕ∗ = α.
Proof. Without loss of generality, we may assume that x = 0 and y = 0.
By Lemma 1.1.9 α(m(Y,0) ) ⊂ m(X,0) . Therefore, α(m(Y,0) k ) ⊂ m(X,0)
k , for all k > 0.
Take w1, . . . , wm generators of the maximal ideal of Om . Let wi be the class of wi in
Om /I (Y, 0) and f i = α(wi ), i = 1, . . . , m. We define

ϕ = ( f 1, . . . , f m ) : (X, 0) → (Cm, 0).

Hence we get a map ϕ∗ : Om → O X,0 .

α
1. ϕ∗ = α̃, where α̃ is the composition Om → O(Y,0) → O(X,0) . In fact, for each i = 1, . . . , k,
ϕ∗ (wi ) = α̃(wi ). Since they are both C-algebra homomorphism it follows that ϕ∗ (g) =
α̃(g) for all polynomial g ∈ C[w1, . . . , wm ]. And given g ∈ Om , we can write

g = gk + g0k ,

where gk is a polynomial of degree smaller then k and g0k ∈ mmk . Then

ϕ∗ (g) − α̃(g) = ϕ∗ (gk ‘) − α̃(g0k ) ∈ m(X,0)

k
.

Since it holds for any k > 0, ϕ∗ (g) − α̃(g)) ∈ ∩ k m(X,0)

k = {0}, by the Krull‘s intersection
theorem.

2. ϕ(X, 0) ⊂ (Y, 0). In fact, since (Y, 0) = (V (I (Y, 0)), 0), it is enough to observe that for
all g ∈ I (Y, 0) the map g ◦ ϕ : (X, 0) → C is the zero map because I (Y, 0) is in the
kernel of ϕ∗ by the construction of α̃.

Corollary 1.3.14. Two germs of analytic spaces (X, x) and (Y, y) are isomorphic if and only if
O(X,x) and O(Y,y) are isomorphic.
Proof. The Nullstellensatz says that there is a 1-1 correspondence between reduced analytic
algebras and germs of analytic spaces. The previous theorem implies that the isomorphism
classes of reduced analytic algebras correspond to isomorphism classes of germs of analytic
spaces 
14 CHAPTER 1. PRELIMINARY RESULTS

1.3.2 Dimension
Definition 1.3.15. Let (X, x) be a germ of a complex space and (O X,x, m) be its local ring.

1. The Krull dimension of (X, x) is the Krull dimension of its local ring, that is, the maximal
length k of chains of prime ideals

p0 ( p1 ( pk

in O X,x .

2. The Chevalley dimension of (X, x) is the least number of generators for an m-primary
ideal of O(X,x) .

3. The Weierstrass dimension of (X, x) is the least number k, such that there exists a Noether
normalization O k ⊂ O(X,x) of (X, x) (see [6, Corollary 3.3.19] for the definition of Noether
normalization).

Remark 1.3.16. All the three dimensions in the previous definition coincide for a germ of
analytic variety. See [6, Section 4.1].

By taking zero sets, it follows that the Krull dimension of (X, x) is the supreme of the length
n of a chain of irreducible subvarieties of (X, x):

0 ( (X1, x) ( ... ( (X n, x) ⊆ (X, x).

Example 1.3.17. For the case O X,x = On , consider the following chain of prime ideals:

(0) ( (x 1 ) ( (x 1, x 2 ) ( ... ( (x 1, ..., x n )

from which it follows that the Krull dimension of (Cn, 0) is at least n. It needs proof, and is in
fact nontrivial that the dimension of (Cn, 0) is indeed equal to n.

Theorem 1.3.18. Suppose (X, x) is a germ of a complex space. Let (X, x) = (X1, x)∪· · ·∪(Xr , x)
be an irreducible decomposition of (X, x). Then

dim(X, x) = max{dim(Xi, x), i = 1, ..., r }.

1.3.3 Finite germs of maps

In general, it is not possible to talk about the image of a germ of a map. For instance, let
f : R2 → R2 the map defined by f (x, y) = (x, x y) and consider its germ at the origin. For
each real positive number r, the sets Xr = {(x, y) ∈ R2 : −r ≤ x ≤ r, −r ≤ y ≤ r } and
Yr = {(x, y) ∈ R2 : −r ≤ x − y ≤ r } may be taken as representatives of (R2, 0) but, for any open
set U of R2 , U ∩ f (Xr ) , U ∩ f (Yr ). For a special kind of maps, the finite maps, this problem
does not occur.

Definition 1.3.19. Let f : X → Y be a continuous map between topological spaces.

1. f is closed if the image f ( A) ⊂ Y is closed for all closed subspaces A ⊂ X.

1.3. GERM OF COMPLEX ANALYTIC SETS 15

2. f is quasi-finite if for all p ∈ Y the fiber f −1 (p) consists of a finite number of points.

3. f is called finite if it is both closed and quasi-finite.

Example 1.3.20. [6, Lemma 3.4.10] Let

P = yr + a1 yr−1 + · · · + ar−1 y + ar

be a polynomial, with coefficients ai ∈ O k . Consider an open neighborhood U ⊂ C k of 0 such

that the power series a1, . . . , ar converge. Define:

X := {P = 0} ∩ (U × C).

Then the canonical projection π : X → U is finite.

Definition 1.3.21. Let (X, p) and (Y, q) be two germs of topological spaces. A germ of
continuous map f : (X, p) → (Y, q) is called finite if it has a finite representative.

Lemma 1.3.22. Let X and Y be topological spaces and f : X → Y be a closed map. Let
p ∈ X and q = f (p). Assume that f −1 (q) = {p}. Let the A, B ⊂ X with p ∈ A ∩ B such that
( A, p) = (B, p). Then ( f ( A), q) = ( f (B), q).

Proof. Let W be an open set in X such that p ∈ W and A ∩ W = B ∩ W . Since f is closed then
the set f (X\W ) is a closed subset of Y and T := Y \ f (X\W ) is an open subset of Y . We will
show that q ∈ T and A ∩ T = B ∩ T.
Since f is continuous, the set U = f −1 (T ) is an open subset of X. Moreover

1. U = f −1 ( f (U)),

2. U ⊂ W and

3. p ∈ U.

Besides, f (U ∩ A) = T ∩ f ( A). In fact,

x ∈ U ∩ A ⇒ f (x) ∈ f (U) ∩ f ( A) ⊂ T ∩ f ( A),

on the other hand,

a ∈ T ∩ f ( A) ⇒ a = f (b), with b ∈ A ∩ f −1 (T ) = A ∩ U ⇒ a ∈ f (U ∩ A).

Similarly, f (U ∩ B) = T ∩ f (B). And U ∩ A = U ∩ B. Hence,

T ∩ f (B) = f (U ∩ B) = f (U ∩ A) = T ∩ f ( A).


We remark that if f : (X, p) → (Y, q) is a finite map germ we can always choose a finite
representative f : X → Y such that the hypothesis f −1 ( f (p)) = {p} is satisfied. In fact, since
for any representative f −1 ( f (p)) is a finite number of points, just reduce the neighborhoods if
necessary. With this, we are ready to define the image of a finite map germ.
16 CHAPTER 1. PRELIMINARY RESULTS

Definition 1.3.23. Let f : (X, p) → (Y, q) be a finite germ of a continuous map. The image
of f is defined by Im( f ) := ( f (X ), q), where f : X → Y is a finite representative such that
f −1 ( f (p)) = {p}.
When f is finite, it is called surjective if Im( f ) = (Y, p).

The following result is known as Remmert’s proper mapping theorem. For a proof see [27,
Chapter V].

Theorem 1.3.24. The image ( f (X ), y) of any finite analytic map-germ f : (X, x) → (Y, y) is
analytic.

Theorem 1.3.25. [36, Lemma D3] Let f : (X, x) → (Y, y0 ) be a finite surjective analytic
map-germ with (Y, y0 ) irreducible. There exists a hypersurface (D, y) ⊂ (Y, y0 ) such that for
all small enough representatives Y of (Y, y0 ) and D of (D, y0 ), the fibre f −1 (y) has constant
cardinality for all y ∈ D.

The constant cardinality from the preview theorem is called the degree of the map f and
denoted deg( f ).

Theorem 1.3.26. Let f : (X, x) → (Cd, 0) be a finite surjective map-germ. If (X, x) is Cohen-
Macaulay then
OX
deg( f ) = dimC ,
( f 1, . . . , f d )
where ( f 1, . . . , f d ) is the ideal generated by the coordinates functions of f .

Proof. See, for instance, [36, Corollary D6]. 

1.3.4 The singular locus and the Jacobian Criterion

Definition 1.3.27. [36, Definition D1] Let (X, x) be a germ of complex space. We say that (X, x)
is smooth or that x is a regular point of X if (X, x) is isomorphic (as a complex space-germ) to
(Cd, 0), for some d. Otherwise, we say that (X, x) is singular.

The above definition of regular point means that O X,x is isomorphic to Od so we necessarily
must have d = dim(X, x 0 ). We fix representatives of X and the functions f i on some open subset
U ⊂ Cn such that X is given by the vanishing of f i on U. Then, it makes sense to consider the
set Σ of points x ∈ X such that X is not regular at x. The set-germ (Σ, x 0 ) is called the singular
locus of (X, x 0 ) and is denoted by Sing(X, x 0 ).
We can characterize the regular points by means of the Jacobian Criterion.

Theorem 1.3.28. Let (X, 0) ⊂ (Cn, 0) be a germ of complex space, and let the ideal of (X, 0)
be generated by f 1, . . . , f s ∈ C{x 1, . . . , x n }. We denote by rank0 ( f 1, . . . , f s ) the rank of the
Jacobian matrix
∂ fi
!
(0) .
∂xj 1≤i≤s, 1≤ j ≤n
Then
edim(X, 0) + rank0 ( f 1, . . . , f s ) = n,
where edim(X, x) := edim(O(X,x) ).
1.4. ISOLATED COMPLETE INTERSECTION SIGULARITIES 17

Remark 1.3.29. Let X be an analytic subset of an open subset U of Cn , say locally defined by
holomorphic functions f 1, ..., f s on U. Suppose that for all x ∈ X, the germ (X, x) has dimension
n − c. It follows directly from the Jacobian Criterion that the singular locus of X is contained
in the zero set of the c-minors of the Jacobian matrix, which is an analytic set. In fact, it is true
that these sets are equal but the proof depends on the Coherence Theorem of Oka-Cartan, the
which is very advanced for the purpose of these notes.

Example 1.3.30. Let (X, 0) ⊂ (C4, 0) be the germ of analytic space generated by xz − y 2, yw −
z 2, xw − yz. The dimension of (X, 0) is two. Therefore,
 z −2y x 0
Sing(X, 0) = V . (xz − y 2, yw − z 2, xw − yz) + I2  0 w −2z y  +/
*
, w −z −y x  -
= V (xz − y 2, yw − z 2, −x 2, 2y 2 + xz, −yz − xw, −x y, 4yz − xw, −2z 2, −2xy,
z 2 − 2yw, −2y 2, −zw, −xz, −yz, 2z 2 + yw, −2zw, y 2 − 2xz, −w 2, −yz − xw, yw)
= V (x 2, x y, y 2 − 2xz, xz, yz, 2z 2 + yw, xw, yw, zw, w 2 )
= {0}

1.4 Isolated Complete intersection Sigularities

Finally, we are ready for the definition of isolated complete intersection singularity (ICIS).
The complete intersections are the natural generalization of hypersurfaces. That is, a complete
intersection is a germ of analytic space defined by the zero set of a map germ from Cn to C p
which keeps the codimension: the codimension of the germ of analytic space is equal to the
codimension of {0} in C p .

Definition 1.4.1. Let (X, 0) ⊂ (Cn, 0) be a germ complex space defined by an ideal I in On . Let
k be the minimal number of generators of I. Then (X, 0) is called a complete intersection if the
dimension of (X, 0) is n − k.

Example 1.4.2. We consider (X, 0), the coordinate axes in C3 , given by the zero set of the
ideal I = (x y, xz, yz). It is easy to see (using for example the Chevalley dimension), that the
dimension of (X, 0) is one. Therefore, (X, 0) is not a complete intersection.
We remark here that (X, 0) is a determinantal variety. The determinantal varieties are the
natural generalization of complete intersection: they are germs of analytic spaces defined by the
inverse image of the set of the matrices with a fixed rank by an analytic map germ from Cn to
the set of the matrices with size m × k with the additional hypothesis that the codimension is the
expected one. Here, I is generated by the minors of size two of the matrix

x 0 z
" #
.
0 y z

We say that a germ of analytic space is Cohen-Macaulay if its local ring is Cohen-Macaulay.

Theorem 1.4.3. If (X, 0) ⊂ (Cn, 0) is an ICIS then it is Cohen-Macaulay.

For a prove of this theorem see [19, Corollary B.8.10].

18 CHAPTER 1. PRELIMINARY RESULTS

Definition 1.4.4. An isolated complete intersection singularity (ICIS) is a complete intersection

(X, 0) such that Sing(X, 0) = (0, 0).

Example 1.4.5. 1. (X, 0) = (V (xy − z 2, x 2 + y 2 + z 2 ), 0) is an ICIS in (C3, 0). In fact,

dim(X, 0) = 1 and
!!
2 2 2 2 y x −2z
Sing(X,0) = V (xy − z , x + y + z ) + I2
2x 2y 2z
= V (x y − z 2, x 2 + y 2 + z 2, y 2 − x 2, xz + 2yz, yz + 2xz)
= {0}

here I2 (M) denotes the ideal generate by the minors of size two of the matrix M.

2. For germs (Cn, 0) → (C p, 0) with n ≤ p of corank 1 and with finite left-right codimension,
the germs of multiple points, D k ( f ), are ICIS (see [36, Chapter 9]).
!
x y z
3. If we consider the matrix M = and I the ideal generated by the minors of size
y z w
two of M. Then (V (I), 0) is an isolated singularity in (C4, 0) which is not a complete
intersection. In fact it has dimension 2 but the radical ideal I can not be generated by 2
elements.

Corollary 1.4.6. If (X, 0) is an ICIS with dimension greater than or equal to two then it is
reduced.

Proof. In fact, we will show that (X, 0) satisfies Serre’s R0 and S1 conditions. The condition
R0 is satisfied if the singular set of (X, 0) has codimension greater then or equal to one. The
condition S1 is satisfied if the germ is Cohen-Macaulay. 
Chapter 2

The contact group

2.1 The K group

Consider K = R or C.
Definition 2.1.1. The contact group K is the set of pairs of germs of diffeomorphisms (h, H),
where h : (Kn, 0) → (Kn, 0), H : (Kn × K p, 0) → (Kn × K p, 0) such that π1 ◦ H = h,
(π2 ◦ H)(x, 0) = 0 where π1 and π2 are the projections into Kn and K p, respectively.
Notice that H (x, y) = (h(x), H2 (x, y)), H2 (x, 0) = 0.
The set of pairs (h, H) ∈ K , such that h is the identity IKn form a subgroup of K , usually
denoted by C .
Definition 2.1.2. Let f , g ∈ mn On,p . We say that f and g are contact equivalent, f ∼K g, if
there is a pair (h, H) ∈ K such that H (x, f (x)) = (h(x), g(h(x)).
Remark 2.1.3. Notice that if f ∼K g, then the diffeomorphism H : (Kn ×K p, 0) → (Kn ×K p, 0)
sends graph( f ) into graph(g), leaving Kn × {0} invariant. This geometric viewpoint of contact
equivalence was extended by Montaldi [37] as follows: two pairs of germs of submanifolds of
Rm have the same contact type if there is a germ of diffeomorphism of Rm taking one pair to
the other. Moreover, he proved in [37], that the contact type of a pair of germs of manifolds is
completely characterized by the K -equivalence class of a convenient map. This result is one
the fundamental pieces of the applications of singularity theory to differential geometry (see
Bruce and Giblin [2] and Izumiya, Romero-Fuster, Ruas and Tari, [23]).
The tangent space and the extended tangent space of K -equivalence are, respectively

TK f = t f (mn θ n ) + f ∗ (m p )θ f

TKe f = t f (θ n ) + f ∗ (m p )θ f

mn θ f θf
We also define K − codim( f ) = dimK and Ke − codim( f ) = dimK .
TK f TKe f
The Ke -codimension of f is also known as its Tjurina number and denoted by τ( f ).
The following result was first proved by Mather in [33].

19
20 CHAPTER 2. THE CONTACT GROUP

Proposition 2.1.4 (Gibson [18], Proposition 2.2, Mond and Nuño-Ballesteros [36], Section 4.4).
The following statements are equivalent.
(1) Two map-germs f , g ∈ mn On,p are K -equivalent.
(2) There exists a germ of diffeomorphism h : (Kn, 0) → (Kn, 0) such that
h∗ f ∗ (m p )On = g ∗ (m p )On .

The local algebra we introduce now is an useful invariant of K -equivalence. For a given
map-germ f : (Kn, 0) → (K p, 0) we define the local algebra of f as
On
Q( f ) = .
f ∗ (m p )On
It follows from the previous proposition that the isomorphism class of Q( f ) is a K -
invariant. Furthermore, it is a complete invariant of K -equivalence for germs f with finite
K -codimension. More precisely, we have
Theorem 2.1.1. If f and g are map-germs with finite K -codimension it follows that f ∼K g
if and only if the local algebras Q( f ) and Q(g) are isomorphic.
Remark 2.1.5. For complex analytic germs the hypothesis of finite K -codimension in Theorem
2.1.1 is not needed.
Example 2.1.6. Let F : (Kn, 0) → (K p, 0) be a germ of rank r. Then, up to A -equivalence,
we can take F in the normal form F (x, y) = (x, f (x, y)), x ∈ Kr , y ∈ Kn−r , with f : (Kn, 0) →
(K p−r , 0) and j 1 f (0, 0) ≡ 0. Let f 0 : (Kn−r , 0) → (K p−r , 0) be the rank zero germ f 0 (y) =
f (0, y). Then Q(F) = Q( f 0 ).
If K -codim ( f 0 ) < ∞ and Q(F)  Q( f 0 ) it follows that F is K -equivalent to the trivial
unfolding F0 (x, y) = (x, f 0 (y)) of f 0 .
As we shall see in the next section, germs f ∈ mn On,p of finite K -codimension are finitely
K -determined, and in this case K ( f ) = K (z), where z = j k f (0) for some k.
Now, for each positive integer k, we set
On
Qk ( f ) = .
f ∗ (m p )On + mnk+1
Q k ( f ) is the local algebra of z = j k f (0). We can also write Q k ( f ) = Q(z).
It is not hard to show that z ∼K k z0 if and only if Q k (z) and Q k (z0 ) are isomorphic.

Exercises
1. Go to the first part of the mini-course (2021) and recall the definition of right-left equiv-
alence (A -equivalence) and show that if f , g : (Kn, 0) → (K p, 0) are A -equivalent then
they are K -equivalent.This was a proposed exercise of first part.
2. Let f ∈ mn On,p . Prove that (see Lemma 4.1 in [36])
dΦt · f
{ |t=0 : Φt ∈ K is smooth, Φ0 = Id} = t f (mn θ n ) + f ∗ m p θ f = TK f .
dt
 Part II

Lecture 2

21
Chapter 3

K -Determinacy

Let G =R, L , A , C or K be one of Mather’s groups and f : (Kn, 0) → (K p, 0) a smooth

map-germ.
Definition 3.0.1. The map-germ f is k-G -determined if any smooth map-germ g : (Kn, 0) →
(K p, 0) with j k g = j k f is G -equivalent to f . If f is k-G -determined for some k < ∞, then it is
finitely-G -determined, and the least such k is the degree of determinacy.

3.0.1 Infinitesimal criteria of finite K -determinacy

The following result has a similar statement for all Mather’s groups. Its proof can be found in
[36] and [47].
Theorem 3.0.1. Let f ∈ mn On,p . The following are equivalent:

(a) f is finitely-K -determined,

(b) for some k, TK f ⊃ mnk On,p ,

mn On,p
(c) K − codim( f ) = dimK < ∞,
TK f
On,p
(d) Ke − codim( f ) = dimK < ∞,
TKe f
More precisely,

(i) If f is r-K -determined, TK f ⊃ mnr+1 On,p .

(ii) If TK f ⊃ mnr+1 On,p then f is r + 1-K -determined.

(iii) If K − codim( f ) = d < ∞, then TK f ⊃ mnd+1 On,p .

Corollary 3.0.2. Let f ∈ mn On,p . Suppose

TK f + mnr+2 On,p ⊃ mnr+1 On,p,

then f is r + 1-K -determined.

23
24 CHAPTER 3. K -DETERMINACY

3.0.2 Geometric criterion of finite K -determinacy

Let f : N → P be a smooth map between manifolds. We recall that the critical set C( f ) is
by definition the set of points x ∈ N such that the derivative of f at x is not surjective. The
following result can be found in [36].

Theorem 3.0.3 (Mather-Gaffney Criterion). A holomorphic map-germ f : (Cn, 0) → (C p, 0)

has finite K -codimension if and only if there is a representative which is finite-to-one on its
critical set.

Exercises
1. Show that f (x 1, x 2 ) = (x 21, x 22 ) and g(x 1, x 2 ) = (x 21 − x 22, x 1 x 2 ) are both 2-K -determined.

2. Show that f (x 1, x 2 ) = (x 21 ± x 2k , x 1 x 2 ) is k-K -determined.

3.1 Classification of stable germs

Let f ∈ mn On,p such that Ke − codim( f ) < ∞. Let

On,p
NKe f =
TKe f

be the normal space. Consider φi ∈ On,p, i = 1, . . . , r, whose images in NKe f together with
those of the ∂∂y j span NKe f as K-vector space.

Theorem 3.1.1 ([18]). Let F : (Kn × Kr , 0) → (K p × Kr , 0) be the map-germ given by

Xr
F (x, u) = ( f (x) + ui φi (x), u).
i=1

Then f is a stable germ.

Example 3.1.1. Let f ∈ mn be given by f (x 1, . . . , x n ) = x 1k+1 ± x 22 ± . . . ± x 2n (Ak -singularity).

Then TKe f is the ideal of On generated by {x 1k , x 2, . . . , x n } and {1, x 1, x 21, . . . , x 1k−1 } is a basis
for the K-vector space NKe f . Therefore,

F (x, u) = (x 1k+1 ± x 22 ± . . . ± x 2n + u1 x 1 + · · · + u k−1 x 1k−1, u)

is a stable map-germ. For example:

• n = 1, k = 2, F (x, u) = (x 3 + ux, u) : cusp singularity.

• n = 1, k = 3, F (x, u) = (x 4 + u1 x + u2 x 2, u1, u2 ) : swallowtail singularity.

