Math. Z.

242, 481–490 (2002)

Digital Object Identifier (DOI) 10.1007/s002090100352

Residual indices of holomorphic maps relative

to singular curves of fixed points on surfaces
Filippo Bracci
, Francesca Tovena
Dipartimento di Matematica, Università di Roma “Tor Vergata”, Via della Ricerca Scientifica
00133 Roma, Italy (e-mail: [email protected] / [email protected])

Received: 15 May 2000; in final form: 10 July 2001 /

Published online: 1 February 2002 –

c Springer-Verlag 2002

Abstract. Let M be a two-dimensional complex manifold and let f : M →

M be a holomorphic map that fixes pointwise a (possibly) singular, com-
pact, reduced and globally irreducible curve C ⊂ M . We give a notion of
degeneracy of f at a point of C. It turns out that f is non-degenerate at
one point if and only if it is non-degenerate at every point of C. When f is
non-degenerate on C, we define a residual index for f at each point of C.
Then we prove that the sum of the indices is equal to the self-intersection
number of C.

Introduction

In [2], C. Camacho and P. Sad introduced the index of a holomorphic vector

field relative to an invariant non-singular curve and proved an index for-
mula. Their result was generalized by A. Lins Neto [6] to the case of an
algebraic foliation and a (possibly) singular invariant curve in the complex
projective plane. Finally T. Suwa [7] gave a definition of index and proved
a formula when the invariant (singular) curve lies in a generic two dimen-
sional complex manifold. Recently M. Abate [1] (cf. also Sect. 1), studying
discrete dynamical systems, introduced an index for holomorphic self-maps
of a two dimensional complex manifold fixing a smooth compact curve (and
non-degenerate on it), proving an analogue of the Camacho-Sad Theorem.
Here we generalize Abate’s result to the case of singular curves, finding an
analogue of Suwa’s Theorem.


Partially supported by Progetto MURST di Rilevante Interesse Nazionale Proprietà
geometriche delle varietà reali e complesse.
482 F. Bracci, F. Tovena

Let M be a two dimensional complex manifold and C ⊂ M be a reduced,

globally irreducible and compact one dimensional subvariety. We consider
a holomorphic map f : M → M , such that f|C = id|C . Fixing a point
p ∈ C and localizing our study to the germs, fp and Cp , of f and C at p,
we introduce a notion of degeneracy for f on C at p. Roughly speaking, we
say that f is degenerate on C at p ∈ C if one cannot distinguish Cp just
looking at “how much” fp fixes Cp (see Definition 2, 4). It turns out that
the non-degeneracy is a local property, i.e. if f is non-degenerate on C at
p ∈ C then there exists a neighborhood U of p such that for any q ∈ U ,
fq is non-degenerate on Cq . Since C is irreducible, this implies that if f is
non-degenerate on C at a point p ∈ C, then f is non-degenerate at every
point of C (see Proposition 3).
If f is non-degenerate at p ∈ C, we define an index Ind(f, C, p) (see
Definition 5, 6) and we show that the sum of the indices Ind(f, C, p) is
equal to the self-intersection number of C (see Theorem 2). The proof of
this result is quite similar to that in [7] for holomorphic vector fields. We take
a resolution π : M̃ → M of C. The map f induces a holomorphic self-map
f˜ of M̃ which fixes the desingularization C̃ of C and is non-degenerate on
it. Then we compare the indices of f˜ on C̃ and f on C and apply Abate’s
index formula for f˜ on C̃.
The actual effort made in this paper is to find a good definition for the
index of f on C at p ∈ C and prove that it is “natural”, i.e. it is well-
behaving under changes of variables and blow-ups. After that, we formally
have the same ingredients of [7], and we can argue following the same lines.
On the other hand the index here defined is not merely an application of
results on vector fields, since in general it is not possible to associate to f
a (global) foliation leaving C invariant. In conclusion, the analogy between
holomorphic vector fields with an invariant leaf and holomorphic self-maps
with a curve of fixed points which seems to come out from [1] and this paper,
is pretty far from being really understood.

1 The smooth case

In this section we recall Abate’s work [1].