3.2. COMPLETE TRANSVERSAL METHOD 25

3.2 Complete transversal method

The references for this sections are [3], [4].
Proposition 3.2.1. Let G be a Lie group with Lie algebra g acting smoothly on an affine space
A and let W be a subspace of VA such that
g(x + w) = gx,
for all x ∈ A and w ∈ W . Then
1. For any x ∈ A we have
x + gx ∩ W ⊂ Gx ∩ {x + W }.
2. If x 0 ∈ A and T is a vector subspace of W satisfying
W ⊂ T + gx 0,
then for any w ∈ W there exist g ∈ G and t ∈ T such that g(x 0 + w) = x 0 + t.
For each k, let H k be the vector subspace of J k (n, p) of homogeneous k-jets of degree k,
that is,
mnk On,p
H k = k+1 .
mn On,p
Let G be any one of Mather’s groups R, L , A , C or K . Consider G k as the normal
subgroup of G consisting of those germs whose k-jet is equal to that of the identity.
A subgroup H of G is called a Lie subgroup of G1 if for each k the k-jet group H (k) is a
Lie subgroup of G1(k) .
Proposition 3.2.2. Let H be a Lie subgroup of G1 and let σ be a k-jet in J k (n, p). If T ⊂ H k+1
is a vector subspace such that
H k+1 ⊂ T + TH (k+1) σ,
then for each (k + 1)-jet τ with j k τ = σ there exists t ∈ T such that τ is H (k+1) -equivalent to
σ + τ.
Proof. We apply Proposition 3.2.1 with A = J k+1 (n, p), W = H k+1 and G = H (k+1) . Since
H (k+1) ⊂ G1(k+1), we have ϕ(h) = h, for all h ∈ H k+1 and ϕ ∈ H (k+1) . But this implies that
η(σ + h) = ησ, for all h ∈ H k+1 and η ∈ g, which gives the necessary condition.

Definition 3.2.3. A subspace T ⊂ H k+1 satisfying the conditions of above proposition is called
a complete transversal for σ ∈ J k (n, p).
Let f ∈ mn On,p . We recall that J ( f ) = On { ∂∂xf1 , . . . , ∂∂xfn } ⊂ On,p is called Jacobian submod-
ule of f . Since we are interested in the K group, we shall consider the following result.
Proposition 3.2.4. The tangent space to the K k -orbit of a k-jet f ∈ J k (n, p) is given as k-jets
of elements of
mn J ( f ) + On . f ∗ m p {e1, . . . , e p }.
For the K1k -orbit of the k-jet f ∈ J k (n, p), the tangent space is given as k-jets of elements of

mn2 J ( f ) + mn . f ∗ m p {e1, . . . , e p }.
26 CHAPTER 3. K -DETERMINACY

Example 3.2.5. Consider the case of map-germs R2 → R2 with zero 1-jet. The classification
of homogeneous 2-jets reduces to that of pencils of binary quadratic forms. Using a change of
coordinates in source we can reduce the first form of the pair to x 21 ± x 22, x 21 or 0. Now the other
change of coordinates allows us to subtract multiples of the first entry from the second. Using
this reduce any pencil to one of the forms

(x 21, x 22 ), (x 21 − x 22, x 1 x 2 ), (x 21, x 1 x 2 ), (x 21 ± x 22, 0), (x 21, 0), (0, 0).

We have:

1. (x 21, x 22 ) and (x 21 − x 22, x 1 x 2 ) are both 2-K -determined (exercise above).

2. Consider 2-jets σ = (x 21, x 1 x 2 ). Thinking of σ as a k − 1-jet we find that a complete

transversal is given by (x 21 + ax 2k , x 1 x 2 ). In fact, since

TK1 σ = m22 {(2x 1, x 2 ), (0, x 1 )} + m2 {(x 21, 0), (x 1 x 2, 0), (0, x 21 ), (0, x 1 x 2 )},

we have that H k ⊂ R{(x 2k , 0)} + TK1(k) σ. If a , 0 we obtain (x 21 ± x 2k , x 1 x 2 ) which is

k-K -determined (exercise above).

3. Consider 2-jets σ = (x 21 ± x 22, 0). In the minus case we can use the alternative normal form
(x 1 x 2, 0). We have:

TK1 σ = m22 {(x 1, 0), (x 2, 0)} + m2 {(x 1 x 2, 0), (0, x 1 x 2 )},

so a complete transversal is given by (x 1 x 2, ax 1k +bx 2k ). If ab , 0 we reduce to (x 1 x 2, ±x 1k ±

x 2k ), which are k-K -determined. If one of a and b is zero, say b, we reduce to (x 1 x 2, x 1k ),
which regarded as an l − 1-jet has complete transversal (x 1 x 2, x 1k + bx 2l ). We reduce to the
normal forms (x 1 x 2, x 1k ± x 2l ), which is (l + 1)-determined.

Exercises
1. Let f ∈ mn On,p . Suppose

TK1 f + mnk+2 On,p ⊃ mnk+1 On,p,

then f is k-K -determined. Hint: regarding f as a k-jet, this shows that the k + 1-
transversal of f is empty, and then all subsequent transversal are also empty.

2. Fill in all the details of the complete transversal in the case (x 1 x 2, x 1k ) in Example 3.2.5.

3. Investigate the germs emerging from (x 21 + x 22, 0) in Example 3.2.5.

Chapter 4

K -Versal unfoldings of map germs (real

and complex)

4.1 Introduction
Since the beginning of singularity theory it has been clear that in order to understand a singu-
larity you have to understand what happens when you deform it into less degenerate types of
singularities. For instance, when looking at a bent wire from the tangent direction at a point
of the wire where it does not have 0 torsion you see a cusp. If you move your head slightly to
the left and to the right, from one side you will see a regular piece of wire and from the other
side you will see a kind of loop. We have deformed the cusp and by seeing what happens near
the cusp we have understood how this singularity appears. In a certain sense we need a family
(a 1-paramater family in this case) of views in order to grasp the full nature of the singularity.
In this part of the lecture we will give the definition of unfolding and deformation of a smooth
function germ f : (Kn, 0) → (K p, 0). When working with map germs f : (Kn, 0) → (K p, 0),
p > 1 it is natural to consider A -equivalence (i.e. smooth changes of coordinates in source
and target) and the notion of unfolding is crucial. In this chapter we want to set up the theory
of deformations of germs under K -equivalence. This theory plays a fundamental role in the
classification of stable map-germs. Furthermore, K -equivalence is an important tool to study
ICIS, as we will see in the following chapter. We will follow the approach in [30] and [36].

4.2 Basic definitions and examples

We recall from the previous lecture, that in the analytic case, studying K -classes of mapgerms
is equivalent to studying isomorphism classes of the analytic spaces, then we define. Let
f : (Kn, 0) → (K p, 0) be a smooth function germ, where smooth means C ∞ when K = R or
holomorphic when K = C.
Definition 4.2.1. A r-parameter unfolding of f is a map-germ
F 0 : (Kr × Kn, 0) → (Kr × K p, 0)
of the form F 0 (u, x) = (u, F (u, x)), such that F (0, x) = f (x). If we denote F (u, x) by f u (x), the
above condition becomes f 0 = f . The map-germ
F : (Kr × Kn, 0) → (K p, 0)

27
28 CHAPTER 4. K -VERSAL UNFOLDINGS OF MAP GERMS (REAL AND COMPLEX)

is a r-parameter deformation of f .
In this chapter we want to set up the theory of deformations of germs under K −equivalence.
This theory plays a fundamental role in the classification of stable map germs, as we discuss in
section 4.4.
Definition 4.2.2. ii) Two r-parameter deformations F, G of f are K -equivalent if there exists
a germ of diffeomorphism
Φ : (Kr × Kn, 0) → (Kr × Kn, 0)
of the form Φ(u, x) = (u, ϕu (x)) where ϕ(0, x) = x (i.e. Φ is a deformation of the identity in
Kn ), such that
Φ∗ (G∗ (M p )) = F ∗ (M p ).
In this case we call Φ a K -equivalence of deformations. It is clear that this relation implies that
F, G are K -equivalent as germs, notice however that the change of coordinates at the source
respects the product structure on Kr × Kn .
iii) A deformation F is called K -trivial if it is equivalent to the constant deformation
F̃ (x, u) = f (x).
iv) A map-germ is called K -stable if any deformation of it is trivial.
Example 4.2.3. Consider the germ f : (K, 0) → (K, 0) given by f (x) = x and the deformations
F (u, x) = x and G(u, x) = x + ux 2 . Taking the diffeomorphism Φ(u, x) = (u, x + ux 2 ) we get
Φ∗ (hFi) = (hGi), where hFi denotes the ideal generated by the coordinate functions F1, . . . , Fp of
F. It follows that F and G are K -equivalent. Furthermore, since F is the constant deformation,
this means that G is K -trivial.
In fact, given any deformation H (u, x), since H (0, x) = x, by considering the diffeomorphism
Φ(u, x) = (u, H (x, u)) we see that H is trivial. We have shown that f (x) = x is stable.
θ( f )
Remark 4.2.4. A function germ f : (Kn, 0) → (K p, 0) is Ke -stable if and only if dimK T K e( f )
=
0, and, thus, if and only if d f 0 is surjective, i.e. the function is regular.
Deformations allow us to see what happens around a singularity, but in order to understand
the singularity completely we want to know what are all the possible phenomena that appear
around it. The idea of a versal deformation is that it captures all the possible less degenerate
singularities into which a certain singularity can be deformed.
Definition 4.2.5. i) Let F : (Kr × Kn, 0) → (K p, 0) be a deformation of a map-germ f and let
h : (Ks, 0) → (Kr , 0) be a map-germ. The pull-back of F by h is the deformation

h∗ F : (Kn × Ks, 0) → (K p, 0)

given by
h∗ F (v, x) = F (h(v), x).
The map-germ h is called the base change map.
ii) An r-parameter deformation F of f is versal if for any s-parameter deformation G there
is a map-germ h : (Ks, 0) → (Kr , 0) such that G is equivalent to h∗ F. It is called miniversal if
there is no versal deformation with less than r parameters.
iii) Two r-parameter deformations F and G of f are isomorphic if there exists a diffeomor-
phism h : (Kr , 0) → (Kr , 0) such that G is equivalent to h∗ F.
4.3. CHARACTERIZATIONS OF VERSALITY 29

Example 4.2.6. Consider the function f (x, y) = x 2 + y 3 and the deformations F (x, y, u1, u2 ) =
x 2 + y 3 + u1 y + u2 and G(x, y, v) = x 2 + y 3 + 3vy 2 . Given the map-germ h(v) = (−3v 2, 2v 3 ),
we get h∗ F (x, y, v) = x 2 + y 3 − 3v 2 y + 2v 3 . Using the diffeomorphism Φ(v, x, y) = (v, x, y + v)
we see that G = h∗ F ◦ Φ, i.e. G is equivalent to h∗ F. We will see in the next section that F is,
in fact, versal.

4.3 Characterizations of versality

The definition of versality is not very useful to verify if a certain deformation F is versal, as we
cannot find a map germ h for any other possible unfolding G, so we need certain characterizations
to find or prove when an unfolding is versal.
Remember that we have defined
TKe ( f ) = t f (θ(n)) + f ∗ (M p )θ( f )
θ( f )
and the Ke -codimension of f is dimK T K e( f )
. Let F (u, x) = f u (x) be a K -deformation of
a function germ f . We denote ∂∂uf ui |u=0 by Ḟi . The following is an infinitesimal criterion for
versality due to Martinet ([29]). We will prove only the necessity of the criterion, the proof
of sufficiency is longer and exceeds the reach of this lecture notes. The proof is based on a
fundamental geometrical lemma of the theory of deformations concerning the existence of a
certain liftable vector fields, and also relies on the Preparation Theorem. A detailed account can
be found in Chapters IV and XI in [30] or Theorem 5.1 in [36] for the A -equivalence version.
Theorem 4.3.1. The r-parameter deformation F of f is versal if and only if
TKe ( f ) + SpK { Ḟ1, . . . , Ḟr } = θ( f ),
θ( f )
i.e. the map germs Ḟ1, . . . , Ḟr generate Ke ( f ) as a K-vector space.
Remark 4.3.2. An unfolding F satisfying the condition
TKe ( f ) + SpK { Ḟ1, . . . , Ḟd } = θ( f )
is called transversal.
Proof. (Proof of Necessity) Suppose F is K − versal. We will show that F is K -transversal,
that is
TKe ( f ) + SpK { Ḟ1, . . . , Ḟr } = θ( f ).
Let g ∈ θ( f ) and consider a 1-parameter deformation G(v, x) = f (x) + vg(x) of f . Notice
that Ġ = g. Since F (u, x) = f u (x) is versal, there exists h : (K, 0) → (Kr , 0), h = (h1, . . . , hr )
such that G(v, x) is equivalent to h∗ F (v, x) = f h(v) (x). We write H (v, x) = f h(v) (x). Applying
the chain rule we get
d( f h(v) ) Xr ∂ fu Xr
h∗˙F = |v=0 = hi0 (0) |u=0 = hi0 (0) Ḟi ∈ SpK { Ḟ1, . . . , Ḟr }.
dv i=1 ∂ui i=1

Since G is equivalent to H, there exists a diffeomorphism Φ(v, x) = (v, ϕv (x)) such that
G∗ (M p ) = Φ∗ (H ∗ (M p )) or, alternatively,
D E D E
G1, . . . , G p = f 1 h(v) ◦ ϕv, . . . f p h(v) ◦ ϕv .
30 CHAPTER 4. K -VERSAL UNFOLDINGS OF MAP GERMS (REAL AND COMPLEX)

Applying the chain rule again and taking into account that f h(0) = f and that ϕ0 is the identity
we get
d
Ġ − Ḣ = g − ( f h(v) ◦ ϕv )|v=0 ∈ TKe ( f ).
dv
Then,
g ∈ TKe ( f ) + SpK { Ḟ1, . . . , Ḟd }.

Example 4.3.3. i) The deformation F (u1, u2, x, y) = x 2 + y 3 + u1 y + u2 of Example 4.2.6 is K -
versal since J ( fO)+h
n On 2
f i = J ( f ) is generated by y and 1. This explains why G(x, y, v) = x +y +3vy
3 2

is equivalent to h∗ F for some h. In fact, any other deformation H will be equivalent to a pull-back
of F.
ii) Consider f (x) = x 4 , JO( nf ) is generated by {1, x, x 2 } so a versal deformation is F (x, u1, u2, u3 ) =
x 4 + u1 x 2 + u2 x + u3 . The parameter u3 is just a translation. If you consider the plane u1, u2 , for
every point in the plane you get a different function. It is interesting to see how this function
varies and what singularities appear. For instance, along the curve (−6s2, 8s3 ), the function has
an inflection point at the origin. On one side of this curve the function has two local minima
and one local maximum, on the other side there is just one local minimum. This is called a
bifurcation diagram, we refer the reader to [36] for more details on this set.
If we consider the unfolding G(x, u1 ) = x 4 + u1 x 2 , as u1 varies we will appreciate changes in
the function, namely it has 3 critical points when u1 < 0 and 1 critical point otherwise. However,
this deformation is not versal, in particular it does not show how in any neighbourhood of the
function f there are functions with inflection points.
On the other hand, the unfolding H (x, y, u1, u2, u3, u4 ) = x 4 + u1 x 2 + u2 x + u3 + u4 x 3 is also
versal but it is not miniversal, since F has less parameters than H. In fact, H can be seen as a
trivial deformation of F.
Corollary 4.3.4. A map germ f admits a K -versal deformation if and only if its Ke -codimension
is finite. Moreover, the Ke -codimension is equal to the number of parameters in a miniversal
deformation.
Proof. Given a versal r-parameter deformation F, by the versality criterion, Ḟ1, . . . , Ḟr generate
θ( f ) θ( f )
T Ke ( f ) as a K-vector space, so Ke -cod( f ) = dimK T Ke ( f ) ≤ r. Converserly, if Ke -cod( f ) = r,
θ( f )
there exist g1, . . . , gr ∈ θ( f ) whose classes generate T Ke ( f ) over K, so F (u, x) = f (x) +
i=1 ui gi (x) is a miniversal deformation of f .
Pr

Corollary 4.3.5. Let F, G be K -versal deformations of a germ f : (Kn, 0) → (K p, 0) of finite
K-codimension r. Then F, G are K - isomorphic deformations.
Proof. Suppose first that F, G are two r-parameter miniversal deformations. Since F is K -
versal, there exists h : (Kr , 0) → (Kr , 0) such that G is equivalent to h∗ F. Since G is versal,
h∗ F is versal too. Applying the chain rule to h∗ F = f h we get
Xr ∂h j
(h∗˙F)i = (0) Ḟj
i=1 ∂ui

for i = 1, . . . , r. Since both F and h∗ F are miniversal, { Ḟ1, . . . , Ḟr } and { (h∗˙F)1, . . . , (h∗˙F)r } are
θ( f ) ∂h
bases of T K e( f )
, and so ( ∂uij (0)) is an invertible matrix. This means that h is a diffeomorphism
and so G and F are isomorphic.
4.4. RELATION BETWEEN K -EQUIVALENCE AND A -EQUIVALENCE 31

u2

u1

Figure 4.1: Different functions for different values of u1 and u2 represented in the {u1, u2 }-plane.

Now suppose F, G are versal m-parameter deformations with m > r. We have dimK SpK { Ḟ1, . . . , F˙m } =
r, so there are m − r linear combinations of the Ḟi which give 0. This means that there exists a
linear change of parameters h1 : (Km, 0) → (Km, 0) such that h1∗ F verifies that there exists m − r
partials (h1∗˙F)i which are 0, i.e. h1∗ F is a constant deformation of a miniversal deformation.
Similarly, there exists h2 such that h2∗ G is a constant deformation of a miniversal deformation.
Since h1 and h2 are diffeomorphisms, F and G are isomorphic. 

In fact, we can obtain a little more information from the above proof, as we can see in the
next corollary.

Corollary 4.3.6. Let f : (Kn, 0) → (K p, 0) be a germ of finite K − codimension r, and let

F be a K − versal deformation of f . For m ≥ r any m-parameter K -versal deformation F 0
of f is K -isomorphic to the (m − r)-parameter constant deformation of F. Hence, any two
m-parameter K -versal deformations of f are K -isomorphic.

4.4 Relation between K -equivalence and A -equivalence

The contact group K defined by Mather in [32] plays a fundamental role in the classification
of stable mappings. In this section we discuss the relation between K -equivalence and A -
equivalence, as an important tool to classify stable singularities.
32 CHAPTER 4. K -VERSAL UNFOLDINGS OF MAP GERMS (REAL AND COMPLEX)

Proposition 4.4.1. Let G : (Kn, 0) → (K p, 0) be a germ of rank r. Then, there exists an invertible
germ h : (Kn, 0) → (Kn, 0) for which F = G ◦ h is an r-parameter unfolding of a germ of rank
0.
Proof. With linear changes of coordinates in source and target, we can assume that g : (Kn, 0) →
(Kr , 0), is a rank r map-germ, where g = π ◦ G, and π : Kn → Kr is the usual projection. Hence,
applying the implicit function theorem, we can find a diffeomorphism h : (Kn, 0) → (Kn, 0)
such that g ◦ h = π. Then, F = G ◦ h is the required germ. 
Let F : (Kr × Kn, 0) → (Kr × K p, 0) be a r-parameter unfolding of a germ of rank 0,
f F : (Kn, 0) → (K p, 0). For the purposes of this section, we can assume that f F is K -finitely
determined. We now consider the correspondence
F 7→ f F .
Proposition 4.4.2. Let F, F 0 : (Kr × Kn, 0) → (Kr × K p, 0) be r-parameter unfoldings of germs
f F, f F 0, of rank 0. If F, F 0 are A -equivalent then f F, f F 0 are K -equivalent.
Proof. Since F (u, x) = (u, f (u, x)) and F 0 (u, x) = (u, f 0 (u, x)) and F 'A F 0, it follows that the
local algebras Q(F) and Q(F 0 ) are isomorphic. Now,
Or+n On
Q(F) = w = Q( f F ).
hu, f (u, x)i h f F (x)i
Similarly, we get that Q(F 0 ) w Q( f F 0 ), and then Q( f F ) w Q( f F 0 ), which imply that f F 'K f F 0,
and it follows that f F and f F 0 are K -equivalent.

It follows from above that the correspondence F 7→ f F induces a mapping from A -orbits to
K -orbits. We want to understand this mapping in detail.
Let f : (Ks × Kn, 0) → (K p, 0) be a s-deformation which is submersive. We denote by V f
the germ of smooth manifold of Ks × Kn defined by f −1 (0). Let
π f : (V f , 0) → (Ks, 0)
be the germ at 0 of the restriction to V f of the projection π : Ks × Kn → Ks .
Proposition 4.4.3. Let f , g : (Ks × Kn, 0) → (K p, 0) be K -versal s-parameter deformations of
the germs f 0, g0 : (Kn, 0) → (K p, 0) of rank 0. If f 0 and g0 are K -equivalent then π f , πg are
A -equivalent.
Proof. We first notice that a K -versal deformation of a germ of rank 0 is always submersive, so
that V f is a non singular manifold in (Ks ×Kn, 0). It follows from ???? that f , g are K -isomorphic
deformations.
As in [18], (4.3), the proof is given in two steps.
Step 1. We first consider the case f 0 = g0 . Since f , g are K -isomorphic deformations, there is
a commutative diagram
Φ
(Ks × Kn, 0) −→ (Ks × Kn, 0)
π↓ π↓
h
(Ks, 0) −→ (Ks, 0)
4.4. RELATION BETWEEN K -EQUIVALENCE AND A -EQUIVALENCE 33

such that Φ, h are invertible germs for which

( f ◦ Φ) ∗ (M p ) = g ∗ (M p ).

Then, it follows that Φ induces a mapping from Vg onto V f , yelding a commuting diagram of
germs
Φ
(Vg, 0) −→ (V f , 0)
πg ↓ πf ↓
h
(Ks, 0) −→ (Ks, 0)
expressing the fact that π f , πg are A -equivalent.
Step 2. We now consider the general case, when f 0 wK g0 . Then, there exist an invertible
germ h : (Kn, 0) → (Kn, 0) and an invertible p × p-matrix M (x) with entries in On such that
g0 (x) = M (x). f 0 (h(x)). The s-parameter deformation g0 (u, x) = M (x). f (u, h(x)) of g0 is
K -versal as well. It follows from Step 1 that πg, πg 0 are A -equivalent. To finish the proof,
notice that π f and πg 0 are A -equivalent, since 1 × h maps Vg 0 onto V f .

Let F : (Kr × Kn, 0) → (Kr × K p, 0) be an r-parameter unfolding of a K -finitely determined
germ f F : (Kn, 0) → (K p, 0) of rank 0, given by F (u, x) = (u, f (u, x)).
To the unfolding F we can associate the germ DF : (Kr × K p × Kn, 0) → (K p, 0) given by

(u, w, x) 7→ −w + f (u, x),

thus DF is an (r + p)-parameter submersive deformation of f F . Notice that there is a geometric

connection between F and DF, since VDF is the graph of f and π DF can be identified with F.
Theorem 4.4.4. The map-germ F is A -stable if and only if DF is a K -versal deformation of
fF.
Studying the fibers of the map π f , if DF is a K -versal deformation of f F is equivalent to
studying all the “nearby” analytic sets to f F = 0. The theorem shows these nearby analytic sets
fit into a stable map. Thus when n ≥ p, the study of the geometry of stable maps is inseparable
from studying the deformations of the analytic space defined by f F .
The proof of this theorem relies on the following well known results (see [31], [36].)
Theorem 4.4.5 (Preparation Theorem). Let f : (Rn, 0) → (R p, 0) be a C ∞ map-germ, E a
finitely generated E n -module. If dimR ( f ∗ (ME p )·E ) < ∞, then E is finitely generated as E p -module
(via f ).
Remark 4.4.6. When K = C, we replace Malgrange Preparation Theorem by Weierstrass
Preparation Theorem.
Proposition 4.4.7. Let f : (Kn, 0) → (K p, 0) be a K -finite map-germ. The following conditions
are equivalent:
1. f is stable.

2. TAe ( f ) = θ( f )
34 CHAPTER 4. K -VERSAL UNFOLDINGS OF MAP GERMS (REAL AND COMPLEX)

3. TAe f + f ∗ (M p )θ( f ) = θ( f )

Proof. Proof of Theorem 4.4.4

Let F (u, x) = (u, f (u, x)) be a rank r unfolding of a map-germ f 0 (x) of rank 0.
Let E be the finitely generated Or+n -module given by
θ(F)
E= . (4.1)
TAe (F) + F ∗ (Mr×p )θ(F)
Then

E θ( f 0 )
= (4.2)
Mu E TKe ( f 0 ) + SpK { Ḟ1, . . . , Ḟr , e1, . . . , e p, }
where e1, . . . , e p is the canonical basis in K p .
We now suppose F is stable. Then, it follows from Proposition ?? that E = 0. Since (4.1)
implies (4.2), we get that

θ( f 0 ) = TKe ( f 0 ) + SpK { Ḟ1, . . . , Ḟr , e1, . . . , e p },

and hence DF = −w + f (u, x) is a K -versal deformation of f 0 .

To prove the converse, it suffices to prove that (4.2) imply (4.1) and apply again Proposition
??. The proof uses Theorem 4.4.5. We refer to [29], chapter XIV for details of the proof.


Let

S(r, n, p) = {A − orbits of stable germs F : (Kr × Kn, 0) → (Kr × K p, 0)},

of rank r.
K (r, n, p) = {K − orbits of germs (Kn, 0) → (K p, 0)}
of rank 0 and K -codimension ≤ r + p.
We are now prepared to state the main result of this section.

Theorem 4.4.8. The mapping

S(r, n, p) → K (r, n, p) (4.3)
induced by the correspondence F → f F is a bijection.

Proof. It follows from Proposition 4.4.2 that the map is injective. We only need to prove that
the map is surjective. Let f 0 : (Kn, 0) → (K p, 0) be a a germ of rank 0 and K -codimension
≤ r + p. Then, we can construct an (r + p) − K -versal deformation of the form −w + f (u, x)
with f an r-parameter deformation of f 0 . This is precisely the deformation DF associated to
the r-parameter unfolding F : (Kr × Kn, 0) → (Kr × K p, 0) given by F (u, x) = (u, f (u, x)). It
follows from Theorem that DF is K -versal, and we can apply Theorem 4.4.4 to get that F is
stable. The trivial observation that F has rank r concludes the proof.


We now define the Kodaira-Spencer map of f , which can be seen as the infinitesimal
counterpart of the map 4.3.
4.4. RELATION BETWEEN K -EQUIVALENCE AND A -EQUIVALENCE 35

Definition 4.4.9. The Kodaira-Spencer map of a map-germ f : (Kn, 0) → (K p, 0) is defined as

θ( f )
ρ̄ f : θ p −→ ,
TKe f
given by ρ̄ f (ξ) = [ξ ◦ f ]. The reduced Kodaira-Spencer map is

θ( f )
ρ f : T0 C p −→ ,
TKe f
given by ρ f (v) = [ξ ◦ f ], such that ξ0 = v.

Obviously, ρ f is surjective if and only if so is ρ̄ f . Moreover, the following holds.

Theorem 4.4.10. The map ρ f is surjective if and only if f is A -stable.

θ( f )
Proof. The image of ρ̄ f (θ p ) in T Ke f is the submodule

ρ̄ f (θ p ) ρ̄ f (θ p ) + TKe ( f )
M= '
ρ̄ f (θ p ) ∩ TKe ( f ) TKe ( f )

Then the quotient of these two modules is given by

θ( f )
T Ke f θ( f )
E= ' (4.4)
M TKe f + ρ̄ f (θ p )
Notice that
TKe f + ρ¯f (θ p ) = TAe f + f ∗ (M p )θ( f ).
Then it follows that ρ̄ f is surjective if and only if dim E = 0, and we apply Proposition 4.4.7 to
get the result. 
We end this section with the following result due to Mather which is the fundamental tool to
classify stable singularities.

Theorem 4.4.11. Let F and G stable map-germs. Then

F wK G if and only if F wA G.

Proof. The proof follows from Theorem 4.4.8.



Example 4.4.12. Normal forms for (real) stable singularities whose local algebra are B2,2 ± =

(x 2 ± y 2, xy) (We use here du Plessis and Wall notation [7]. They are denoted I2,2 = (x 2 + y 2, xy)
and II2,2 = (x 2 − y 2, x y) by Mather [33].)

F : (R2 × R2, 0) → (R2 × R2, 0)

(x, y, u, v) 7→ F (x, y, u, v) = (x 2 ± y 2 + ux + vy, x y, u, v).
36 CHAPTER 4. K -VERSAL UNFOLDINGS OF MAP GERMS (REAL AND COMPLEX)

Exercises
1. Let f (x, y) = x 3 + y 2 x.
i) Find a K -miniversal deformation.
ii) Show that F (x, y, u) = x 3 + y 2 x + 2ux 2 y is a trivial deformation of f .

2. Let f (x, y) = (x 2, y 2 ). Find a K -miniversal deformation.

3. Let f : (R3, 0) → (R2, ) f (x, y, z) = (x, z 2 + y 3 ± x k y), k ≥ 2. Find a K -miniversal

deformation.

4. Show that if two deformations are equivalent, then they are isomorphic.

5. Show that if h is a diffeomorphism in the parameter space, then F and h∗ F are isomorphic.

6. Show that f is K -stable if and only if the Ke -codimension is 0.

7. Show that if f is K - stable, then all deformations are K -versal.

8. Prove Corollary 4.3.6.

Part III

Lecture 3

37
Chapter 5

Deformations of ICIS

5.1 Basic invariants of ICIS

We recall that an ICIS is a germ of complex space (X, x) such that:
1. it is a complete intersection, that is, if dim(X, x) = n, it can be embedded in (C N , x) as
the zero locus of an ideal I ⊆ O N,x generated by N − n functions I = ( f 1, . . . , f N−n );
2. it has isolated singularity, that is, there exists a representative X such that X \{x} is smooth.
The local ring of (X, x) is O X,x = O N,x /I.
It follows that (X, x) can be seen as the fibre of a holomorphic map germ
f = ( f 1, . . . , f N−n ) : (C N , x) → (C N−n, 0),
which is K -finite, by the Geometric Criterion (see Theorem ??). Conversely, the fibre (X, x)
of any K -finite map germ f : (C N , x) → (C N−n, 0) is an ICIS of dimension n. Moreover,
O X,x = O N,x / f ∗ m N−n,x = Q( f ), the local algebra of f .
We now from Theorem 2.1.1 that two holomorphic map germs f , g : (C N , x) → (C N−n, 0)
are K -equivalent if and only if Q( f )  Q(g). Thus, we have a bijection between the set of
isomorphism classes of ICIS (X, x) embedded in (C N , x) and the set of K -classes of K -finite
map germs f : (C N , x) → (C N−n, 0).
Remark 5.1.1. Suppose that (X, x) is embedded in (C N , x) and (Y, y) is embedded in (C M , y),
with N ≤ M. After taking translations, we can assume, for simplicity, that x = 0 in C N and
y = 0 in C M . Then (X, 0) is the fibre of some f : (C N , 0) → (C N−n, 0) and (Y, 0) is the fibre
of some g : (C M , 0) → (C M−n, 0). Instead of f we take the constant unfolding f × idC M−N and
now we have that (X, 0)  (Y, 0) if and only if f × idC M−N and g are K -equivalent.
Let (X, x) be an ICIS given as the fibre of a K -finite map germ f : (C N , x) → (C N−n, 0).
We define the O X,x -module
1
O X,xN−n
TX,x = ,
J MX ( f )
where J MX ( f ) is the Jacobian submodule, that is, the submodule of O X,x N−n generated by the

residue classes of the partial derivatives ∂ f /∂ x i , i = 1, . . . , n. We have an isomorphism of

O X,x -modules
1 θ( f )
TX,x ≡ ,
TKe f

39
40 CHAPTER 5. DEFORMATIONS OF ICIS

1 has finite dimension over C. Its C-dimension is called the Tjurina number
so TX,x

1 θ( f )
τ(X, x) := dimC TX,x = dimC ,
TKe f

which coincides with the Ke -codimension of f .

If M is an R-module and N is an S-module, an isomorphism between M and N will be a pair
(ϕ, L), where ϕ : R → S is a ring isomorphism and L : M → N is an R-module isomorphism
(where we consider N as an R-module via ϕ).
1 and T 1 ,
Lemma 5.1.2. If (X, x)  (Y, y), there exists an isomorphism (ϕ, L) between TX,x Y,y
where ϕ : OY,y → O X,x . In particular, τ(X, x) = τ(Y, y).

Proof. We assume for simplicity that x = 0 and y = 0. Suppose first that (X, x), (Y, y) are
the fibres of K -equivalent map germs f , g : (C N , 0) → (C N−n, 0), respectively. There exist a
pair (φ, A), where φ : (C N , 0) → (C N , 0) is a diffeomorphism and A ∈ Gl N−n (OC N ,0 ) such that
g = A · ( f ◦ φ). We consider the pair (φ∗, Ã), where φ∗ : O N → O N and Ã: θ( f ) → θ(g) is
given by Ã(ξ) = A · (ξ ◦ φ). This is an isomorphism between θ( f ) and θ(g).
Moreover, we now we know that φ∗ ( f ∗ m N−n ) = g ∗ m N−n and Ã(TKe f ) = TKe g. Hence
(φ∗, Ã) induces an isomorphism (ϕ, L), where now ϕ : O X,x → OY,y and

1 θ( f ) θ(g) 1
L : TX,x ≡ −→ ≡ TY,y . (5.1)
TKe f TKe g

Suppose now that (X, x), (Y, y) are the fibres of map germs f : (C N , 0) → (C N−n, 0) and
g : (C M , 0) → (C M−n, 0), respectively, with M ≥ N. By the previous case, we can assume
that g = f × idCr , with r = M − N. Here we consider i : (C N , 0) → (C N × Cr , 0) given by
i(x) = (x, 0). We have a pair (i ∗, Ã), where i ∗ : O M → O N , Ã: θ(g) → θ( f ) is given by
Ã(ξ) = π ◦ ξ ◦ i and π(y, u) = y.
In this case, i ∗ (g ∗ m N−n ) = f ∗ m N−n and Ã(TKe g) = TKe f . So (i ∗, Ã) induces an isomor-
phism (ϕ, L), with ϕ : OY,y → O X,x and

1 θ(g) θ( f ) 1
L : TY,y ≡ −→ ≡ TX,x . (5.2)
TKe g TKe f


Remark 5.1.3. We see in the proof of Lemma 5.1.2 that when (X, x)  (Y, y) are the fibres of
K -equivalent map germs f , g : (C N , 0) → (C N−n, 0), then there exists an isomorphism (ϕ, L)
between TX,x 1 and T 1 , for some isomorphism between the ambient local rings ϕ : O → O . We
Y,y N N
will see in Section 6.4 a theorem due to Gaffney and Hauser [15] which shows the converse. That
1 and T 1 , for some isomorphism ϕ : O → O ,
is, if exists an isomorphism (ϕ, L) between TX,x Y,y N N
then (X, x)  (Y, y).

Finally, we recall that the Kodaira-Spencer map of a map germ f , g : (C N , 0) → (C N−n, 0) is

θ( f ) 1
ρ̄ f : θ N−n −→ ≡ TX,x ,
TKe f
5.2. VERSAL DEFORMATIONS OF ICIS 41

given by ρ̄ f (ξ) = [ξ ◦ f ], where (X, x) is the fibre of f . The reduced Kodaira-Spencer map is

θ( f ) 1
ρ f : T0 C N−n −→ ≡ TX,x ,
TKe f
given by ρ f (v) = [ξ ◦ f ], such that ξ0 = v. Obviously, ρ f is surjective if and only if so is ρ̄ f .
Moreover, ρ f is surjective if and only if f is A -stable (see Theorem 4.4.10).
We remark that the Kodaira-Spencer maps depend on the choice of f . Nevertheless, if
f , g : (C N , 0) → (C N−n, 0) are A -equivalent then we have some kind of uniqueness (see Exercise
3).

Exercises
1. Consider the germs f , g : (C2, 0) → (C2, 0) given by f (x, y) = (x, y 3 + x y) and g(x, y) =
(x, y 3 ). Show:

(a) f , g are K -finite and they define the same 0-dimensional ICIS with τ = 2.
(b) ρ f is an isomorphism, so f is stable.
(c) ρg is not surjective, so g is not stable. In particular, ρ f , ρg .

2. Let f : (C N , 0) → (C N−n, 0) be holomorphic and take F (x, u) = ( f u (x), u) an r-parameter

unfolding. Show that there exists an O N+r -isomorphism
θ(F) θ( f )
L: −→
TKe F TKe f
sending the class of ∂/∂ yi into the class of ∂/∂ yi , for i = 1, . . . , N − n and the class of
∂/∂u j into the class of ∂ f u /∂u j |u=0 , for j = 1, . . . , r.

3. Show that if f , g : (C N , 0) → (C N−n, 0) are A -equivalent, we have commutative diagrams

ρ̄ f ρf
θ N−n / T1 T0 C N−n / T1
X,x X,x

 ρ̄ g
  ρg

θ N−n / T1 T0 C N−n / T1
Y,y Y,y

where the rows are isomorphisms induced by the A -equivalence and (X, x), (Y, y) are the
fibres of f , g, respectively.

4. Show that any ICIS (X, x) is isomorphic to the fibre of a K -finite map germ f : (C N , 0) →
(C N−n, 0) of rank 0 (in that case N is called the embedding dimension of (X, x)).

5.2 Versal deformations of ICIS

In this section, we study versal deformations of ICIS, which will be closely related to the notion
of K -versal unfoldings of K -finite map germs f : (C N , 0) → (C N−n, 0).
42 CHAPTER 5. DEFORMATIONS OF ICIS

Definition 5.2.1. Let (X0, x 0 ) be an ICIS of dimension n. A deformation of (X0, x 0 ) is a pair

(ι, f ), where f : (X, x) → (S, s) is a holomorphic map germ between smooth spaces with
dim(X, x) − dim(S, s) = n and ι is an isomorphism from (X0, x 0 ) to the fibre (X s, x) of f , so ι
induces an isomorphism between O X,x /mS,s O X,x and O X0,x 0 .
A morphism from a deformation (ι0, f 0 ) to another (ι, f ) is a pair (g, g̃) such that the diagram
g̃
(X 0, x 0 ) / (X, x) (5.3)
f0 f
 g 
(S0, s0 ) / (S, s)

is Cartesian and g̃ ◦ ι0 = ι.
A deformation (ι, f ) is called versal if for any other deformation (ι0, f 0 ) there exists a
morphism (g, g̃) from (ι0, f 0 ) to (ι, f ). In general, we do not require the morphism (g, g̃) to be
unique in any sense. However, if the differential dgs 0 : Ts 0 S0 → Ts S is unique, we say that the
deformation is miniversal.
Remark 5.2.2. Let (ι, f ) be a deformation, with f : (X, x) → (S, s). Assume we have an
A -equivalence (φ, ψ) between f and f 0 : (X 0, x) → (S0, s). Then (ι0, f 0 ), with φ ◦ ι, is also
a deformation isomorphic to (ι, f ) (see Exercise 1). In particular, (ι, f ) is isomorphic to a
deformation of the form (ι0, f 0 ), with f 0 : (C N , 0) → (C N−n, 0).
Lemma 5.2.3. Let (g, g̃) be a morphism from (ι0, f 0 ) to (ι, f ). Then we have a commutative
diagram
ρf 0
Ts 0 S0 / T10 (5.4)
X s 0 ,x 0

dgs 0 L
 ρf

Ts S / T1
X s ,x

for some isomorphism (ϕ, L).

Proof. We split the morphism (g, g̃) as the composition of two morphisms:

(X 0, x 0 ) / (X × S0, (x, s0 )) / (X, x) (5.5)

f0 f ×idS 0 f
 (g,idS 0 )  π1 
(S0, s0 ) / (S × S0, (s, s0 )) / (S, s)

where π1 is the projection onto the first component. It is enough to show that for each morphism
we have a commutative diagram as in (5.4).
In the left hand side of (5.5), ϕ = (g, idS 0 ) is an embedding. The fact that (X 0, x 0 ) is smooth
implies that ϕ is transverse to F = f × idS 0 . Thus, we can choose coordinates in such a way that
such Cartesian square is transformed into another one of the form
j
(C N , 0) / (C N × Cr , 0)
h H
 
(C N−n, 0)
i / (C N−n × Cr , 0)
5.2. VERSAL DEFORMATIONS OF ICIS 43

where i(y) = (y, 0), j (x) = (x, 0) and H is an unfolding of h. We have a commutative diagram
ρh
T0 C N−n / θ(h)
T KO e h

di 0 L

ρH
T0 (C N−n × Cr ) / θ(H)
T Ke H

where L is the isomorphism of Exercise 5.1.2. In fact, for i = 1, . . . , N − n,

∂ ∂ ∂ ∂ ∂
! ! " # " # !
L ◦ ρ H ◦ di 0 = L ◦ ρH =L = = ρh .
∂ yi 0 ∂ yi 0 ∂ yi ∂ yi ∂ yi 0

Now we look at the Cartesian square in the right hand side of (5.5). By taking coordinates
again, we can see it as a Cartesian square of the form
q
(C N × Cr , 0) / (C N , 0)
H h
 
p
(C N−n × Cr , 0) / (C N−n, 0)

where p(y, u) = y, q(x, u) = x and H = h × idCr . Now it is obvious that we have a commutative
diagram
ρH
T0 (C N−n × Cr ) / θ(H)
T Ke H

dp0 L
 
ρh
T0 C N−n / θ(h)
T Ke h

where L is the isomorphism given in (5.2).



Lemma 5.2.4. Let (X0, x 0 ) be an ICIS. There exists a deformation whose reduced Kodaira-
Spencer map is an isomorphism. Any deformation admits a morphism to a deformation with
surjective Kodaira-Spencer map.

Proof. The first part is just Mather’s method to construct stable map germs. Assume that
(X0, x 0 ) is the fibre of a K -finite map germ f : (C N , 0) → (C N−n, 0) of rank 0 (see Exercise 4).
It follows that TKe f ⊆ m N θ( f ). Since f is K -finite,
θ( f )
Ke − codim( f ) = dimC < ∞,
TKe f

so m N θ( f )/TKe ( f ) has also finite dimension as a C-vector space. We choose g1, . . . , gd ∈ θ( f )

whose residue classes give a C-basis of m N θ( f )/TKe ( f ). So, ∂/∂ y1, . . . , ∂/∂ y N−n together
with g1, . . . , gd give a C-basis of θ( f )/TKe ( f ) and Ke − codim( f ) = N − n + d.
We consider the d-parameter unfolding
 Xd 
F (x, u) = f (x) + ui gi (x), u ,
i=1
44 CHAPTER 5. DEFORMATIONS OF ICIS

so (ι, F) is a deformation of (X0, x 0 ), with ι(x) = (x, 0). Its reduced Kodaira-Spencer map
θ( f )
ρ F : T0 (C N−n × Cd ) −→
TKe f
now sends ∂/∂ yi |0 to the class of ∂/∂ yi and ∂/∂u j |0 to the class of g j . Therefore, it is an
isomorphism.
To prove the second part, let (ι0, f 0 ) be any deformation of (X0, x 0 ). By taking charts in
(X, x) and (S, s) we can assume that f 0 : (C M , 0) → (C M−n, 0), for some M ≥ N. Now, we can
proceed as before, but in this case we get a d-parameter unfolding F 0 whose Kodaira-Spencer
map
θ( f )
ρ F 0 : T0 (C M−n × Cd ) −→
TKe f 0
is only surjective. The required deformation is (ι00, F 0 ), where ι00 (x) = (ι0 (x), 0) and the
morphism from (ι0, f 0 ) to (ι00, F 0 ) is defined in the obvious way. 
Theorem 5.2.5. Let (X0, x 0 ) be an ICIS. Then:
1. A deformation of (X0, x 0 ) is versal if and only if its Kodaira-Spencer map is surjective.

2. Two versal deformations of (X0, x 0 ) are isomorphic if their base germs have the same
dimension.

3. (X0, x 0 ) admits a miniversal deformation and any two are isomorphic.

4. A deformation of (X0, x 0 ) is miniversal if and only if its reduced Kodaira-Spencer map is

an isomorphism.
Proof. 1. Let (ι, f ) be a versal deformation. By Lemma 5.2.4, there exists a deformation (ι0, f 0 )
whose Kodaira-Spencer map ρ f 0 is surjective. By versality, there exists a morphism (g, g̃) from
(ι0, f 0 ) to (ι, f ). Now Lemma 5.2.3 gives ρ f ◦ dgs = L ◦ ρ f 0 , for some isomorphism L. Hence,
ρ f is also surjective.
Suppose now that (ι, f ) is a deformation such that ρ f is surjective and let (ι0, f 0 ) be any
deformation. By Lemma 5.2.4, there exists another deformation (ι00, f 00 ) such that ρ f 00 is
surjective and a morphism (ι0, f 0 ) → (ι00, f 00 ). By Theorem 4.4.10, f and f 00 are A -stable.
Suppose that f : (X, x) → (S, s) and f 00 : (X 00, x 00 ) → (S00, s00 ), with dim S = p and dim S00 = q.
Take k = max{p, q} and consider f × idCk−p and f 00 × idCk−q . These two maps are also A -stable
map germs between spaces of the same dimension and they are also K -equivalent. By Theorem
4.4.11, f × idCk−p and f 00 × idCk−q are A -equivalent and hence, the corresponding deformations
are isomorphic, by Exercise 1. Now it is obvious that we have a morphism (ι00, f 00 ) → (ι, f ).
2. Let (ι, f ) and (ι0, f 0 ) be versal deformations such that the base spaces have the same
dimension. By 1, the Kodaira-Spencer mappings are surjective, so f and f 0 are A -stable, by
Theorem 4.4.10. Since f and f 0 are K -equivalent, they are A -equivalent by Theorem 4.4.11
and hence, (ι, f ) and (ι0, f 0 ) are isomorphic (see Exercise 1).
3. and 4. Let (ι, f ) be a deformation such that ρ f is an isomorphism. By 1, it is versal
and hence, for any deformation (ι0, f 0 ) there exists a morphism (g, g̃) from (ι0, f 0 ) to (ι, f ). By
Lemma 5.2.3, ρ f ◦ dgs = L ◦ ρ f 0 , for some isomorphism L. Since ρ f is invertible, dgs is unique,
so (ι, f ) is miniversal.
5.2. VERSAL DEFORMATIONS OF ICIS 45

Conversely, let (ι, f ) be a miniversal deformation. Then ρ f is surjective, by 1. By Lemma

5.2.4, there exists a deformation (ι0, f 0 ) such that ρ f 0 is an isomorphism. By 2, (ι, f ) is
isomorphic to (ι0, f 0 × idC` ), for some ` ≥ 0. We must show that ` = 0.
In fact, let (g, g̃) be the isomorphism (ι, f ) → (ι0, f 0 × idC` ). We have another morphism
(g0, g̃0 ) given as a composition

(ι, f ) 7−→ (ι0, f 0 ) 7−→ (ι0, f 0 × idC` ).

By miniversality, dgs and dg0s must coincide, but this is only possible if ` = 0.


Remark 5.2.6. The proof of Lemma 5.2.4 gives an algorithm to construct a minimersal defor-
mation (this is Mather’s algorithm to construct stable germs).

Example 5.2.7. Let f : (C3, 0) → (C2, 0) be given by

(x, y, z) 7→ (x 2 + y 2 + z 2, x y).

Then TKe f is the submodule of O32 generated by

(2x, y), (2y, x), (2z, 0)

plus the submodule (x 2 + y 2 +z 2, x y)O32 . A C-basis of m3 O32 /TKe f is (computed with Singular):

(y, 0), (0, y), (0, z)

A miniversal deformation of the fibre (X, 0) of f is (ι, F), where ι(x, y, z) = (x, y, z, 0) and
F : (C3 × C3, 0) → (C2 × C3, 0) is given by

F (x, y, z, u) = ( f 1 (x, y, z) + u1 y, f 2 (x, y, z) + u2 y + u3 z, u).

In this case, we have τ = 5.

The following corollary gives the relationship between versality of deformations of ICIS and
K -versality of unfoldings of map germs.

Corollary 5.2.8. Let (X0, x 0 ) be an ICIS given as the fibre of a K -finite map germ f : (C N , 0) →
(C N−n, 0). Let F (x, u) = ( f u (x), u) be an r-parameter unfolding of f . The following statements
are equivalent:

1. (ι, F) is a versal deformation of (X0, x 0 ), where ι(x) = (x, 0),

2. F is A -stable,

3. F̃ (x, u, v) = ( f u (x) + v, u, v) is a K -versal deformation of f .

46 CHAPTER 5. DEFORMATIONS OF ICIS

Exercises
1. Let (ι, f ) be a deformation, with f : (X, x) → (S, s). Assume we have an A -equivalence
(φ, ψ) between f and f 0 : (X 0, x) → (S0, s). Then (ι0, f 0 ), with φ ◦ ι, is also a deformation
isomorphic to (ι, f ).

2. Compute a versal deformation and the Tjurina number of the ICIS defined by the following
map germs:
f : (C2, 0) → (C2, 0), f (x, y) = (x 3, y 2 ),
f : (C3, 0) → (C2, 0), f (x, y) = (x 3 + y 2 + z 2, x y).

5.3 Some analytic properties of versal deformations

Definition 5.3.1. Let (Z, 0) be a germ of analytic subset of (C N , 0). A vector field ξ ∈ θ N is
called logarithmic for (Z, 0) if ξ (h) ∈ I (Z, 0), for all h ∈ I (Z, 0). This is equivalent to the fact
that ξ is tangent to Z at any regular point z of Z. The subset of all logarithmic vector fields is
denoted by Der(− log Z ) and it is not difficult to see that Der(− log Z ) is an O N -submodule of
θN .

Definition 5.3.2. A free divisor is a hypersurface (Z, 0) ⊂ (C N , 0) such that Der(− log Z ) is a
free O N -module of rank N.

Definition 5.3.3. Consider a holomorphic map germ f : (Cn, 0) → (C p, 0). A vector field ξ ∈ θ p
is called liftable if there exists η ∈ θ n such that d f ◦ η = ξ ◦ f . The subset of θ p of liftable
vector fields is denoted by Lift( f ). We observe that Lift( f ) is the kernel of the morphism ω f
defined as the composition
ωf
θp / θ( f ) / θ( f ) ,
T Re f

so Lift( f ) is a O p -submodule of θ p .

We recall that the discriminant of a K -finite map germ f : (Cn, 0) → (C p, 0), with n ≥ p is
D := f (C), the image of the critical locus C of f . By the geometric criterion of K -finiteness
(Theorem ??), the restriction to the critical locus f : C → (C p, 0) is finite and hence its image
D is analytic in (C p, 0).

Lemma 5.3.4. The discriminant D of a K -finite map germ f : (Cn, 0) → (C p, 0) with n ≥ p is

a hypersurface of (C p, 0).

Proof. On one hand, C is defined as the zero locus of the ideal in On generated by the p × p-
minors of the Jacobian matrix of f . This implies that dim D = dim C ≥ n − (n − p + 1) = p − 1.
On the other hand, D is a null subset in (C p, 0), by Sard’s Theorem and thus dim D = p − 1. 

Proposition 5.3.5. Let f : (Cn, 0) → (C p, 0) be A -stable, n ≥ p ≥ 2. Then, the restriction

f : C → D is the normalisation of D.

Proof. We know that C is the zero locus of the ideal in On generated by the p × p-minors of the
Jacobian matrix of f and dim C = n − (n − p + 1) = p − 1, by Lemma 5.3.4. This implies that
5.3. SOME ANALYTIC PROPERTIES OF VERSAL DEFORMATIONS 47

C is determinantal and hence, Cohen-Macaulay. Moreover, since f is stable, the differential

d f : (Cn, x) → HomC (Cn, C p ) is multitransverse to the rank stratification of HomC (Cn, C p ).
On one hand, the singular locus Σ(C) of C is the inverse image of the subset of linear maps
of rank ≤ p − 2, which has dimension n − 2(n − p + 2) ≤ p − 4. By Serre’s conditions S2 and
R1, C is normal.
On the other hand, f : C → D is finite by the geometric criterion of K -finiteness (Theorem
??). Moreover, the multitransversality also implies that the set of pairs (x, x 0 ) ∈ C × C such that
x , x 0 and f (x) = f (x 0 ) has dimension p − 2, so f : C → D is generically one-to-one. Hence,
f : C → D is the normalisation of D.

Proposition 5.3.6. Let f : (Cn, 0) → (C p, 0) be A -stable, n ≥ p, with discriminant D. Then,
Lift( f ) = Der(− log D).
Proof. Suppose ξ ∈ Lift( f ), so d f ◦ η = ξ ◦ f , for some η ∈ θ n . By integrating η and ξ we get
flows φt and ψt , respectively, such that f ◦ φt = ψt ◦ f . Given any point y ∈ D, with y = f (x)
and x ∈ C, we have a commutative diagram
f
(Cn, x) / (C p, y)
φt ψt
 f 
(Cn, φt (x)) / (C p, ψt (y))
which gives φt (x) ∈ C and ψt (y) ∈ D. This shows that ξ is tangent to D at any regular point
y ∈ D, hence ξ ∈ Der(− log D).
Let us see the converse. The case p = 1 is trivial, because the only stable singularities are
regular points or Morse critical points and in both cases the equality Lift( f ) = Der(− log D) is
obvious. We can assume p ≥ 2 and hence, the restriction f : C → D is the normalisation of D,
by Lemma 5.3.5.
Let ξ ∈ Der(− log D). By integrating ξ we get a flow Ψ : (C p × C, 0) → (C p, 0) with the
property that ψt (D) ⊂ D. We consider the diagram
f
CO / DO
Φ Ψ
f ×idC
(C × C, 0) / (D × C, 0)
Since C is normal, (C × C, 0) is also normal. By the universal property of the normalisation,
there exists a unique analytic mapping Φ : (C × C, 0) → C such that f ◦ Φ = Ψ ◦ ( f × idC ). We
take Φ̄ : (Cn × C, 0) → (Cn, 0) an analytic extension of Φ and define η ∈ θ n as η x = ∂∂tΦ̄ (x, 0).
By construction, t f (η) = ω f (ξ) on C. Since C is normal, it is reduced and hence the
ideal of functions vanishing on C is J ( f ), the ideal in On generated by the p × p-minors of the
Jacobian matrix of f . This gives
t f (η) − ω f (ξ) ∈ J ( f )θ( f ).
Observe that J ( f ) is the 0th-Fitting ideal of θ( f )/TRe f . By Exercise 2 we have J ( f )θ( f ) ⊆
TRe f , hence
t f (η) − ω f (ξ) = t f (η 1 ),
48 CHAPTER 5. DEFORMATIONS OF ICIS

for some η 1 ∈ θ n . This implies

ω f (ξ) = t f (η − η 1 ),
so ξ ∈ Lift( f ). 
Theorem 5.3.7. Let f : (Cn, 0) → (C p, 0) be A -stable, n ≥ p. Then its discriminant D is a free
divisor.
Proof. By Proposition 5.3.6, Lift( f ) = Der(− log D), so we must show that Lift( f ) is free of
rank p. The module θ( f )/TRe f has a matrix presentation over On

Onn
df
/ On
p / θ( f ) / 0
T Re f

and the support of θ( f )/TRe f is the critical locus C, which has dimension p − 1. By a theorem
of Buchsbaum and Rim [5], θ( f )/TRe f is Cohen-Macaulay. In particular,
θ( f ) θ( f )
depth = dim = p − 1.
TRe f TRe f
But the depth of θ( f )/TRe f is the same when regarded as an O p -module via f ∗ : O p → On .
Consider the sequence of O p -modules
ωf
0 / Lift( f ) / θp / θ( f ) / 0 (5.6)
T Re f

Here ω f is surjective because f is A -stable and by definition, Lift( f ) is the kernel of ω f . So,
the sequence (5.6) is exact. By the depth Lemma,
θ( f )
( )
depth Lift( f ) ≥ min depth θ p, depth + 1 = p.
TRe f
But also depth Lift( f ) ≤ dim Lift( f ) ≤ dim θ p = p, so depth Lift( f ) = p. By the Auslander-
Buchsbaum formula, the projective dimension of Lift( f ) is 0, which means that it is a free
O p -module.
Finally, the rank of θ( f )/TRe f over O p is zero (see Exercise 3). Again the exactness of
(5.6) implies that Lift( f ) must have rank p.


Exercises
1. Let M be an R-module. Show that F0 (M) ⊆ Ann M, where F0 (M) is the 0-th Fitting
ideal of M, that is, F0 (M) is the ideal Ip (ϕ) in R generated by the p × p-minors of a matrix
presentation
ϕ
Rn / Rp / M / 0
and Ann M = {a ∈ R | aM = 0} is the annihilator of M. (Hint: Use the Cramer’s rule).

2. Let f : (Cn, 0) → (C p, 0) be holomorphic. Show that J ( f )θ( f ) ⊆ TRe f , where J ( f ) is

the ideal generated by the p × p minors of the Jacobian matrix of f .

3. Let f : (Cn, 0) → (C p, 0) be K -finite. Show that θ( f )/TRe f has rank zero over O p .
(Hint: Let h ∈ O p be a reduced equation of D and show that h ∈ AnnO p (θ( f )/TRe f ).)
Chapter 6

Topology of the generic fibres

6.1 The link of an isolated singularity

Let X be a closed analytic subset of an open set U ⊆ C N and x ∈ X such that X \ {x} is smooth
of constant dimension n. The main result of this section will be that X is homeomorphic at x to
a cone on a C ∞ -manifold, which is unique up to diffeomorphism.
Lemma 6.1.1 (Curve Selection Lemma). Let V be an open neighbourhood of p in Rm and let
f 1, . . . , f k , g1, . . . .g` be real analytic functions on V such that p is in the closure of

Z := {x ∈ V | f i (x) = 0, i = 1, . . . , k, g j (x) > 0, j = 1, . . . , `}.

Then there exists a real analytic curve γ : [0, δ) → V with γ(0) = 0 and γ(t) ∈ Z for t ∈ (0, δ).
Proof. [35, Lemma 3.1] for f i, g j polynomials, but the proof works also for real analytic
functions. 
Lemma 6.1.2. Let r : X → [0, ∞) be the restriction of a real analytic function r̃ : U → R such
that r −1 (0) = {x}. Then 0 is not an accumulation point of critical values of r | X\{x} .
Proof. See [28, Lemma 2.2]. 
Definition 6.1.3. Let r : X → [0, ∞) be as in Lemma 6.1.2. Then we say that r defines the point
x in X. We use the following notation

Xr ≤ := {x 0 ∈ X | r (x 0 ) ≤  }

and similarly Xr< , Xr= , X0<r< , etc.

Remark 6.1.4. Usually we take r : X → [0, ∞) as r (x 0 ) = k x 0 − xk 2 . Then

Xr ≤ = X ∩ B  , Xr< = X ∩ B , Xr= = X ∩ S ,

where B  , B and S are the closed ball, the open ball and the sphere of radius  centered at x,
respectively.
We recall that the cone on a topological space Z is the quotient space
Z × [0, 1]
CZ = .
Z × {0}

49
50 CHAPTER 6. TOPOLOGY OF THE GENERIC FIBRES

Proposition 6.1.5. Let r : X → [0, ∞) define x in X and let  > 0 such that Xr ≤ is compact
and r | X\{x} has no critical values in (0, ]. Then Xr= is a compact real analytic submanifold of
U and there exists a homeomorphism H from the cone on Xr= onto Xr ≤ such that

C(Xr= )
H / Xr ≤
π 1
r
% 
[0, 1]

commutes, where π is induced by the projection into the second component.

Proof. See [28, Proposition 2.4]. 

Definition 6.1.6. Let r : X → [0, ∞) and  > 0 be as in Proposition 6.1.5. The submanifold
Xr= is called the link of X at x.
When r (x 0 ) = k x 0 − xk 2 , then  > 0 is called a Milnor radius for X at x and B  (resp. B ,
S ) is called a closed Milnor ball (resp. open Milnor ball, Milnor sphere) (see Remark 6.1.4).

In the next proposition we show that, up to isotopy, the link does not depend on the choice
of r and .

Proposition 6.1.7. Let r, r 0 : X → [0, ∞) define x in X. Then there exists  > 0 such that

1. the hypotheses of 6.1.5 are satisfied for r and ;

2. if  0 > 0 is such that Xr 0 ≤ 0 ⊆ Xr ≤ , then the hypotheses of 6.1.5 are satisfied for r 0 and  0
and there exists a diffeomorphism of Xr ≤,r 0 ≥ 0 onto [0, 1] × Xr= which maps Xr= (resp.
Xr 0= 0 ) onto {0} × Xr= (resp. {1} × Xr 0= 0 ).

Proof. See [28, Proposition 2.5]. 

Corollary 6.1.8. The diffeomorphism type of the link of X at x only depends on the abstract
analytic set germ (X, x) (i.e., on the C-algebra O X,x ).

Proof. Suppose that (X, x) and (X 0, x 0 ) are analytic set germs with isolated singularity such
that O X,x  O X 0,x 0 . This implies that there exists a biholomorphism φ : (X, x) → (X 0, x 0 ). If
r : X → [0, ∞) defines x in X, then r 0 = r ◦ φ−1 : X 0 → [0, ∞) defines x 0 in X 0 and for  > 0
small enough, we have φ(Xr= ) = Xr0 0= . 

6.2 The Milnor fibration

We assume that X is closed analytic set of pure dimension n + k in some open U ⊆ C N , x ∈ X
and F : U → C k is an analytic map with the property that at each point of F −1 F (x) \ {x}, X is
nonsingular and F | X is a submersion. As this is clearly a property of the restriction f := F | X ,
we say that f defines an isolated singularity at x (even when f is a submersion at x). For
convenience we assume that f (x) = 0. We also suppose that r : X → [0, ∞) is a real analytic
function such that r | f −1 (0) defines f −1 (0) at x. We take  > 0 such that the hypotheses of 6.1.5
are statisfied for r | f −1 (0) and .
6.2. THE MILNOR FIBRATION 51

Since f −1 (0)r= is compact, there exists S a neighbourhood of 0 in C k such that f | Xr = is a

submersion along f −1 (S)r= . We take S contractible. We introduce the following notation
X := f −1 (S)r< , X := f −1 (S)r ≤ , ∂ X := f −1 (S)r= ,
so that X is open in f −1 (S) and X (resp. ∂ X) is its closure (resp. boundary) in f −1 (S).
We also denote by C f (or just C) the critical locus of f (i.e., the set of points of X which
are singular or where f is not a submersion). If k = 0, then C f is just the singular locus of X,
which we denote by X sing and we write Xreg = X \ X sing . For each s ∈ S, let X s (resp. X s ) be
the intersection of f −1 (s) with X (resp. X). Similarly, if A ⊆ S, we put
X A := X ∩ f −1 ( A), X A := X ∩ f −1 ( A).
The image D f := f (C f ) is called the discriminant of f . We call the restriction f : X → S (resp.
f : X → S) a good representative (reps. good proper representative) of f .
Remark 6.2.1. As in Remark 6.1.4, usually we take r (x 0 ) = k x 0 − xk 2 and it also common
to take S = Bη , the open ball in C k centered at 0 of radius η. In such a case, a good proper
representative has the form
f : X ∩ B  ∩ f −1 (Bη ) −→ Bη, (6.1)
where 0 < η    1 and analogously for a good representative taking B instead of B  . The
expression 0 < η    1 means that we have to choose first  > 0 small enough and once we
have fixed , then we choose η > 0 small enough, depending of .
Theorem 6.2.2. With the above notation, we have:
1. f : X → S is proper and f : ∂ X → S is a C ∞ -trivial bundle.
2. C f is analytic in X and closed in X. Moreover, f |C f is finite (i.e., proper with finite fibres).
3. X sing has dimension ≤ k and C f \ X sing has pure dimension k − 1.
4. D f is an analytic subset of S of the same dimension as C f . It is a hypersurface in S (or
void) if C f \ X sing is dense in C f .

5. The mapping f : (X S\D f , ∂ X S\D f ) → S \ D f is a C ∞ -fibre bundle pair, of which each fibre
pair (X s, ∂ X s ) is a complex analytic n-manifold with boundary.
6. f defines an ICIS at every point of X reg .
Proof. See [28, Theorem 2.8]. 
Definition 6.2.3. With the notation of Theorem 6.2.2, the fibre X s (resp. X s ), with s ∈ S \ D is
called a Milnor fibre (resp. a compact Milnor fibre) and the fibre bundle of item 5 is referred as
the Milnor fibration.
Remark 6.2.4. When we take r (x 0 ) = k x 0 − xk 2 and S = Bη as in (6.1), the Milnor fibration is
given by
f : X ∩ B  ∩ f −1 (Bη \ D f ) −→ Bη \ D f , (6.2)
and the Milnor fibre (resp. compact Milnor fibre) is X ∩ B ∩ f −1 (s) (resp. X ∩ B  ∩ f −1 (s)),
with s ∈ Bη \ D f .
52 CHAPTER 6. TOPOLOGY OF THE GENERIC FIBRES

The Milnor fibration was considered for the first time by Milnor in the case k = 1 and
X smooth [35]. The next proposition shows that the Milnor fibration only depends, up to
diffeomorphism, on the germ f : (X, x) → (C k , 0).
0 0
Proposition 6.2.5. Let f : X → S and f : X → S0 be good proper representatives of the germ
f : (X, x) → (C k , 0). Then there exist a neighbourhood T of 0 in S∩S0 and a C ∞ -diffeomorphism
0 0
H : (X T , ∂ X T ) → (X T , ∂ X T ) which is the identity on a neighbourhood of C f ∩ f −1 (T ) and
0
commutes with the projection onto T. In particular, H induces a diffeomorphism X s → X s , for
all s ∈ T.
Proof. See [28, Proposition 2.9]. 

6.3 The homotopy type of the Milnor fibre

In this section we consider the compact Milnor fibre X s of an n-dimensional ICIS (X0, x) given
as the fibre of a K -finite map germ f : (Cn+k , x) → (C k , 0). The following theorem was proved
first by Milnor [35] for the hypersurface case (k = 1) and later by Hamm [21] in the general
case.
Theorem 6.3.1. Let (X0, x) be an ICIS of dimension n and take a good proper representative
f : X → S. For any s ∈ S, X s has the homotopy type of a wedge of a finite number of n-spheres
X s ' Sn ∨ · · · ∨ Sn.
Proof. See [28, 5.7 and 5.9]. 
We use this theorem in the particular case of a compact Milnor fibre X s , with s ∈ S \ D, in
order to define the Milnor number of an ICIS.
Definition 6.3.2. Let (X0, x) be an ICIS of dimension n. The number of n-spheres in the
compact Milnor fibre X s ' S n ∨ · · · ∨ S n is called the Milnor number of (X0, x) and is denoted
by µ(X0, x).
The first important property of the Milnor number is that it only depends on the isomorphism
class of the ICIS.
Proposition 6.3.3. Let (X0, x) and (Y0, y) be ICIS such that (X0, x)  (Y0, y), then µ(X0, x) =
µ(Y0, y).
Proof. We know from Proposition 6.2.5 that µ(X0, x) only depends on the K -finite map germ
f : (Cn+k , x) → (C k , 0) which defines (X0, x). We assume x = 0 for simplicity.
By Corollary 5.2.8, there exists F (x, u) = ( f u (x), u) a stable r-parameter unfolding of f .
We take a good proper representative fo F,
0
F : B  ∩ F −1 (Bη0 ) −→ Bη0 ,
0
where 0 < η    1, where B  is a closed ball in Cn+k × Cr and Bη0 is an open ball in C k × Cr .
After shrinking  and η if necessary, we can assume that

f : B  ∩ f −1 (Bη ) −→ Bη
6.3. THE HOMOTOPY TYPE OF THE MILNOR FIBRE 53

is also a good proper representative of f where now B  is a closed ball in Cn+k and Bη is an
open ball in C k . If s ∈ Bη is a regular value of f , then (s, 0) ∈ Bη0 is a regular value of F and
moreover,
0
(B  ∩ f −1 (s)) × {0} = B  ∩ F −1 (s, 0).
This shows that f and F define the same Milnor number.
Analogously, (Y0, y) is the fibre of a K -finite map germ g : (Cn+`, 0) → (C`, 0) and there
exists G(x, v) = (gv (x), v) a stable s-parameter unfolding of g, so g and G define the same
Milnor number.
After multiplying F or G by the identity, we can assume that F, G : (C N , 0) → (C p, 0). Since
(X0, x)  (Y0, y), F and G are K -equivalent and hence A -equivalent, by Theorem ??. We have
a commutative diagram
(C N , 0)
F / (C p, 0)
φ ψ
 
(C N , 0)
G / (C p, 0)
where φ and ψ are diffeomorphisms. We can choose good proper representatives of F and G so
that we have a commutative diagram

X
F / S
φ ψ
 
0
X
G / S0

0
and we have ψ(DF ) = DG . For all s ∈ S \ DF , φ(X s ) = X ψ(s) and hence, F and G define the
same Milnor number. 
Let (X0, x) be an ICIS of dimension n, defined as the fibre of a K -finite map germ
f : (Cn+k , x) → (C k , 0). By Theorem 6.2.2, the discriminant (D, 0) is a hypersurface in
(C k , 0). Hence, there exists a line L in C k such that L ∩ D = {0}. After a linear change
of coordinates in C k we can assume that L is the line y k = 0. This is equivalent to that
f 0 := ( f 1, . . . , f k−1 ) : (Cn+k , x) → (C k−1, 0) defines an ICIS (X00 , x) of dimension n + 1. The
following result is known as the Lê-Greuel formula and was proved independently by Lê [44]
and Greuel [1].
Theorem 6.3.4 (Lê-Greuel formula). With the above notation we have
On+k,x
µ(X0, x) + µ(X00 , x) = dimC , (6.3)
I ( f 0) + J ( f )
where I ( f 0 ) = ( f 1, . . . , f k−1 ) and J ( f ) is the ideal in On+k,x generated by the maximal minors
of the Jacobian matrix of f .
Proof. See [28, 5.10]. 
The Lê-Greuel formula (6.3) allows us to compute the Milnor number of any ICIS by means
of a recursive formula. In fact, we can choose generic linear coordinates in C k such that for each
` = 1, . . . , k, the map germ f (`) := ( f 1, . . . , f ` ) : (Cn+k , x) → (C`, 0) defines an ICIS (X0(`), x)
of dimension n − ` + k.
54 CHAPTER 6. TOPOLOGY OF THE GENERIC FIBRES

Corollary 6.3.5. With the above notation we have

Xk On+k,x
µ(X0, x) = (−1) k−` dimC .
`=1 I ( f `−1 ) + J ( f ` )
Observe that in the hypersurface case k = 1, we recover the well known formula for the
Milnor number
On+1,x
µ(X0, x) = dimC ,
J( f )
where now J ( f ) is the Jacobian ideal, generated by the partial derivatives of f . Another
interesting particular case is when (X0, x) is 0-dimensional, in which case the formula for
µ(X0, x) can be simplified as follows:
Proposition 6.3.6. Let (X0, x) be a 0-dimensional ICIS. Then,

µ(X0, x) = dimC O X0,x − 1.

Proof. See [28, Proposition 5.13]. 

Example 6.3.7. Consider (X, 0) = V (x 2 + y 2 + z 2, x y) ⊂ (C3, 0). We have (X 0, 0) = V (x 2 +
y 2 + z 2 ), so µ(X 0, 0) = 1. By the Lê-Greuel formula
O3
µ(X, 0) + µ(X 0, 0) = dimC ! =6
2x 2y 2z
(x 2 + y 2 + z 2 ) + I2
y x 0

(computed with Singular). Hence µ(X, 0) = 5.

Let (X0, x) be an ICIS of dimension n, defined as the fibre of a K -finite map germ
f : (Cn+k , x) → (C k , 0). We take a good proper representative f : X → S. The following
property is known as the conservation of the Milnor number for ICIS. It was proved, more
generally, for deformations of isolated determinantal singularities in [38].
Theorem 6.3.8 (Conservation of the Milnor number). For any s ∈ S,
X
µ(X0, x) = 0
µ(X s, x 0 ) + βn (X s ),
x ∈S(X s )

where S(X s ) is the singular locus of X s and βn (X s ) is the nth-Betti number of X s (that is, the
number of n-spheres in the wedge X s ' S n ∨ · · · ∨ S n ).
Corollary 6.3.9 (Upper semicontinuity of the Milnor number). For any s ∈ S and for any
x 0 ∈ S(X s ), µ(X s, x 0 ) ≤ µ(X0, x).
Theorem 6.3.10. Let (X0, x) be an ICIS. Then µ(X0, x) = 0 if and only if (X0, x) is smooth.
Proof. If (X0, x) is smooth then it is obvious that µ(X0, x) = 0. Conversely, assume that (X0, x)
is not smooth. This is equivalent to that f has corank ` > 0 at x. We must show that µ(X0, x) > 0
and we will prove it by induction on `.
Let ` = 1 and assume x = 0 for simplicity. We can choose coordinates in (Cn+k , 0) and in
(C k , 0) such that
f (x) = (x 1, . . . , x k−1, f k (x)),
6.4. THE GAFFNEY-HAUSER THEOREM 55

2 . By the Lê-Greuel formula (6.3),

for some function f k ∈ mn+k

C{x 1, . . . , x n+k } C{x k , . . . , x n+k }

µ(X0, 0) = dimC = dimC = µ(g),
(x 1, . . . , x k−1 ) + J (x 1, . . . , x k−1, f k ) J (g)
2 , J (g) ⊆ m
where g(x k , . . . , x n+k ) = f k (0, . . . , 0, x k , . . . , x n+k ). Since g ∈ mn+1 n+1 and hence,
µ(g) > 0.
Suppose now the result is true for corank ` and assume that the corank is ` + 1. Again we
put x = 0 for simplicity. We choose coordinates in (Cn+k , 0) and in (C k , 0) such that

f (x) = (x 1, . . . , x k−`−1, f k−` (x), . . . , f k (x)),

2 . We take the 1-parameter unfolding F (x, t) = ( f (x), t),
for some functions f k−`, . . . , f k ∈ mn+k t
where
f t (x) = f (x) + t(0, . . . , 0, x k−`, 0, . . . , 0).
For all t , 0, f t has corank ` at 0. By hypothesis induction, µ(Xt, 0) > 0, where Xt = f t−1 (0).
On the other hand, by the upper semicontinuity 6.3.9, µ(X0, 0) ≥ µ(Xt, 0) > 0 and we are
done. 
The following corollary is now a direct consequence of Theorems 6.3.8 and 6.3.10.

Corollary 6.3.11. Suppose that µ(X s, x 0 ) = µ(X0, x), for some s ∈ S and x 0 ∈ X s . Then X s \ {x 0 }
is smooth.

Exercises
1. Compute the Milnor number of the ICIS defined by the following map germs:
f : (C2, 0) → (C2, 0), f (x, y) = (x p, y q ),
f : (C3, 0) → (C2, 0), f (x, y) = (x 3 + y 2 + z 2, xy).

6.4 The Gaffney-Hauser theorem

In [15], Gaffney and Hauser showed a very general theorem which states that if two germs of
complex spaces (X, x) and (Y, y) have isomorphic singular loci (in some specific sense), then
(X, x)  (Y, y). Here we present a simplified version in the case that (X, x) and (Y, y) are ICIS.
The case of hypersurfaces with isolated singularities was obtained previously by Mather and
Yau in [34].

Theorem 6.4.1. Let (X, x) and (Y, y) be ICIS given as the fibres of K -finite map germs
f : (C N , x) → (C p, 0) and g : (C N , y) → (C p, 0), respectively. Assume that there exists an
1 and T 1 , for some ϕ : O
isomorphism (ϕ, L) between TX,x Y,y N,y → O N,x . Then (X, x)  (Y, y).

Proof. For simplicity we assume x = y = 0 in C N . By hypothesis we have an isomorphism

(ϕ, L) between θ( f )/TKe f and θ(g)/TKe g, for some ϕ : O N → O N . We must show that f , g
are K -equivalent.
56 CHAPTER 6. TOPOLOGY OF THE GENERIC FIBRES

Step 1. We can assume that TKe f = TKe g.

We have ϕ = φ∗ for some diffeomorphism φ : (C N , 0) → (C N , 0). We can substitute f by
f˜ = L · ( f ◦ φ), so f˜ and f are K -equivalent, but now TKe f˜ = TKe g.
Step 2. Let f t = (1 − t) f + tg, with t ∈ C. Then TKe f t = TKe f , for all t ∈ C.
For all i = 1, . . . , n and for all j, k = 1, . . . , p, we have:
∂ ft ∂f ∂g
= (1 − t) +t ∈ TKe f ,
∂ xi ∂ xi ∂ xi
and
∂ ∂ ∂
( f t )k = (1 − t) f k + tgk ∈ TKe f ,
∂yj ∂yj ∂yj
therefore TKe f t ⊆ TKe f . The opposite inclusion is analogous, since for t , 1, we have
1
f = 1−t ( f t − tg) and for t = 1 is obvious.
Step 3. For each t 0 ∈ C, we consider the unfolding F : (Cn × C, (0, t 0 )) → (C p × C, 0)
given by F (x, t) = ( f t (x), t). Then F is Ke -trivial, that is, there exist Φ : (Cn × C, (0, t 0 )) →
(Cn × C, (0, t 0 )) unfolding of the identity and L = L t ∈ Gl p (OCn ×C,(0,t 0 ) ) such that L 0 = Ip and

L t · ( f t ◦ φt ) = f t 0 .

Let On+1 = OCn ×C,(0,t 0 ) . We have

∂ ft ∂ ft ∂ ft
( )
p
= g − f ∈ TKe f t ⊆ On+1 , . . ., + (( f t )1, . . . , ( f t ) p )On+1,
∂t ∂ x1 ∂ xn
where the right hand side is the relative version of the Ke -tangent space of the unfolding F.
That is, we can write
∂ f t Xn ∂ f t Xp ∂
− = ξi + a jk ( ft )j ,
∂t i=1 ∂ x i j,k=1 ∂ yk
for some functions ξi, a j k ∈ On+1 . We consider the funcions ξ as the components of a time
dependent vector field
Xn ∂
ξt = ξi .
i=1 ∂ x i

and the functions a j k as the component of a time dependent matrix At = (a j k ) ∈ Mp×p (On+1 ).
The above equality can be rewritten now as
∂ ft
+ d f t · ξt + At · f t = 0.
∂t
We consider the following system of differential equations
∂φt
 ∂t = ξt ◦ φt, φt 0 = id,


 ∂L t = L t · ( At ◦ φt ), L t 0 = Ip,
 ∂t
By integrating this we get Φ = (φt, t) and unfolding of the identity in (Cn × C, (0, t 0 )) and
L = L t ∈ Gl p (OCn ×C,(0,t 0 ) ). Let us see that

L t · ( f t ◦ φt ) = f t 0 .
6.4. THE GAFFNEY-HAUSER THEOREM 57

In fact,

∂
∂L t ∂φt ∂ f t
!
L t · ( f t ◦ φt ) = · ( f t ◦ φt ) + L t · (d f t ◦ φt ) · + ◦ φt
∂t ∂t ∂t ∂t
∂ ft
!
= L t · ( At ◦ φt ) · ( f t ◦ φt ) + L t · (d f t ◦ φt ) · (ξt ◦ φt ) + ◦ φt
∂t
∂ ft
!
= L t · At · f t + d f t · ξt + ◦ φt = 0.
∂t

This shows that L t · ( f t ◦ φt ) does not depend on t. But for t = t 0 we have L t 0 · ( f t 0 ◦ φt 0 ) = f t 0 .

Step 4. By Step 3, the germ of f t 0 at 0 is K -equivalent to the germ of f t at φt−1 (0),
for all t in a neighbourhood of t 0 . On one hand, we have (Xt 0, 0)  (Xt, φt−1 (0)), hence
µ(Xt 0, 0) = µ(Xt, φt−1 (0)) by Proposition 6.3.3) and Xt \ {φt−1 (0)} is smooth by Corollary 6.3.11.
On the other hand, by Step 2, TKe f t = TKe f , for all t ∈ C, which implies that 0 is a singular
point of Xt . Therefore, φt−1 (0) = 0.
This shows that F is not only Ke -trivial, but also K -trivial. The K -class of the germ f t
is locally constant and since C is connected, it is globally constant. In particular, f 0 = f and
f 1 = g are K -equivalent.

58 CHAPTER 6. TOPOLOGY OF THE GENERIC FIBRES
Part IV

Lecture 4

59
Chapter 7

Equisingularity and ICIS

7.1 Introduction and some basic examples

As mentioned earlier in these lectures, to understand a singularity X we want to understand the

“nearby" singularities, that is the singularities that appear in deformations of X. A first question
in studying suchdeformations is: given a family of sets or maps, when are all the members the
same? When are some of the members different? Equisingularity is the study of these questions.
It is easier to say when a member of family is different, than it is to say when two sets or two
maps are the same. Often the change in a single invariant suffices to pick out the members which
are out of step with the rest.
As we shall see in this lecture, studying families of sets makes the introduction of infinitesimal
methods natural and powerful, even when the integration of the resulting vectorfields only give
homeomorphisms.
A basic question is what do we mean by “the same"? And how do we tell when a family of
sets are the same using invariants of the members of the family? These questions are explored in
this lecture for families of ICIS; we shall see that the invariants typically have both a topological/
geometric and infinitesimal character.
We start with some notation to describe a family of sets. In the diagram:
X d (0) ⊂ X d+k ⊂ Y × C N
pY πY

0∈ Y = Ck

the parameter space is Y , X (0) denotes the fiber of the family over {0}, X d+k denotes the
total space of the family which is contained in Y × C N . We usually assume Y ⊂ X d+k , and
X = F −1 (0), X (y) = f y −1 (0), where f y (z) = F (y, z).
Given a family of map germs as above, we say the family is smoothly trivial if there exists a
smooth family of origin preserving bi-holomorphic germs r y such that r y (X (0)) = X (y). If the
map-germs are only homeomorphisms we say the family is C 0 trivial.

Example 7.1.1. Let X be the family of two moving lines in the plane with equation F (y, z1, z2 ) =
z1 (z2 − yz1 ) = 0. Here y is the parameter, the z2 axis is fixed, a component of every member of
the family while the line z2 − yz1 = 0 moves with y. Our intuition says that all of these sets are the

61
62 CHAPTER 7. EQUISINGULARITY AND ICIS

“same", and we know the family of functions F (y, z1, z2 ) = z1 (z2 − yz1 ) are all right equivalent
to f 0 (z1, z2 ) = z1 z2 , because they are all Morse functions. Hence the family is smoothly trivial.

Problem 7.1.2. Show that the family of 3 moving lines in C2 is smoothly trivial, by showing
that the family of functions F (y, z1, z2 ) = z1 z2 (z1 − (1 + y)z2 ) is right equivalent to f 0 (z1, z2 ) =
z1 z2 (z1 − z2 ) for y , −1. This will be a good review of some of the ideas of course A lecture 2.

In fact, our intuition suggests that the family of n moving distinct lines should be “equisingu-
lar". The next example shows that we must use a weaker notion of equisingularity than smooth
triviality if we want a notion that agrees with our intuition about the n moving lines.

Example 7.1.3. Let X be the family of four moving lines in the plane with equation F (x, y, z) =
z1 z2 (z2 + z1 )(z2 − (1 + y) Z1 ) = 0. Here y is the parameter, the z1 and z2 axis are fixed, as is
the line z2 + z1 = 0 while the line z2 − (1 + y)z1 = 0 moves with y. Here is a picture of the total
space of the family:

This family is not smoothly trivial as the next exercise shows,

Problem 7.1.4. Show that the family of 4 lines is not smoothly trivial by following the hints and
proving them: If r y is a trivialization of the family of sets, Dr y (0) must carry the tangent lines
of X (0) to X (y). If a linear map preserves the lines defined by z1 = 0, z2 = 0, z2 = −z1 then the
linear map must be a multiple of the identity. Hence r y can’t map z2 = z1 to z2 = (1 + y)z1 ,
y , 0.

Even though the family of four lines is not smoothly trivial, we would like to use infinitesimal
methods as the foundation of our theory of equisingularity. The infinitesimal approach using
vectorfields, promises to reduce equisingularity problems to algebra, just as Mather’s work does
for smooth equivalence. We discuss the kind of vectorfields we will use in the next section.

Problem 7.1.5. Show that an analytic set defined by a homogeneous polynomial of degree d in
2 variables in C2 consists of d lines counted with multiplicity.
7.2. RUGOSE VECTORFIELDS AND VERDIER’S CONDITION W 63

7.2 Rugose vectorfields and Verdier’s condition W

Given a family of hypersurfaces X over Y 1 , defined by F (y, z), the family X is smoothly trivial
if we can find a smooth vectorfield of the form V = ( ∂∂y − ξ) defined on X, ξ (y, 0) = 0, such that

DF (V ) = 0, on X.

Geometrically, this means that V is tangent to Y and to X on X0 , the set of smooth points of X,
so the flow induced by V must preserve Y and X.
If we only ask V to be real analytic at points of X0 , then, there is a canonical way to define
V , which works for every F. Here is the ξ that works.
Let
Pn ∂F ∂F ∂
i=1 ∂ y (y, z) ∂z i (y, z) ∂z i
ξ (y, z) = P .
n ∂F ∂F
i=1 ∂z i (y, z) ∂z i (y, z)
This means that any 1-parameter family of hypersurfaces has a canonical tangent vectorfield.
This is not true in general, but nonetheless, for any 1-parameter family of equidimensional
analytic sets, there does exist a cover {Ui } of X0 and a collection of vectorfields Vi real analytic
on Ui tangent to Ui .

Problem 7.2.1. Show that with this definition of ξ (y, z), DF (V ) = 0 and V is real analytic
whenever D z (F (y, z)) , 0. (Here D z (F (y, z)) is the vector of partial derivatives with respect
to the z variables.)

Verdier showed that the vectorfield V could be integrated to give a family of homeomorphisms
which trivialized X provided the inequality

kξ (y, z)k ≤ Ckzk

held on a neighborhood of the origin in X, for some C > 0 [46]. Verdier called a vector field
satisfying such an inequality a rugose vectorfield. He also defined a stratification condition,
condition W, which ensured, that if it held between all pairs of incident strata, smooth vectorfields
on the smallest stratum lifted to rugose vectorfields on larger strata.
The basic pair of strata is the case where X0 is the set of smooth points of a complex analytic
set X, and Y is a smooth subset of X at a point y ∈ Y . Condition W says that the distance
between between the tangent space to X at a point x i of X0 and the tangent space to Y at y goes
to zero as fast as the distance between x i and Y . We first need to define what we mean by the
distance between two linear spaces.
Suppose A, B are linear subspaces at the origin in C N , then define the distance from A to B
as:
k(u, v)k
dist( A, B) = sup .
kuk kvk
u ∈ B⊥ − {0}
v ∈ A − {0}
In the applications B is the “big” space and A the “small” space. The inner product is the
Hermitian inner product when we work over C. The same formula also works over R.
64 CHAPTER 7. EQUISINGULARITY AND ICIS

Example 7.2.2. For this example, we work with linear subspaces of R3 . Let A = x-axis, B a
plane with unit normal u0 , then the formula for the distance from A to B reduces to cos θ, where
θ is the small angle between u0 and the x-axis, in the plane they determine. So when the distance
is 0, B contains the x-axis.

We recall Verdier’s condition W.

Definition 7.2.3. Suppose Y ⊂ X̄, where X, Y are strata in a stratification of an analytic space,
and dist(TY0, T X x ) ≤ Cdist(x, Y ) for all x close to Y . Then the pair (X, Y ) satisfies Verdier’s
condition W at 0 ∈ Y ([46]).

Example 7.2.4. For a family of n-lines, the pair X0, Y is easily seen to satisfy this condition,
because X is made up of n smooth surfaces, intersecting along Y , and the intersection of their
tangent spaces, at points of Y is just Y . Since each component of X0 satisfies W over Y , so does
X0 .

We use the W condition for the definition of equisingularity which we will study.

Definition 7.2.5. A family X is W-equisingular (or just equisingular) if X has a stratification in

which adjacent pair of strata satisfy condition W, and the parameter space Y is a stratum.

Verdier introduced condition W after the Whitney conditions were introduced; these played
a central role in the development of the topological equisingularity of sets and maps developed
by Thom and Mather. Here is their definition in the analytic case.
If X is an analytic set, X0 the set of smooth points on X, Y a smooth subset of X, then the
pair (X0, Y ) satisfies Whitney’s condition A at y ∈ Y if for all sequences {x i } of points of X0 ,

{x i } → y
⇒ T ⊃ TYy
{T X x i } → T

The pair (X0, Y ) satisfies Whitney’s condition B at y ∈ Y if for all sequences {x i } of points
of X0 ,
{x i } → y
{T X x i } → T ⇒T ⊃ L
sec(x i, πY (x i )) → L

Problem 7.2.6. Show that the family of 4 lines satisfies the Whitney conditions. (Hint: The
family consists of submanifolds meeting pairwise transversely.)

Teissier showed that in the complex analytic case, conditions W and the Whitney conditions
are equivalent. (See [42].) As we shall see, W connects fairly easily with vectorfields, Jacobian
ideals, and modules. This is not so true for the Whitney conditions.

Example 7.2.7. This is a famous example used in many singularities talks. X is defined by
F (y, z1, z2 ) = z23 + z12 − y 2 z22 = 0. The members of the family X (y) consist of node singularities
where the loop is pulled smaller and smaller as y tends to zero, becoming a cusp at y = 0. Here
is a picture:
7.2. RUGOSE VECTORFIELDS AND VERDIER’S CONDITION W 65

Our intuition says that for y = 0 there is a drastic change in the family, and this family should
not be equisingular for any reasonable definition.

Problem 7.2.8. Show that W fails for Teissier’s example for X0, Y where Y is the y-axis at the
origin, by following these hints. If W holds, then the analytic inequality must hold along every
curve. Consider the curve φ(t) = (t, 0, t 2 ). Check that the image of φ lies in X. Now compute
each side of the inequality restricted to the image of φ. You should end up looking at

k ∂F 2
∂ y (t, 0, t )k
≤ Ckt 2 k.
kDF (t, 0, t 2 )k

Show that this cannot hold by comparing orders in t on each side of the inequality.

Problem 7.2.9. Show that Whitney’s condition b also fails for Teissier’s example, directly from
the definition. (Hint: use the curve φ(t) = (t, 0, t 2 ) again. )

If we project the surface to the y, z2 plane, the critical set of the projection is the closure of the
smooth points of the surface where the line y = 0, z2 = 0 is tangent to the surface; this happens
when F and Fx are 0, and is the curve φ. The curve φ is the polar curve of X for the projection
onto the y, z2 plane. Later on we will see that a family of plane curves is W-equisingular if and
only if the polar curve at the origin is empty.
As a first step to understanding the W condition, we consider the case where X is a hyper-
surface in Cn . We would like to re-write this condition in terms of F where F defines X. This
will allow us to develop an algebraic formulation of the W condition.
Set-up: We use the basic set-up with X k+n a family of hypersurfaces in Y k × Cn+1 .
66 CHAPTER 7. EQUISINGULARITY AND ICIS

Proposition 7.2.10. Condition W holds for (X0, Y ) at (0, 0) if and only if there exists U a
neighborhood of (0, 0) in X and C > 0 such that

∂F ∂F
k (y, z)k ≤ C sup kzi (y, z)k
∂ yl i, j ∂z j

for all (y, z) ∈ U and for 1 ≤ l ≤ k.

Proof. In this set-up, Y is a k-plane, so we will set A = Y , and calculate the distance between
Y and a tangent plane to X0 at (y, z) which is our B. At a smooth point of X k+n , we can use
DF (y, z)/kDF (y, z)k for u ∈ B⊥ , and the standard basis for the vectors from A.
Then the distance formula says that condition W holds if and only if

k ∂∂Fyl (y, z)k

sup ≤ C 00dist((y, z), Y ) = C 0 sup kzi k
1≤l≤k kDF (y, z)k 1≤i≤n+1

This is equivalent to

∂F ∂F
k (y, z)k ≤ C sup kzi k sup k (y, z)k
∂ yl 1≤i≤n+1 1≤ j ≤n+1 ∂z j

From which the desired result follows.



Denote the ideal generated by the partial derivatives of F with respect to the z variables by
∂F
Jz (F), and the ideal generated by z j by mY . Then zi ∂z j
are a set of generators for mY Jz (F). The
inequality above says that the partial derivatives of F with respect to yl go to zero as fast as the
ideal mY Jz (F). We will examine the implications of this in the next section.

7.3 The Theory of Integral Closure of Ideals and Modules

Many operations on ideals and submodules of a free module come from operations on rings.
(For other examples of this, see [13], [12], [17].)
We illustrate this idea by reviewing the notions of the integral closure of a ring and the
normalization of an analytic space, then relating these to the integral closure of an ideal in the
next section.

Definition 7.3.1. Let A, B be commutative Noetherian rings with unit, A ⊂ B a subring. Then
h ∈ B is integrally dependent on A if there exists a monic polynomial f (T ) = T n + f iT i ,
P
i=0
f i ∈ A such that f (h) = 0. The integral closure of A in B consists of all elements of B integrally
dependent on A.

Example 7.3.2. Let A be the ring of convergent power series in the germs t 2 and t 3 , denoted
C{t 2, t 3 }, B = C{t}. Then if f (T ) = T 2 − t 2 we have f (t) = 0, so t is integrally dependent on A.
In fact, B is the integral closure of A in B.
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 67

Definition 7.3.3. Let A be the local ring of an analytic space X, x, B the ring of meromorphic
functions on X at x; the space associated with the integral closure of A in B is the normalization
of X.

Example 7.3.4. Let A = C{t 2, t 3 }, B = C{t}. Then A is the local ring at the origin of the cusp
x 3 − y 2 = 0, and since t 3 /t 2 = t, the ring of meromorphic functions on X at the origin is C{t}.
So by the previous example the normalization of the cusp is a line.

In this context a ring A is normal if the integral closure of A in its quotient field is A. A space
germ is normal if its local ring is normal. Normal spaces have nice properties–they are non-
singular in codimension 1 and the Riemann removable singularities theorem is true for them.
Given a space germ X, we always have a map π N X from the normalization of X, denoted N X, to
X which is finite and generically 1-1. N X and π N X are unique up to smooth right equivalence.
You can read proofs of these facts in [20] p 154-163, working backwards as necessary.
The following exercise is easy assuming the facts in the last paragraph.

Problem 7.3.5. Show that the normalization of an irreducible curve germ X, x is C, 0.

If you know a little bit about singularities of maps, the next exercise is also easy.

Problem 7.3.6. Suppose f : Cn, 0 → C p, 0, n < p and f is an A-finitely determined map-germ.

Show (Cn, 0), f is a normalization of the image of f .

Basic Results from the Theory of Integral Closure for Ideals

The operation of integral closure of rings creates, as we shall see, an operation on ideals,
the operation of forming the integral closure of I, which is an ideal, denoted I. The integral
closure of mn J ( f ) in On plays the same role in a theory of equisingularity of functions built on
condition W as T R ( f ) does for right equivalence, and mn J ( f ) in O X,x plays a similar role for
the theory of equisingularity of hypersurfaces based on condition W.
Assume I is an ideal in O X,x , f ∈ O X,x . In discussing the properties of integral closure,
sometime we work on a small neighborhood of X. In this case, I refers to the coherent sheaf I
generates on U.
List of Basic Properties ([26]) f is integrally dependent on I if one of the following equivalent
conditions obtain:
(i) There exists a positive integer k and elements a j in I j , so that f satisfies the relation
f k + a1 f k−1 + · · · + a k−1 f + a k = 0 in O X,0 .
(ii) There exists a neighborhood U of 0 in C N , a positive real number C, representa-
tives of the space germ X, 0 the function germ f , and generators g1, . . . , gm of I on U,
which we identify with the corresponding germs, so that for all x in U we have: k f (x)k ≤
C max{kg1 (x)k, . . . , kgm (x)k}.
(iii) For all analytic path germs φ : (C, 0) → (X, 0) the pull–back φ∗ f = f ◦ φ is contained
in the ideal generated by φ∗ (I) in the local ring of C at 0. If for all paths φ∗ f is contained in
φ∗ (I)m1 , then we say f is strictly dependent on I and write f ∈ I † .
68 CHAPTER 7. EQUISINGULARITY AND ICIS

Let N B denote the normalization of the blowup of X by I, D̄ the pullback of the exceptional
divisor of the blowup of X by I to N B by the normalization map. Then we have:
(iv) For any component C of the underlying set of D̄, the order of vanishing of the pullback
of f to N B along C is no smaller than the order of the divisor D̄ along C. This implies that the
pullback of f lies in the ideal sheaf generated by the pullback of I.
The set of all elements of O X,x which are integrally dependent on I is the integral closure of
I and is denoted I.
Proposition 7.3.7. If I is an ideal in O X,x , then so is I.
Proof. We use property iii). Let φ : (C, 0) → (X, 0) be any analytic curve, g ∈ O X,x , f 1 , f 2 in
I. Then (g f 1 + f 2 ) ◦ φ = (g ◦ φ)( f 1 ◦ φ) + ( f 2 ◦ φ) ∈ φ∗ (I), since φ∗ (I) is an ideal in O1 . 
The proof of this for general rings is Corollary 1.3.1 of [22].
The first property is usually taken as the definition, and shows that integral dependence is
an algebraic idea. This permits the extension of the concept to ideals in any ring. For the
development of the idea of the integral closure of an ideal or module from the algebraic point of
view see [22].
The second property is used to control equisingularity conditions. It already appeared in
the discussion of Verdier’s condition W in the hypersurface case earlier, and we will revisit it
shortly.
The third property is convenient for computations, and often for proofs as the proof of the
previous proposition shows. It is also helpful in understanding conditions involving limits. In
the analytic setting, definitions that use sequences of points, such as the Whitney conditions,
can be checked with curves, often leading to an interpretation of the condition in terms of the
integral closure of an ideal or module. We will see an example of this in the study of limiting
tangent hyperplanes in the next section.
Given a curve φ(s), and a germ f , if f ◦ φ is defined, it is equal to csr mod mr+1
1 for c , 0
for some r. We call r the order of f on φ and write f φ = r, and Jφ for the order of an ideal J on
φ. Then f ∈ I † if and only if f φ > Iφ , for all curves φ.
Because the exceptional divisor of the blow-up of the Jacobian ideal tracks limiting infinites-
imal information, the fourth property is perhaps the most important. Since N B is normal, each
component of the exceptional divisor is generically a smooth submanifold of a manifold, so the
ideal vanishing on the component is locally principal. This means we can talk about the order
of vanishing on each component. The order of the divisor D̄ is just the order of vanishing along
the component of the pullback of I to N B. Concretely, pick a local generator u of the ideal of
the component, and write the elements of I in terms of u. The smallest power of u that appears
is the order of I along C.
The fourth property also shows how a closure operation on rings gives a closure operation
on ideals– start with a ring and an ideal, enlarge the ring by a closure operation, look at the ideal
generated in the new ring, then intersect with the original ring to define the closure operation on
the ideal.
The next problem is another way to see this principle for the operation of integral closure,
and gives some insight into the form of the first property.
Problem 7.3.8. Let A = O X [IT], I an ideal of O X , and T an indeterminate. Let B = O X [T].
(So A and B are rings.)Then h ∈ I in O X if and only if hT is integrally dependent on A in B.
(You can read the solution in [26].)
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 69

When we have a generic property, ie. one that holds on a non-empty Zariski-open set of X,
we would like to know when the property holds on all of X. The fourth property provides a key
step in answering this question.

Proposition 7.3.9. Let X, x be a germ of an analytic set, I an ideal of O X,x f ∈ I except perhaps
at x and suppose E ⊂ BI (X ) has no component that projects to x. Then f ∈ I.

Proof. Because f ∈ I except perhaps at x, it follows that the order of the pullback of f to any
component of D̄ is greater than or equal to the order of the pullback of I, except for components
that project to x. But there is no component of D̄ over x, because the normalization map is
finite, and there is no component of E over x. So, the order condition holds for all components
of D̄. 

This proposition says that one way of proving integral closure conditions which hold gener-
ically, is to control the exceptional divisor. We will see different ways to do this in the rest of
the notes.

Problem 7.3.10. We can improve the last proposition. Suppose X d, 0 is the germ of an analytic
set, V k , 0 a subvariety of X, I an ideal which vanishes along V . Suppose f ∈ I¯ off V k . Find a
bound on the dimension of the fiber of E over 0, which will ensure that f ∈ I¯ on X.

Problem 7.3.11. This is a corollary of the last problem. Suppose X d+k ⊂ Cn × C k is a k-

parameter family of analytic sets over Y k = 0 × C k . Suppose f ∈ I¯ off V k−1 , V k−1 ⊂ Y . Find
a bound on the dimension of the fiber of E over 0, which will ensure that f ∈ I¯ on X. If I (0)
vanishes only at the origin in X (0), what is the expected dimension of the exceptional divisor of
BI (0)(X (0))?

The meaning of the last two problems is that a generic integral closure condition in the set-up
of the last problem, will extend over the whole space, provided that the dimension of E(0) is the
same as the dimension of the exceptional divisor of BI (0) (X (0)).
Reading For detailed proofs of the equivalences between these properties see [26] p 18-27.
You can download this paper from Teissier’s list of publications–it is #15. Try this after reading
the proofs of the equivalences contained here.
In the next example, we practice using the first property.

Example 7.3.12. Let A = O2 , I = (x n, y n ). Suppose f = x i y j , i + j ≥ n. Consider the monic

polynomial h(T ) = T n − (x n )i (y n ) j . Since (x n )i (y n ) j is in (I i )(I j ) ⊂ I i+ j ⊂ I n , and h( f ) = 0,
then f ∈ I.

Now we do a computation using the third property.

Example 7.3.13. Let A = O2 , I = (x a, y b ). Given m = x i y j define the weight of m to be bi + a j,

given f (x, y), define the weight of f to be the minimum weight of all monomials appearing in a
power expansion of f . We will show that I consists of all f such that weight of f ≥ ab.

First, we’ll show weight of m ≥ ab implies m ∈ I. It suffices to check this for curves
φ(t) = (t r , t s ) as higher order terms don’t affect the order of I or the monomial m on the curve.
Since I is an ideal, this will show that f ∈ I.
We have Iφ =min{ra, sb}; assume ra ≤ sb.
70 CHAPTER 7. EQUISINGULARITY AND ICIS

It is convenient to think of the monomial x i y j as the point (i, j) in the x y-plane. Consider
the parallel lines r x + sy = c. Then if m is any monomial on this line, m φ = c, and m φ > c
if m lies above this line. If the weight of m ≥ ab then m lies above or on the line connecting
(a, 0) and (0, b), so it will lie above or on any line passing through (a, 0), which lies below or
on (0, b). This implies that m φ ≥ ra and shows m ∈ I.
Suppose the power expansion of f contains a monomial m which lies below the line con-
necting (a, 0) and (0, b). Then the convex hull of the monomials appearing in f has a vertex m0
which lies below the line connecting (a, 0) and (0, b). We can find a line passing through this
vertex which lies below (a, 0) and (0, b). Then for the curve ψ defined by this line,
f ψ = m0ψ < Iψ

which shows that f < I.

This kind of reasoning is very useful in studying properties of ideals which are well connected
to their Newton polygons. In this example, the Newton polygon of I is all the points of R2 above
or on the line connecting (a, 0) and (0, b) in the first quadrant. For more examples and details
see [43], which is #46 on Teissier’s publication list or [40].
Next, we use property 2 to characterize Verdier’s W in the hypersurface case.
Set-up: We use the basic set-up with X k+n a family of hypersurfaces in Y k × Cn+1 .
∂F
Proposition 7.3.14. Condition W holds for (X0, Y ) at (0, 0) if and only if ∂ yl ∈ mY Jz (F) for
1 ≤ l ≤ k.
Proof. By the last proposition of the first section we know that W holds if and only if
∂F ∂F
k (y, z)k ≤ C sup kzi (y, z)k
∂ yl i, j ∂z j
∂F
But, by property 2 this is equivalent to ∂ yl ∈ mY Jz (F) for 1 ≤ l ≤ k. 

If we have a curve φ on X k+n , φ(0) = 0, and the image of φ in X0k+n except at 0, and
J (F)φ = r then we can calculate the limiting tangent hyperplane to X k+n along φ as
limit (1/sr )(DF (φ(s)))
s→0

If ∂∂Fyl ∈ Jz (F) for 1 ≤ l ≤ k, then the limiting plane is never vertical, but it does not
necessarily contain Y .
Problem 7.3.15. Show that if ∂∂Fyl for 1 ≤ l ≤ k is strictly dependent on Jz (F) then every limit
of tangent planes along every curve φ not in V (Jz (F)) contains Y .
We will prove a few of the implications showing the equivalence of the basic properties.
Proposition 7.3.16. Property 1 implies property 3
Proof. Let f satisfy the relation f k +a1 f k−1 +· · ·+a k−1 f +a k = 0 in O X,0 , and let φ : C, 0 → X, 0.
Choose g ∈ I such that gφ = Iφ . We may assume the image of φ does not lie in V (I). Then
( f ◦ φ) k a1 ◦ φ ( f ◦ φ) k−1 a k−1 ◦ φ ( f ◦ φ) ak ◦ φ
+ +···+ + =0
(g ◦ φ) k (g ◦ φ) (g ◦ φ) k−1 (g ◦ φ) k−1 (g ◦ φ) (g ◦ φ) k
ai ◦φ ( f ◦φ)
and (g◦φ) i is smooth for all i. Since O1 is normal, it follows that (g◦φ) is smooth, hence
f ◦ φ ∈ φ (I).
∗ 
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 71

Proposition 7.3.17. Property 3 implies property 4

Proof. We will only prove this for the case where V (I) = 0.
Consider the components {Ci } of D̄. Since N B is normal and the Ci have codimension 1,
we can pick out points ci on each Ci and curves φ̃i , such that φ̃i (0) = ci , and φ̃i is transverse to
Ci . We can choose ci so that π ∗N B (I) vanishes only on Ci in a neighborhood of ci , and the same
is true for f ◦ π N B . If ui defines Ci at ci , then we have f ◦ π N B = hi ui f i , hi a unit. The exponent
f i is the order of vanishing of f along Ci . Since φ̃i is transverse to Ci at ci , ui ◦ φi (t) = t, so
f ◦ π N B ◦ φi (t) = hi0 (t)t f i , h0 a unit.
We can also find local generators of π ∗N B (I) of form ui Ii where Ii is the order of I along Ci .
Now π N B ◦ φ̃i is a map from C, 0 → X, 0, since π N B (Ci ) = 0, and hence π N B (ci ) = 0. (This
is the reason for restricting to this case.) Hence, if property 3 holds, f i ≥ Ii for all i. If we
work at any point of D̄ since π ∗N B (I) is principal, we can find g ◦ π N B a local generator then
f ◦ π N B /g ◦ π N B is a meromorphic function which is well defined off a set of codimension 2.
Since N B is normal, the function is analytic, so f ◦ π N B ∈ π ∗N B (I). 

Proposition 7.3.18. Property 4 implies property 2

Proof. Choose a compact neighborhood U of 0, and consider its inverse image in N B. The
inverse image must be compact as well. So, since f ◦ π N B ∈ π ∗N B (I), we can cover π −1
N B (U) with
a finite number of sets and choose elements of I such that

k f ◦ π N B (p0 )k ≤ C max{kg1 ◦ π N B (p0 )k, . . . , kgm ◦ π N B (p0 )k}

holds on π −1
N B (U). Then it is clear that

k f (π N B (p0 ))k ≤ C max{kg1 (π N B (p0 ))k, . . . , kgm (π N B (p0 ))k}.

Since π N B surjects on U, this finishes the proof. 

There is a nice corollary of the method of proof used in the previous proposition and of
property 2 which we now describe. Given a subset S of an analytic set X, f : X, S → Y, y where
S = f −1 (y) denotes the germ of an analytic map along S. Given an ideal I in OY,y , f ∗ (I) denotes
the ideal sheaf along S obtained by pulling back I by f .

Proposition 7.3.19. Suppose f : X, S → Y, y where S = f −1 (y), f proper and surjective.

Suppose I an ideal of OY,y , h ∈ OY,y . Then h ∈ I if and only if h ◦ f ∈ f ∗ (I) along S.

Proof. Since f is proper, S is compact, and as in the last proof we can cover S with a collection
of neighborhoods such that on the union the germ of a function along S is in f ∗ (I) if an only if
it satisfies an analytic inequality of the type described by property 2. Since f is surjective, the
inequalities push down/pullback to Y, y. 

Problem 7.3.20. Use the finite map f (x, y) = (x b, y a ) to give another proof that (x a, y b ) consists
of all g such that weight of g ≥ ab.

Problem 7.3.21. Show that if f ∈ On , then f ∈ mn J ( f ) in On . (Hint: let φ be a curve on Cn

and compare f φ with the order of ( f ◦ φ)0 using the chain rule to expand ( f ◦ φ)0. (For the
solution cf [41].)
72 CHAPTER 7. EQUISINGULARITY AND ICIS

We have Prop 2.10 to describe W for hypersurfaces, but what about sets of higher codimen-
sion? We will see that the theory of integral closure of modules provides the tools we need to
describe the higher codimension case.

The Theory of Integral Closure for Modules: Motivation

Verdier’s condition W is based on the distance between the tangent space T X x to X at smooth
points x and the tangent space T to Y . Recall this distance is defined as

k(u, v)k
dist(T, T X x ) = sup .
kuk kvk
u ∈ T X x⊥ − {0}
v ∈ T − {0}

If u ∈ T X x⊥ − {0}, then the set of points perpendicular to u consists of a hyperplane which

contains T X x . These hyperplanes are called tangent hyperplanes; denote a tangent hyperplane
to X, x by H x , and the collection of all tangent hyperplanes to X, x by C(X ) x . Then we can
rephrase the distance formula as

dist(T, T X x ) = sup dist(T, H x )

H x ∈C(X ) x

If X = F −1 (0) where F : Cn → C p , then at a smooth point p of X, the projectivisation of the

rowspace of the matrix of partial derivatives of F is C(X ) p . Since the tangent hyperplanes are
what we need to control the distance between the tangent space of X, p and TY, 0, this suggests
we should look at the module generated by the partial derivatives of F denoted J M (X ), just as
we looked at J (F) in the hypersurface case.

Basic Results from the Theory of Integral Closure for Modules

Notation: M ⊂ N ⊂ F p , F p a free O X,x module of rank p, M, N submodules of F. If M is

generated by g generators {mi }, then let [M] be the matrix of generators whose columns are the
{mi }.
We will develop properties for modules similar to those for ideals; however a convenient
entry way into the theory is:

Definition 7.3.22. If h ∈ F p then h is integrally dependent on M, if for all curves φ, h ◦ φ ∈

φ∗ (M). The integral closure of M denoted M consists of all h integrally dependent on M.

A good very basic reference on properties of integral closure of modules is [8, p. 301-307].
The development of these ideas in the setting of modules over commutative rings can be found
in [22] starting with chapter 16.

Problem 7.3.23. M is a module, M = M

x y 0
" #
Example 7.3.24. Let [M] = , then M = m2 O22 .
0 x y
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 73
!
y
It is clear that M ⊂ m2 O22 ; we will show that ∈ M.
0
y◦φ x◦φ
! !
Given a curve φ we can assume yφ < x φ otherwise ∈ O1 .
0 0
Then
0
! ! !
y y
◦φ= ◦ φ − x/y ◦ φ ◦φ
0 x y
where x/y ◦ φ ∈ O1 .
Connection with the theory of integral closure of ideals I
Notation: Given an element h ∈ F and a submodule M, then (h, M) denotes the submodule
generated by h and the elements of M. Given a submodule N of F, Jk (N ) denotes the ideal
generated by the set of k by k minors of a matrix whose columns are a set of generators of N.
If M is an O X module then the rank of M is k on a component V of X if Jk (M) , (0) on V and
k is the largest value for which this is true. We also denote this ideal of largest non-vanishing
minors by J (M)
Theorem 7.3.25. (Jacobian principle) Suppose the rank of (h, M) is k on each component of
(X, x). Then h ∈ M if and only if Jk (h, M) ⊂ Jk (M)
Proof. The complete proof appears in [8, p. 304]. The easy part is to show that h ∈ M implies
Jk (h, M) ⊂ Jk (M).
We have
φ∗ (Jk (h, M)) = Jk (φ∗ (h, M)) = Jk (φ∗ (M) = φ∗ (Jk (M))
which implies the result.
The problem in the other direction is checking for curves which lie in the set of points where
the rank is less than maximal, so that all the elements of Jk (h, M) vanish, but h doesn’t vanish.
We approach this problem in two steps.
Assume first that the image of our curve φ does not lie entirely in V (Jk (h, M)).
Then, by hypothesis φ∗ (Jk (h, M)) = φ∗ (Jk (M)) , 0. So, there is a non-zero minor of the
matrix of generators [M], of M, J (I, K ) such that J (I, K ) ◦ φ is generator of φ∗ (Jk (M)). Here
I is an index of the rows and K an index of the columns which comprise the k × k submatrix
whose determinant is J (I, K ).
Consider MI,K the submodule of F k defined using as matrix of generators the square sub-
matrix of [M] whose determinant is J (I, K ), and let h I be the element obtained from h by using
the entries indexed by I.
Applying Cramer’s rule, we have that h I ◦φ ∈ φ∗ (MI,K ), where h I ◦φ(t) = ([MI,K ]◦φ(t))ξ (t)
for some column vector ξ (t), given by composing the output of Cramer’s rule with φ(t). Let [MK ]
be the submatrix of [M] using the columns indexed by K. Consider h I ◦ φ(t) − ([MK ]◦ φ(t))ξ (t).
If this is zero, we have checked the condition for φ. If it is not zero, then φ∗ (h, M) has rank
greater than k which is a contradiction.
Now suppose the image of φ does lie entirely in V (Jk (h, M)), so φ∗ (Jk (h, M)) = 0.
Here the argument breaks into two parts again. We first assume X is smooth so that we can
vary the curve freely, then we use the resolution of singularities to reduce to the smooth case.
Suppose φ∗ (M) , φ∗ (h, M). Now, by the Artin-Rees theorem we know that there exists
ν0 > 0, ν0 ∈ Z such that
m1l O1 ∩ φ∗ (h, M) = m1l−ν0 (m1ν0 O1 ∩ φ∗ (h, M)).
p p
74 CHAPTER 7. EQUISINGULARITY AND ICIS

This implies, that in fact,

p
φ∗ (M) , φ∗ (h, M) mod m1l O1
p
for any l > ν0 . If not, then h ◦ φ = g mod φ∗ (M), with g ∈ m1l O1 , and so
p
g ∈ m1l O1 ∩ φ∗ (h, M),

hence
g, h ◦ φ ∈ φ∗ (M) + m1 (m1ν0 O1 ∩ φ∗ (h, M)).
p

Since φ∗ (M) + m1 φ∗ (h, M) = φ∗ (h, M), Nakayma’s lemma would imply the result.
Now choose l > ν0 ; since X is smooth, we can find a curve φ1 , by changing terms of the
power series expansion φ of order ≥ l, such that the image of φ1 does not lie in V (Jk (h, M)).
This implies that
p
φ∗1 (M) = φ∗ (M) mod m1l O1
p
φ∗1 (h, M) = φ∗ (h, M) mod m1l O1
φ∗1 (M) = φ∗1 (h, M)
This gives a contradiction in this case.
If X is not smooth, then we can make a resolution, X̃, π, of singularities of X, lift φ to φ̃ on
X̃. Then φ∗ (M) , φ∗ (h, M) if and only if φ˜∗ π ∗ (M) , φ˜∗ π ∗ (h, M), then we can again vary φ˜∗
as before.

If h ∈ M, this last proposition allows us to to do more than show h ∈ M along curves.

Proposition 7.3.26. Suppose h ∈ M, then there exists an open cover {UI,K } of the complement
of V (J (M)), such that on each UI,K , h = [M]ξ I,K , where the entries of ξ I,K are locally bounded
on UI,K .

Proof. The open cover {UI,K } is constructed by constructing an open cover {VI,K } of the fiber
over the origin in N B J (M) (X ) such that on each VI,K , the pullback of J (I, K ) is a local generator
of the pullback of J (M). Then Cramer’s rule applies, and the pullbacks of the ξ I,K are smooth,
hence locally bounded on the images of the VI,K which are the UI,K . 
As another application we can develop the analogue of property 2 for ideals.
p p
Proposition 7.3.27. ([8], Prop 1.11) Suppose h ∈ O X,x , M a submodule of O X,x of generic rank
k on each component of X. Then h ∈ M if and only if for each choice of generators {si } of M,
there exists a constant C > 0 and a neighborhood U of x such that for all ψ ∈ Γ(Hom(C p, C)),

kψ(z) · h(z)k ≤ C sup kψ(z) · si (z)k

i

for all z ∈ U.

For each choice of ψ, the {ψ · si (z)} give a linear combination of the rows of [M] at each
point, while ψ(z) · h(z) is the analogous combination of the entries of h. So the inequality of the
theorem relates the size of row vectors of [M (x)] to corresponding combinations of the entries
of h. The constant C and the neighborhood U depend on h and M but not on ψ.
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 75

Proof. We will use the Jacobian principle to show that the inequality implies the integral closure
inclusion, by using special ψi .
Let SI be a k × (k − 1) submatrix of [M], going through all such submatrices as I varies, let
h I be a k-tuple gotten by dropping the same entries from h as rows from [M] in forming SI . Let
ψ I (z)(h(z)) = det[h I (z), SI (z)]. Note that ψ I (z)si (z) = det[si (z), SI (z)], a generator of Jk (M).
The inequality which we are assuming then shows that Jk (h, M) ⊂ Jk (M), which gives the
result by the Jacobian principle.
A weaker version of the other direction is easy; if h ∈ M, then for any curve φ, (ψ(z) ·
h(z)) ◦ φ ∈ φ∗ ({ψ(z) · si (z)}), hence (ψ(z) · h(z)) ∈ ({ψ(z) · si (z)}). Then the result follows
by property 2 for ideals. However, here the constant does depend on ψ.
Instead we argue like this. Let {si } be a set of generators of M. Applying property 2 to
the finite set of elements {gi } that make up the numerators of the entries of the ξ I,K in the last
proposition, we have that there exists U and C such that if gi is such a numerator, then

kgi (z)k ≤ C sup k JI,K (z)k.

We have that JI,K (z)h(z) = gi si for appropriate gi . Then working first at z < V (J (M))
P
X
kψ(z) · h(z)k = k (gi /J (I, K ))(z)ψ(z) · si (z)k ≤ C N sup kψ(z) · si (z)k
i

where N is the number of terms in the sum. Since the inequality is between continuous functions
and holds on an open dense subset of U it holds on U. 
p p
Corollary 7.3.28. Suppose h ∈ O X,x , M a submodule of O X,x of generic rank k on each
component of X. Then h ∈ M if and only if for each choice of generators {si } of M, there exists
a constant C > 0 and a neighborhood U of x such that for all T ∈ C p ,

kT · h(z)k ≤ C sup kT · si (z)k

i

for all z ∈ U.
Proof. In one direction, take ψ to be constant; in the other we can replace T by ψ, using the fact
that the constant C is independent of the choice of T. 
There is a useful variant of the last Proposition.
p
Proposition 7.3.29. ([16]) For a section h ∈ O X to be integrally dependent on M at 0, it is
necessary that, for all maps φ : (C, 0) → (X, 0) and ψ : (C, 0) → (Hom(C p, C), λ) with λ , 0,
the function ψ(h ◦ φ) on C belong to the ideal ψ(M ◦ φ).
Conversely, it is sufficient that this condition obtain for every φ whose image meets any given
dense Zariski open subset of X.
We will use these ideas to extend our criterion for condition W to equidimensional sets of
any codimension, but first we develop the analogue of property 4 for modules.

Blowing up modules and Connection with Ideals II

We now develop the analogue of property 4 for modules. We will want a construction that
works for pairs of submodules, not just a single submodule.
76 CHAPTER 7. EQUISINGULARITY AND ICIS

Given a submodule M of a free O X d module F of rank p, we can associate a subalgebra

R (M) of the symmetric O X d algebra on p generators. This is known as the Rees algebra of M.
If (m1, . . . , m p ) is an element of M then mi Ti is the corresponding element of R (M). Then
P
Projan(R (M)), the projective analytic spectrum of R (M) is the closure of the projectivised row
spaces of M at points where the rank of a matrix of generators of M is maximal. Denote the
projection to X d by c, or by cM where there is ambiguity.
Example 7.3.30. If M is the Jacobian module of X, then Projan(R (M)) is C(X ), the projec-
tivised conormal space of X.
If M is a submodule of N or h is a section of N, then h and M generate ideals on Projan R (N );
denote them by ρ(h) and M. If we can express h in terms of a set of generators {ni } of N as
gi ni , then in the chart in which T1 , 0, we can express a generator of ρ(h) by gi Ti /T1 .
P P

Example 7.3.31. If M is the Jacobian module of X and N = F p then V (M) consists of pairs
(x, L) where x ∈ X and L ∈ PHom(C p, C), and L ◦ DF (x) = 0. If H is the hyperplane which
is the kernel of L, then the image of DF (x) lies in H.
Using 7.3.29 it is easy to show that h is integrally dependent on M at the origin, if and only
the ideal sheaf induced from h is integrally dependent as an ideal sheaf on M along 0 × P p−1 .
In other words, if and only if ρ(h) is integrally dependent on M. The combination ψ(t), φ(t)
amounts to giving path on X × P p−1 . This is the second connection between integral closure of
ideals and modules.
Looking at a pair (M, N ) allows us to “strip out" one copy of N from M, as the following
example shows.
Example 7.3.32. Let M = I = (x 2, x y, z) = J (z 2 − x 2 y) and N = J = (x, z). M is the Jacobian
ideal of the Whitney umbrella, and N defines the singular locus of the umbrella. So, working
on C3 , we have Projan R (N ) = B J (C3 ), which has ring R = O3 [T1, T2 ]/(zT1 − xT2 ), and where
the map from R (N ) to R is given by x → T1, z → T2 . Writing the generators of I in terms of the
generators of J as x 2 = x · x, x y = y · x, z = z the map from R (I) to R has image (xT1, yT1, T2 )
and this induces the ideal sheaf I on Projan R (N ). We see that this is supported only at the
point (0, [1, 0]).
The next proposition and the ideas behind it, is very useful in the study of determinantal
singularities. It is also a good example of stripping a copy of a module N from M.
p p
Proposition 7.3.33. Suppose M ⊂ N ⊂ O X,0 are O X modules with matrix of generators [M],
[N], and [F] is a matrix such that [M] = [N][F]. Let F be the ideal sheaf induced on
Projan(R (N )) by the module F with matrix of generators [F]. Then M = N if and only if V (F )
is empty.
Proof. We are going to apply 7.3.29, so we must show that for all maps φ : (C, 0) → (X, 0) and
ψ : (C, 0) → (Hom(C p, C), λ), that the order in t of ψ(t)[M] ◦ φ(t) and ψ(t)[N] ◦ φ(t) are the
same. We have

ψ(t)[M] ◦ φ(t) = ψ(t)[N][F] ◦ φ(t).

Suppose the order of ψ(t)[N]◦φ(t) in t is k. Then we can lift φ, ψ to a curve on Projan(R (N ))
as follows. Define Φ : C, 0 → X × Pg(N )−1 , by Φ(t) = (φ(t), [(1/t k )(ψ(t)[N] ◦ φ(t)]). We have
Φ(0) = (0, lim (1/t k )(ψ(t)[N] ◦ φ(t)), and the image of Φ for t , 0 clearly lies in Projan(R (N )).
t→0
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 77

Given an element f ∈ F , the value of f along Φ is (φ(t), [(1/t k )(ψ(t)[N] f˜ ◦ φ(t)]), where
f˜ is the element of F which induces f . Then V (F ) is empty if and only if the order of F along
all Φ is zero. Since [M] = [N][F] this is equivalent to the order of M and N being the same on
(ψ, φ).

Notice that if M ⊂ N and F are as above then the inclusion of M in N always induces a map
from Projan(R (N )) \ V (F ) to Projan(R (M)). The map is given by taking (x, p) to (x, F (p)),
where F (p) is evaluation of the set of generators of F which come from the columns of [F].
The next corollary includes this setting in our discussion of reduction.
Corollary 7.3.34. Suppose M and N as above, then the following are equivalent:
1. M is reduction of N.
2. V (F ) is empty.
3. The induced map is a finite map from Projan(R (N )) to Projan(R (M)).
Proof. 1) and 2) are equivalent by the previous proposition. The material in section 2 of [24]
shows that the induced map is finite if and only if V (F ) is empty. 
Here is a typical way that 3) is used.
Proposition 7.3.35. Suppose N ⊂ F, F a free O X,x module, and suppose the fiber of Projan R (N )
over x has dimension k. Then N has a reduction M, where M is generated by k + 1 elements.
Proof. Let g be the number of generators of N, so we view Projan R (N ) as a subset of X × Pg−1 .
For a generic choice of plane P in Pg−1 of codimension k + 1, the intersection of P and the fiber
of Projan R (N ) over x is empty. We can choose coordinates on Pg−1 so that the plane given by
T1 = · · · = Tk+1 = 0 is such a plane, Ti coordinates on Pg−1 . Choosing coordinates on Pg−1 is
equivalent to choosing generators on N. Let M be the submodule of N generated by the first
k + 1 generators of N after the new choice of generators. Then the projection onto the first k + 1
coordinates of Pg−1 , when restricted to Projan R (N ) gives a finite map to Projan R (M). Hence
M is a reduction of N by 3). 
Corollary 7.3.36. Suppose N ⊂ F, F a free O X,x module, X d equidimensional, N has generic
rank e on each component of X, x, then N has a reduction with d + e − 1 generators.
Proof. Since the generic rank of N is e, the generic fiber dimension of Projan R (N ) is e − 1, so
the dimension of Projan R (N ) is d + e − 1. Then d + e − 2 is the largest the dimension of the
fiber of Projan R (N ) over x can be, so N has a reduction with (d + e − 2) + 1 generators. 
Having defined the ideal sheaf M, we blow up by it. The advantages of this we will see in
the notes on determinantal singularities, as it gives a constructive/geometric way to calculate
the multiplicity of a pair of modules. But for now, this gives the context for which property 4 in
the ideal case holds. As an example of how the blow up comes up, if we are in the basic set-up,
and M = mY J M (X) then the blow up by M is the blowup of the conormal of X by the ideal
defining the stratum Y .
To state our result some more notation is needed. Given M a submodule of N ⊂ F p , h ∈ N,
let N BM (Projan R (N )), πM be the normalized blow-up of Projan R (N ) by M with projection
πM to Projan R (N ).
78 CHAPTER 7. EQUISINGULARITY AND ICIS

Proposition 7.3.37. (Analogue of Property 4 for ideals) In the above set-up h ∈ M if and only
if πM
∗ ( ρ(h)) ∈ π ∗ (M).
M

Proof. We give the proof for the case where N is free for simplicity. We apply proposition
7.3.29, so h ∈ M if and only if for all φ : (C, 0) → (X, 0) and ψ : (C, 0) → (Hom(C p, C), λ),
we have the function ψ(h ◦ φ) on C belongs to the ideal ψ(M ◦ φ). Giving the pair (φ, ψ) is
equivalent to giving a path on X × P p−1 , the order of ρ(h) on the path is the order of ψ(h ◦ φ).
So 7.3.29 is equivalent to h ∈ M if and only if the ideal sheaf induced by ρ(h) is in the integral
closer of the ideal sheaf M. In turn, by property 4 for ideals, this implies the result.


As an application we can extend our criterion for condition W to equidimensional sets of

any codimension.
Set-up: We use the basic set-up with X k+n an equidimensional family of equidimensional sets,
X k+n ⊂ Y k × C N , J M (X ) ⊂ O p .
∂F
Proposition 7.3.38. Condition W holds for (X0, Y ) at (0, 0) if and only if ∂ yl ∈ mY J M (F) for
1 ≤ l ≤ k.

Proof. We re-work the form of Verdier’s condition W to fit our current framework. If we work
at a smooth point x of X, then a conormal vector u of X at x can always be written as S · DF (x),
where S ∈ C p ; S is not unique unless DF (x) has rank p. Conversely, any such S gives a
conormal vector. It is clear also that W holds if the distance inequality holds for the standard
basis for the tangent space T of Y . Then

k(u, v)k
dist(T, T X x ) = sup .
⊥ kuk kvk
u ∈ T X x − {0}
v ∈ T − {0}
becomes
∂f
k
kS · ∂ yi
dist(T, T X x ) = sup
kS · DF (x)k
S ∈ C p − {0}
1 ≤ i ≤ k, S · DF (x) , 0
because kuk = kS · DF (x)k, and kvk = 1.
So Verdier’s condition W becomes:

∂f
sup kS · k ≤ Ckzk kS · DF (x)k .
∂ yi
S ∈ Cp
1≤i≤k
Since the functions are analytic and the inequality holds on a Z-open set of X, we can assume
it holds on a neighborhood of the origin.
Now consider the integral closure condition, ∂∂Fyl ∈ mY J M (F) for 1 ≤ l ≤ k. Using
∂F
Corollary 2.5, we have ∂ yl ∈ mY J M (F) for 1 ≤ l ≤ k if and only if
7.3. THE THEORY OF INTEGRAL CLOSURE OF IDEALS AND MODULES 79

∂f
sup kS · k ≤ C sup kzi S · DF (x)k .
p ∂ yi 1≤i≤n
S∈C
1≤i≤k
But this is easily seen to be equivalent to the previous inequality. 

This last result shows that Verdier’s condition W is exactly the geometric meaning of the ideal
sheaf induced by the ∂∂yfi being in the integral closure of the ideal sheaf induced by mY J M (X )
on X × P p−1 .
In the next section we will see how to describe and control equisingularity conditions using
multiplicity of ideals and modules.
First though, we will look at an interesting variant of W-equisingularity.
We say that a deformation F : Y k × Cn → C p with smooth parameter space Y k is WV
equisingular if ∂∂Fyl ∈ mY J Mz (F) + F ∗ (mY )On+k for 1 ≤ l ≤ k in On+k . We say each f y
p p

is WV equivalent. This is a rephrasing of the definition of WV equivalence in [8]. In the

literature, two map-germs are V-equivalent if they define isomorphic set germs. The zero sets
of WV equivalent map-germs can be placed in families which are W-equisingular. Working
with map-germs instead of sets often has big advantages, which we illustrate by a discussion of
k-WV-determinancy.
p
We say that f ∈ On is l-WV-determined, if every family of sets defined by F (y, z) =
p
f (z) + g(y, z) is WV-equisingular, where g ∈ mYl+1 On+k .
p p
Theorem 7.3.39. Suppose TK ( f ) ⊃ mnl On in On , n ≥ p. Then f is l-WV-determined.
p
Proof. The hypothesis implies that TK ( f ) = On off the origin. This follows because TK ( f ) =
p
On off the origin, and this can only happen if f is a submersion at points of X = f −1 (0) off the
origin. The inequality p ≤ n then implies X is an ICIS.
The proof then follows the standard line of ideas worked out by Mather, up to a certain
point. Consider the trivial deformation G of f . The generators of TK ( f ) as an On module also
p
generate mY J Mz (G) + G∗ (mY )On+k as an On+k module. Denote by πn the projection to Cn .
p
Given a curve φ on Cn+k it is clear that the order of the generators of mY J Mz (G) + G∗ (mY )On+k
along φ is the same as the order along πn ◦ φ, and the same is true for the generators of mYl . Thus
p p
mY J Mz (G) + G∗ (mY )On+k ⊃ mYl Ok+n
p p
in Ok+n . Since the generators of mY J Mz (F) + F ∗ (mY )On+k agree with the generators of
p
mY J Mz (G) + G∗ (mY )On+k mod mYl+1 , it follows that:

p p
mY J Mz (G) + G∗ (mY )On+k ⊃ mY J Mz (F) + F ∗ (mY )On+k .

We show the opposite inclusion is true as well. Suppose φ is a curve on Cn+k , φ(0) is the
origin. We have
p p
φ∗ (mY J Mz (F) + F ∗ (mY )On+k ) + m1 (φ∗ (mY J Mz (G) + G∗ (mY )On+k )
p
= φ∗ (mY J Mz (G) + G∗ (mY )On+k ),
80 CHAPTER 7. EQUISINGULARITY AND ICIS

because
p p p
m1 (φ∗ mY J Mz (G) + G∗ (mY )On+k ) ⊃ m1 φ∗ (mYl On+k ) ⊃ φ∗ (mYl+1 On+k ).

Applying Nakayama’s lemma shows that

p p
φ∗ (mY J Mz (F) + F ∗ (mY )On+k ) = φ∗ (mY J Mz (G) + G∗ (mY )On+k ).

Since
∂F
◦ φ ∈ φ∗ (mYl+1 On+k ) ⊂ φ∗ (mY J Mz (F) + F ∗ (mY )On+k ),
∂yj
F is WV-equisingular. 
Note that if we tried to argue on G−1 (0) and F −1 (0), we would be working on different
spaces so that the integral closure operations would not be comparable. This problem already
appears in the smooth case if we tried to compare J M ( f y ) and J M ( f 0 ) as submodules of their
respective free modules. This is why we work with the maps defining the spaces on their ambient
spaces instead of working directly with the sets.
Problem 7.3.40. Show that if n < p in the above theorem, then we still have F −1 (0) = Y k .
Problem 7.3.41. Suppose f : Cn, 0 → C p, 0, each component of f homogeneous of degree d,
p
f = 0 an ICIS. Show that TK ( f ) ⊃ mnd On , hence f is d-WV-determined. (Hint: This is easier
to do if you use the Jacobian Principle, Theorem 7.3.25.)
Problem 7.3.42. Suppose f : Cn, 0 → C p, 0. Suppose the initial form of each component of
f has degree d. Let f d be the map-germ whose component functions are these initial forms.
Suppose f d = 0 is an ICIS. Show that f is d-WV-determined.
More results on WV-equivalence can be found in [8].
p p
Problem 7.3.43. Find an example in which h ∈ J M (X ) in O X , but h is not in TKe ( f ) in On , f
defines X.

7.4 Multiplicities and Integral closure

The multiplicity of an ideal or module or pair of modules is one of the most important invariants
we can associate to an m-primary module. It is intimately connected with integral closure. It
has both a length theoretic definition and intersection theoretic definition. We give the definition
in terms of length first, for ideals, and submodules of a free module. Denote the length of a
module M by l (M).
Theorem/Definition 7.4.1. (Buchsbaum-Rim [5]) Suppose M ⊂ F, M, F both A-modules, F
free of rank p, A a Noetherian local ring of dimension d, F/M of finite length, F = A[T1, . . . , Tp ],
R (M) ⊂ F , then
λ(n) = l (Fn /M n ) is eventually a polynomial P(M, F) of degree d+p-1.
Writing the leading coefficient of P(M, F) as e(M)/(d + p − 1)!, then we define e(M) as the
multiplicity of M.
7.4. MULTIPLICITIES AND INTEGRAL CLOSURE 81

It is possible to compute simple ideal examples by hand as we show:

Example 7.4.2. Let M = I = (x 2, x y, y 2 ) ⊂ O2 . Then e(M) = 4.
We have p = 1, F = O2 , and we work with F = O2 [T1 ]. (Notice that Projan F = C2 .)
Now M n = I nT n = m22nT n , so

l (Fn /M n ) = l (O2 /m2n ) = (2n)(2n + 1)/2 = 4n2 /2! + (l.o.t.)

So e(M) = 4.
Problem 7.4.3. Let M = I = (x 2, y 2 ) ⊂ O2 . Show e(M) = 4. (Hint: Try to show that the terms
that are missing in this problem due to the missing x y term, grow only linearly with n, so the
leading term of the polynomial is the same.)
It is possible to do the very simplest module examples by hand easily as well.
Problem 7.4.4. Let M = m2 O22 . Show e(M) = 3.
The next problem is harder–try to use the same strategy as in Problem 3.3.
x y 0
" #
Problem 7.4.5. Let [M] = . Show e(M) = 3.
0 x y
Remark 7.4.6. If O X d,x is Cohen-Macaulay, and M has d + p − 1 generators where M ⊂ F p ,
then there is a useful relation between M and its ideal of maximal minors. The multiplicity of M
is the colength of M, and is also the colength of the ideal of maximal minors, by some theorems
of Buchsbaum and Rim [5], 2.4 p.207, 4.3 and 4.5 p.223.
A proof of this theorem in the context of analytic geometry using the Multiplicity Polar
theorem is given in [11]. Using this result, it is easy to do Problem 4.5. Note however, that
the colength of the ideal of maximal minors is, in general, not the multiplicity of the ideal of
maximal minors.
x y 0
" #
Problem 7.4.7. Let [M] = . Show e(J (M)) = 4.
0 x y
By the last paragraph if I = ( f 1, . . . , f d ) is an ideal of finite colength with d generators and
O X d,x is Cohen-Macaulay, then e(I) is just the degree of the mapgerm F : X, x → Cd, 0 defined
by the generators of I. So, if y ∈ Cd − ∆(F), then F −1 (y) has e(I) pre-images. Then F −1 (t y)
gives a deformation of the fiber of F over 0 to e(I) smooth points.
An important theorem both for computational and theoretical purposes was proved by Rees
in the ideal case. A proof of a generalization to modules appears in [24]
Theorem 7.4.8. Suppose M ⊂ N are m primary submodules of F p , and M = N. Then
e(M) = e(N ). Suppose further that O X,x is equidimensional, then e(M) = e(N ) implies
M=N
Several generalizations of this result exist: Kleiman and Thorup [[24], (6.8)(b)] proved a
similar result in which F p is replaced by an arbitrary finitely generated module whose support
is equidimensional; they also proved an additivity result in Theorem (6.7b)(i) of [24] for the
three pairs of modules arising from three nested modules. Generalizations also exist where the
82 CHAPTER 7. EQUISINGULARITY AND ICIS

multiplicity is not defined. Gaffney and Gassler did the case of ideals [14], and Gaffney for
modules [10], while Ulrich and Valadoshti have an approach using the epsilon multiplicity.
For computational purposes, this is coupled with another result–given any M ⊂ F p , M a
module over a local ring of dimension d, there exists a submodule R of M with d + p − 1
generators such that M = R. Such an R is called a reduction of M.
So if O X d,x is Cohen-Macaulay, we can try to find a reduction R of M with the right number
of generators d + p − 1, then calculate the length of F/R. (This length is also called the colength
of R.) Here is a very simple example.

Problem 7.4.9. Suppose I is any ideal in m2n O2 which contains x n , y n . Then e(I) = n2 .

Here is another example where it is easy to calculate e(J M (X )).

Proposition 7.4.10. Let X 1, 0 ⊂ Cn, 0 be an ICIS, defined by f = ( f 1, . . . , f n−1 ), where f i is

homogeneous of degree di . Then
Xn−1  Yn−1
e(J M (X )) = (di − 1) ( di ).
i=1
i=1

Proof. Note that X consists of a finite number of lines. If we treat the equations of X as equations
on Pn−1 , the zeroes are a discrete set of points, and these points are the lines that make up X.
The number of such points by Bezout’s theorem is ( i=1
Qn−1
di ).
We can choose n−1 columns of the matrix of partial derivatives, such that the submatrix, [N]
gotten has rank n − 1 on X except at 0. Then det[N] is homogeneous of degree ( i=1 (di − 1)),
Pn−1
since each row is homogeneous of degree di − 1. The multiplicity of e(N ), N the submodule
of O Xn−1 generated by the columns of [N], is the colength of det[N] in O X by the theorem of
Buchsbaum and Rim (7.4.6), since N is of finite colength and generated by 1 + (n − 1) − 1 = n − 1
generators. The colength of det[N] in O X , since X is Cohen-Macauley, is the degree of det[N]
on X. In turn since degree is additive, this is the sum of the degrees of det[N] on each line.
This degree is just the degree of det[N] as a homogeneous polynomial. So,
Xn−1  Yn−1
e(N ) = (di − 1) ( di )
i=1
i=1

The same computation would work for any submodule of M defined using n − 1 linear
combinations of generators of M, provided the generic rank of the submodule was n − 1 on each
line. Hence, e(N ) = e(R), where R is a reduction of M, and so e(N ) = e(M). 
We can give a topological interpretation of the e(J M (X )), X an ICIS, using the Lê -Greuel
formula.
Recall, Lê [44] and Greuel [1] proved the following formula:
OCn,0
µ(X ) + µ(X 0 ) = dimC ,
I
where X is the ICIS defined by F : (Cn, 0) → (C k , 0) ; F the map with components f 1, . . . , f k
and X 0 the ICIS defined by F 0 : (Cn, 0) → (C k+1, 0); F 0 the map with components f 1, . . . , f k+1 ,
∂( f ,..., f
I is the ideal generated by f 1, . . . , f k , and the k + 1 × k + 1-minors ∂(x i 1,...,x ik+1) ) .
1 k+1
7.4. MULTIPLICITIES AND INTEGRAL CLOSURE 83

Proposition 7.4.11. (Module form of the Lê-Greuel formula) Let X d, 0 be an ICIS, d > 0, H a
hyperplane which is not a limit tangent hyperplane to X at the origin. Then

e(J M (X ), 0) = µ(X ) + µ(X ∩ H).

Proof. Let L be the linear form defining H. We let L be f k+1 in the formula. So the right hand
side of the formula becomes µ(X ) + µ(X ∩ H). Since H is a hyperplane which is not a limit
tangent hyperplane to X at the origin, we know J M (X ) H is a reduction of J M (X ). Further, the
ideal of k + 1 × k + 1 minors of a matrix of generators of J M (X ∩ H) is the same as the ideal
of k × k minors of a matrix of generators of J M (X ) H . This implies that the colength of I in the
formula is the colength of k × k minors of J M (X ) H , which by the Buchsbaum-Rim theorem is
e(J M (X ) H ), which is e(J M (X )), since J M (X ) H is a reduction of J M (X ). 
In the ICIS case we can use multiplicity to find Milnor numbers inductively. We first do the
case of dimension 0.
Proposition 7.4.12. Suppose I defines an ICIS X of dimension 0; then µ(X ) = e(I, On ) − 1
Proof. The hypothesis implies we can find n generators f 1, . . . , f n of I; then e(I) = deg( f 1, . . . , f n )
at 0 as a map from Cn, 0 → Cn, 0. Then the inverse image of a non-critical value has e(I) points.
Fixing one point, as a common point for every 0 sphere, we get a bouquet of e(I) − 1 0-spheres.
So the Milnor number is e(I, On ) − 1. 
Now we show how the method works in an example.
Corollary 7.4.13. Let X 1 be a homogeneous ICIS, then
Xn−1  Yn−1 n−1
Y
µ(X ) = (di − 1) * di + − di + 1.
i=1
, i=1 - i=1
Proof. We know e(J M (X ), 0) = µ(X ) + µ(X ∩ H). Solving for µ(X ) we get

µ(X ) = e(J M (X ), 0) − µ(X ∩ H).

Since X has dimension 1, µ(X ∩ H) = m(X ) − 1 by the previous proposition. Since X is a union
of lines we know e(J M (X ), 0) = i=1 (di − 1)) i=1 di , while m(X ) − 1 = ( i=1 di ) − 1, from
Pn−1 Qn−1 Qn−1
which the result follows. 
Using the proof of 7.4.11, we can give another interpretation of e(J M (X )).
Proposition 7.4.14. Let X be a versal deformation of an ICIS (X, 0), defined by f : Cn, 0 →
C p, 0, n > p, let L be linear form on Cn such that L −1 (0) is not a limit tangent hyperplane to
X, 0, let π denote the projection to the base of X. Then the degree of π restricted to Σ(π, L) is
the number of critical points of L restricted to a smooth fiber of X, is e(J M (X ), 0).
Proof. We know that Σ(π, L) is Cohen-Macauley, so the degree of π restricted to Σ(π, L), is
the colength of the ideal (u1, . . . , u k ) in the local ring of Σ(π, L) at the origin, (u1, . . . , u k )
coordinates on the base of the deformation. This the same as the colength of the ideal I in
7.4.11, which we know by the proof of 7.4.11 is e(J M (X )), and if we choose a non-critical
value u of π restricted to X and to Σ(π, L), the degree is the number of critical points of L on
(X )(u) which is smooth. 
84 CHAPTER 7. EQUISINGULARITY AND ICIS

This result shows we can realize e(J M (X )) as the number of critical points of L on any
smoothing X̃ of X, provided L only has Morse singularities on X̃.
We want a theorem which extends 7.3.9 and the material in problems following it, to the
module case. The next theorem due to Kleiman and Thorup provides the necessary generaliza-
tion.
Set-up: X the germ of a reduced analytic space of pure dimension d, F a free O X -module,
M ⊂ N ⊂ F two nested submodules with M , N, M and N are generically equal and free of
rank e. Set r := d + e − 1. Set C := Projan(R (M)) where R (M) ⊂ SymF is the subalgebra
induced by M in the symmetric algebra on F. Let c : C → X denote the structure map. Let W
be the closed set in X where N is not integral over M, and set E := c−1W .

Theorem 7.4.15. (Kleiman-Thorup, [24],[25]) If N is not integral over M, then E has dimension
r − 1, the maximum possible.

A recent proof in a more general setting appears in [39].

If we have a family of analytic sets X as in the basic set-up, and M a module on X, we can
get a family of modules M (y) by restricting M to the fibers of X over Y . There are some facts
about the integral closure of ideals and their multiplicities we want to extend to modules.

Proposition 7.4.16. Suppose N ⊂ M are modules on X Assume that there is a dense Zariski
open subset V of Y such that, for each y in V , the image in O X (y) of N is a reduction of M (y).
Then there is a smaller dense Zariski open subset U of Y over which N is a reduction of M.

Proof. Cf.[16] lemma 1.2. 

Suppose M is a module on X; assume M has finite co-support over Y , so
X
e(y) = (e(M (y), (y, z))
z∈X (y)

is finite for all y. Then

Proposition 7.4.17. The function y 7→ e(y) is Zariski upper semi-continuous on Y .

Proof. Cf.[16] lemma 1.1. 

Theorem 7.4.18. (Principle of Specialization of Integral Dependence PSID) Assume that X is

equidimensional, and that y 7→ e(y) is constant on Y k . Let h be a section of a free O X module
E whose image in E(y) is integrally dependent on the image of M (y) for all y in a dense Zariski
open subset of Y . Then h is integrally dependent on M.

Proof. (Cf. Theorem 1.8 [16])

The proof of the PSID proceeds by showing that the constancy of the multiplicity means that
M has a reduction MR which is generated by dim(X (y)) + p−1 generators, which is the minimum
possible if e(M (y)) is well defined for all y. To do this, first we find such an MR whose restriction
MR (0) to X (0) is a reduction of M restricted to X (0), so e(MR (0)) = e(M (0)) by Theorem
7.4.8. Then the uppersemicontinuity of the multiplicity implies e(MR (0)) ≥ e(MR (y)), while
MR (y) ⊂ M (y) implies e(MR (y)) ≥ e(M (y)). This gives us the inequality:

e(M (0)) = e(MR (0)) ≥ e(MR (y)) ≥ e(M (y)) = e(M (0)).
7.4. MULTIPLICITIES AND INTEGRAL CLOSURE 85

Thus, by Theorem 7.4.8, MR (y) is a reduction of M (y) for all y.

Now replace M by the submodule generated by MR and g, where g may be h or any element
of M not in MR . We know MR is a reduction of (M, h) off a Zariski closed set of Y as this is
true fiberwise.
Now, the dimension of the fiber of Projan(R (MR )) over our base point x 0 ∈ X is at most
dim(X (y)) + p − 2, which is one less than the number of generators. Now the inverse image of
W in Projan(R (MR )) must have dimension at most dim(X (y)) + p − 2 + k − 1. Then since

(dim(X (y)) + p − 2 + k − 1) ≤ (dim(X (y)) + k) + (p − 1) − 2 = ((dim(X ) + p − 1) − 2,

the Kleiman-Thorup theorem then shows that M̄R = M̄, which gives the result. 

Now we come to the first of the two main results linking the equisingularity of families of
ICIS with multiplicities.

Theorem 7.4.19. Let X be a family of ICIS over Y k as in the basic setup. Suppose e(mJ M (X (y), 0))
is independent of y. Then X − Y is smooth, and the pair (X − Y, Y ) satisfies W.

Proof. Since e(y) is upper semi-continuous, there can be no points on X (y) except the origin
in the co-support of mJ M (X (y)); hence J M (X (y)) has maximal rank except at 0 so X (y) is
smooth except at 0. By 7.3.38 we have ∂∂Fyl ∈ mY J M (F) for 1 ≤ l ≤ k on a Z-open subset of Y .
So by the PSID, we have that it holds at all points.


We have seen that bounding the dimension of the fiber of C := Projan(R (mY J M (X))) over
the origin implies W. Surprisingly, in fact, by results of Teissier [42] and Lê-Teissier [45], we
know that if W holds for the pair (X0, Y ) at the origin, then the fiber dimension of both C and
Projan(R (J M (X))) over the origin in X is minimal, which is n − 2 in these cases. (Cf [42]
Chap. 5, Th. 1.2, and [45] Prop. 2.2.4.2.) The proof of these results is difficult and beyond the
scope of these notes.
For a beginner, it is a little hard to appreciate how significant these results are. Here is an
example which uses the bound on the fiber dimension of Projan(R (J M (X))).

Theorem 7.4.20. Suppose the pair (X0, Y ) satisfy W at the origin; let AY be the set of hyperplanes
in Y × C n which contain Y . For a generic member H of Ay , (H ∩ X0, Y ) also satisfies W .

Proof. Let H be any hyperplane in AY which is not a limiting tangent hyperplane to X at 0.

Since AY has dimension n − 1, and the fiber of Projan(R (J M (X))) over the origin, which is
exactly the limiting tangent hyperplanes of X at the origin, has dimension at most n − 2.
We may assume H is defined by z1 = 0. Then J Mz (X) H is a reduction of J Mz (X), and
mY J Mz (X) H is a reduction mY J Mz (X). Restricting to z1 = 0, the first statement implies that
J Mz (X ∩ H) has maximal rank at points of H ∩ X0 , hence these points are smooth. Further, the
second statement implies the pair (H ∩ X0, Y ) also satisfies W . 

Of course, repeating the construction implies that for a generic flag of planes containing Y ,
the induced family of sections satisfies W . If X d, 0 ⊂ Cn is an ICIS, let µ∗ (X, 0) denote the
sequence of Milnor numbers µi (X ) = µ(X ∩ Hi ) where Hi is a generic plane of codimension
86 CHAPTER 7. EQUISINGULARITY AND ICIS

i, 0 ≤ i ≤ d. It is clear that µi (X ) is a topological invariant of X ∩ Hi for i < d, while µd (X d )

is the multiplicity of X, less 1.
Given the product mJ M (X ), there is an expansion formula which relates e(mJ M (X )) our
infinitesimal invariant to the µ∗ invariants, which are our topological/geometric invariants.

Theorem 7.4.21. Suppose X d, 0 is an ICIS, Hi a generic plane of codimension i for X d then

n−1 Xd−1 n − 1!
!
e(mJ M (X, 0)) = m(X, 0) + e(J M (X ∩ Hi, 0))
d i=0 i

n−1 Xd−1 n − 1!
!
= (µd (X, 0) + 1) + (µi (X, 0) + µi+1 (X, 0))
d i=0 i
Proof. [9]. Note that is suffices for Hi to be part of a flag of planes such that Hi is not a limiting
tangent hyperplane to X ∩ Hi−1, 0 at the origin in Hi−1 . 
Although we refer to [9] for details, we provide some intuition in the ideal case, mJ (X ),
X n ⊂ Cn+1 . Suppose we have collections I and J of n elements of m and J (X ) which give
reductions of m and J (X ), suppose we can form A = (zi f i ), zi ∈ I, f i ∈ J (X ), A a reduction
of mJ (X ). Further suppose any of the ideals Bk = (zi1, . . . zi k , f j1, . . . f jn−k ) have the generic
value of the multiplicity for ideals of this type and e(J M (X ∩ Hk , 0)) also has this generic
value. We observe that given an ideal of the form ( f 1 f 2, g1 . . . , gd−1 ) in O X d of finite colength,
then e(( f 1 f 2, g1 . . . , gd−1 ) = e(( f 1, g1 . . . , gd−1 )) + e(( f 2, g1 . . . , gd−1 )). Using this observation
repeatedly, and the assumptions about the multiplicity of the Bk and A, we get the formula of
the theorem, as the binomial coefficients in the formula tell how many of the different Bk the
expansion process yields.

Corollary 7.4.22. Let X be a family of ICIS over Y k as in the basic setup. Suppose e(mJ M (X (y), 0))
is independent of y. Then the µ∗ sequence of X (y) is independent of y.

Proof. Although this follows indirectly from 7.4.19, and the argument of the next theorem, we
can give a simple, direct proof here. Since the µ∗ (X (y)) sequence is upper semi-continuous in
y, as is e(mJ M (X (y), 0)), all of the terms in the sum must remain constant, if the value of the
sum does. 
Now we can prove our second main result.

Theorem 7.4.23. Suppose X is a family of ICIS, and the pair (X −Y, Y ) satisfies W at the origin.
Then, the µ∗ sequence of X (y) is independent of y, as is e(m y J M (X (y))).

Proof. Since the families of generic plane sections also satisfy W by 7.4.20, it follows that these
families are topologically trivial, hence the µ∗ sequence of X (y) is independent of y. This
implies e(m y J M (X (y)) is independent of y by 7.4.21. 

Challenge Problem 7.4.24. What is the geometric meaning of e(Te K ( f ))? This is well under-
stood for functions; since f ∈ J ( f ), e(Te K ( f )) = e(J ( f )) = µ( f ).
Bibliography

[1] E. Brieskorn and G.-M. Greuel. Singularities of complete intersections. In Manifolds—

Tokyo 1973 (Proc. Internat. Conf., Tokyo, 1973), pages 123–129. Univ. Tokyo Press,
Tokyo, 1975.

[2] J. W. Bruce and P. J. Giblin. Curves and singularities. Cambridge University Press,
Cambridge, 1984. A geometrical introduction to singularity theory.

[3] J. W. Bruce, N. P. Kirk, and A. A. du Plessis. Complete transversals and the classification
of singularities. Nonlinearity, 10(1):253–275, 1997.

[4] James William Bruce. Classifications in singularity theory and their applications. In New
developments in singularity theory (Cambridge, 2000), volume 21 of NATO Sci. Ser. II
Math. Phys. Chem., pages 3–33. Kluwer Acad. Publ., Dordrecht, 2001.

[5] David A. Buchsbaum and Dock S. Rim. A generalized Koszul complex. II. Depth and
multiplicity. Trans. Amer. Math. Soc., 111:197–224, 1964.

[6] Theo de Jong and Gerhard Pfister. Local analytic geometry. Advanced Lectures in
Mathematics. Friedr. Vieweg & Sohn, Braunschweig, 2000. Basic theory and applications.

[7] Andrew du Plessis and Terry Wall. The geometry of topological stability, volume 9 of
London Mathematical Society Monographs. New Series. The Clarendon Press, Oxford
University Press, New York, 1995. Oxford Science Publications.

[8] Terence Gaffney. Integral closure of modules and Whitney equisingularity. Invent. Math.,
107(2):301–322, 1992.

[9] Terence Gaffney. Multiplicities and equisingularity of ICIS germs. Invent. Math.,
123(2):209–220, 1996.

[10] Terence Gaffney. Generalized Buchsbaum-Rim multiplcities and a theorem of Rees. Comm.
Algebra, 31(8):3811–3827, 2003. Special issue in honor of Steven L. Kleiman.

[11] Terence Gaffney. The multiplicity polar theorem and isolated singularities. J. Algebraic
Geom., 18(3):547–574, 2009.

[12] Terence Gaffney. Bi-Lipschitz equivalence, integral closure and invariants. In Real and
complex singularities, volume 380 of London Math. Soc. Lecture Note Ser., pages 125–137.
Cambridge Univ. Press, Cambridge, 2010.

87
88 BIBLIOGRAPHY

[13] Terence Gaffney. The genericity of the infinitesimal Lipschitz condition for hypersurfaces.
J. Singul., 10:108–123, 2014.

[14] Terence Gaffney and Robert Gassler. Segre numbers and hypersurface singularities. J.
Algebraic Geom., 8(4):695–736, 1999.

[15] Terence Gaffney and Herwig Hauser. Characterizing singularities of varieties and of
mappings. Invent. Math., 81(3):427–447, 1985.

[16] Terence Gaffney and Steven L. Kleiman. Specialization of integral dependence for modules.
Invent. Math., 137(3):541–574, 1999.

[17] Terence Gaffney and Marie A. Vitulli. Weak subintegral closure of ideals. Adv. Math.,
226(3):2089–2117, 2011.

[18] C. G. Gibson. Singular points of smooth mappings, volume 25 of Research Notes in

Mathematics. Pitman (Advanced Publishing Program), Boston, Mass.-London, 1979.

[19] G.-M. Greuel, C. Lossen, and E. Shustin. Introduction to singularities and deformations.
Springer Monographs in Mathematics. Springer, Berlin, 2007.

[20] R. C. Gunning. Lectures on complex analytic varieties: The local parametrization theorem.
Mathematical Notes. Princeton University Press, Princeton, N.J.; University of Tokyo
Press, Tokyo, 1970.

[21] Helmut Hamm. Lokale topologische Eigenschaften komplexer Räume. Math. Ann.,
191:235–252, 1971.

[22] Craig Huneke and Irena Swanson. Integral closure of ideals, rings, and modules, volume
336 of London Mathematical Society Lecture Note Series. Cambridge University Press,
Cambridge, 2006.

[23] Shyuichi Izumiya, Maria del Carmen Romero Fuster, Maria Aparecida Soares Ruas, and
Farid Tari. Differential geometry from a singularity theory viewpoint. World Scientific
Publishing Co. Pte. Ltd., Hackensack, NJ, 2016.

[24] Steven Kleiman and Anders Thorup. A geometric theory of the Buchsbaum-Rim multi-
plicity. J. Algebra, 167(1):168–231, 1994.

[25] Steven Kleiman and Anders Thorup. The exceptional fiber of a generalized conormal
space. In New developments in singularity theory (Cambridge, 2000), volume 21 of NATO
Sci. Ser. II Math. Phys. Chem., pages 401–404. Kluwer Acad. Publ., Dordrecht, 2001. With
an appendix by Steven Kleiman and Anders Thorup.

[26] Monique Lejeune-Jalabert and Bernard Teissier. Clôture intégrale des idéaux et équisin-
gularité. Ann. Fac. Sci. Toulouse Math. (6), 17(4):781–859, 2008. With an appendix by
Jean-Jacques Risler.

[27] S. Łojasiewicz. Introduction to complex analytic geometry. Birkhäuser Verlag, Basel,

1991. Translated from the Polish by Maciej Klimek.
BIBLIOGRAPHY 89

[28] Eduard J. N. Looijenga. Isolated singular points on complete intersections, volume 5 of

Surveys of Modern Mathematics. International Press, Somerville, MA; Higher Education
Press, Beijing, second edition, 2013.

[29] Jean Martinet. Déploiements versels des applications différentiables et classification des
applications stables. In Singularités d’applications différentiables (Sém., Plans-sur-Bex,
1975), pages 1–44. Lecture Notes in Math., Vol. 535. 1976.

[30] Jean Martinet. Singularités des fonctions et applications différentiables. Pontifíia Uni-
versidade Católica do Rio de Janeiro, Rio de Janeiro, 1977. Deuxième édition corrigée,
Monografias de Matemática da PUC/RJ, No. 1. [Mathematical Monographs of the PUC/RJ,
No. 1].

[31] John N. Mather. Stability of C ∞ mappings. I. The division theorem. Ann. of Math. (2),
87:89–104, 1968.

[32] John N. Mather. Stability of C ∞ mappings. III. Finitely determined mapgerms. Inst. Hautes
Études Sci. Publ. Math., (35):279–308, 1968.

[33] John N. Mather. Stability of C ∞ mappings. IV. Classification of stable germs by R-algebras.
Inst. Hautes Études Sci. Publ. Math., (37):223–248, 1969.

[34] John N. Mather and Stephen S. T. Yau. Classification of isolated hypersurface singularities
by their moduli algebras. Invent. Math., 69(2):243–251, 1982.

[35] J. Milnor. Singular points of complex hypersurfaces, volume 61 of Ann. of Math. Studies.
Princeton University Press, Princeton, 1968.

[36] D. Mond and J.J. Nuño-Ballesteros. Singularities of Mappings, volume 357 of Grundlehren
der mathematischen Wissenschaften. Springer International Publishing, 2020.

[37] James A. Montaldi. On contact between submanifolds. Michigan Math. J., 33(2):195–199,
1986.

[38] J. J. Nuño Ballesteros, B. Oréfice-Okamoto, and J. N. Tomazella. Equisingularity of

families of isolated determinantal singularities. Math. Z., 289(3-4):1409–1425, 2018.

[39] Antoni Rangachev. Associated primes and integral closure of Noetherian rings. J. Pure
Appl. Algebra, 225(5):Paper No. 106588, 11, 2021.

[40] Marcelo J. Saia. The integral closure of ideals and the Newton filtration. J. Algebraic
Geom., 5(1):1–11, 1996.

[41] Bernard Teissier. Cycles évanescents, sections planes et conditions de Whitney. In Singu-
larités à Cargèse (Rencontre Singularités Géom. Anal., Inst. Études Sci., Cargèse, 1972),
pages 285–362. Astérisque, Nos. 7 et 8. 1973.

[42] Bernard Teissier. Variétés polaires. II. Multiplicités polaires, sections planes, et conditions
de Whitney. In Algebraic geometry (La Rábida, 1981), volume 961 of Lecture Notes in
Math., pages 314–491. Springer, Berlin, 1982.
90 BIBLIOGRAPHY

[43] Bernard Teissier. Monômes, volumes et multiplicités. In Introduction à la théorie des

singularités, II, volume 37 of Travaux en Cours, pages 127–141. Hermann, Paris, 1988.

[44] Lê D ung Tráng. Computation of the Milnor number of an isolated singularity of a complete
intersection. Funkcional. Anal. i Priložen., 8(2):45–49, 1974.

[45] Lê D ung Tráng and Bernard Teissier. Limites d’espaces tangents en géométrie analytique.
Comment. Math. Helv., 63(4):540–578, 1988.

[46] Jean-Louis Verdier. Stratifications de Whitney et théorème de Bertini-Sard. Invent. Math.,

36:295–312, 1976.

[47] C. T. C. Wall. Finite determinacy of smooth map-germs. Bull. London Math. Soc.,
13(6):481–539, 1981.