Let C be a compact one-dimensional (smooth) submanifold of M , with
M being a two dimensional complex manifold. Suppose f : M → M is
holomorphic and f|C = id|C . Let p ∈ C and choose an adapted local chart U
with coordinates (x, y) around p, i.e. such that p = (0, 0), C ∩U = {y = 0}.
In these coordinates we can write


f1 (x, y) = x + y m g(x, y)
f (x, y) = (1)
f2 (x, y) = b(x)y + y n+1 h(x, y)
Residual indices of holomorphic maps relative 483

with m ≥ 1, n ≥ 1 and g(x, 0) = 0 (or m = ∞ if g ≡ 0) and h(x, 0) ≡ 0

(or n = ∞ if h ≡ 0). We avoid considering the case f ≡ idM .
The compactness of C implies that b(x) ≡ b(f ) is constant. The map f
is said non-degenerate on C at p if m ≤ n. If b(f ) = 1 or if M is the total
space of a line bundle over C, then Abate proves that if f is non-degenerate
on C at p, it is so independently of the adapted chart chosen and of the point
p ∈ C. In this situation Abate’s residual index is defined as
ιp (f, C) = Res(k(x)dx, 0),
where
f2 (x, y) − b(f )y
k(x) = lim .
y→0 y(f1 (x, y) − x)

This index is independent of the local coordinates chosen, and the Abate’s
index formula is
Theorem 1 (Abate). Let C be a one dimensional compact submanifold of a
two dimensional complex manifold M and let f : M → M be holomorphic
such that f|C = id|C . Assume b(f ) = 1 or that M is the total space of a line
bundle E over C. Assume moreover that f is non-degenerate on C. Then

ιp (f, C) = b(f ) (C · C).
p∈C

If C has a singularity at p ∈ C, there are no adapted charts available at

p, and Abate’s theory doesn’t apply.

2 The residual index in the irreducible case

Let Op be the ring of germs of holomorphic functions at p ∈ M and O∗p the

field of its invertible elements. If H is a holomorphic map defined on a open
set U ⊆ M , then Hp ∈ Op is the germ defined by H at p ∈ U . If H ∈ Op
then (H)p is the ideal generated by H in Op and V(H) is the germ of the
subvariety defined by H. If Cp is a germ of a subvariety at p, then I(Cp ) is
the ideal of Cp (see [4]).
Let C be a (possibly singular) curve whose germ Cp is irreducible at p ∈
M and let f be a holomorphic self-map of M which pointwise fixes C. We
avoid considering the case f ≡ idM . Let U ⊂ M be a open neighborhood
of p and φ : U → C2 a local chart such that φ(p) = (0, 0). If l ∈ Op is a
defining function for Cp at p, then
φ ◦ f ◦ φ−1 = Id + (l ◦ φ−1 )µ G, (2)
for some germ G = (G1 , G2 ) of holomorphic self-map of C2 at (0, 0),
G ≡ 0 on φ(S) and µ ≥ 1. It is easy to see that µ is independent of φ and l.
484 F. Bracci, F. Tovena

Remark 1. We will omit to write explicitly the local chart φ in the formulae:
for instance we write simply f = Id + lµ G instead of (2). We denote by
H  the gradient of H ∈ Op in the given local chart and by < H, G > the
scalar product of two germs H, G of holomorphic self-map of C2 .
Given H ∈ Op , expanding H ◦ f − H we find that
H ◦f −H
≡< H  , G > mod I(Cp ).
lµ
Definition 1. The map f satisfying (2) is said to be non-degenerate at p on
the locally irreducible curve C if
l◦f −l
≡ 0 mod I(Cp ),
lµ
i.e. if < l , G >≡ 0 on the germ Cp .
Remark 2. For k ∈ O∗p we have
(kl) ◦ f − kl l◦f −l
µ
≡ k 1−µ mod I(Cp ).
(kl) lµ
Thus the definition of non-degeneracy of f on C at p is independent of the
defining function l.
Remark 3. Assume that p is a smooth point for C. As in the first section, we
can choose an adapted local chart with coordinates (x, y) around p, so that f
satisfies (1) and y is a defining function for C. Since y◦f −y = (b(x)−1)y+
y n+1 h(x, y), our definition of non-degeneracy coincides with Abate’s one
whenever b(p) = 1. In the case b(p) = 1, the map f is degenerate according
to Definition 1 but it could be non-degenerate according to Abate’s definition.
However, if C has a singularity at some q ∈ C and f is non-degenerate on
C at q then—as a consequence of Lemma 2—it turns out that f is non-
degenerate on C at every point p ∈ C and in particular b(p) = 1 at any
smooth point of C. Therefore, if C is singular, f is non-degenerate on C
according to Definition 1 if and only if it is non-degenerate at one—and
hence any—smooth point according to Abate’s definition.
Definition 2. We say that H ∈ Op is transverse to (f, Cp ) if
H ◦f −H
≡ 0 mod I(Cp ),
lµ
i.e. < H  , G >≡ 0 on Cp .
Therefore f is degenerate on Cp at p if and only if a defining function of Cp
—and hence any— is transverse to (f, Cp ).
Residual indices of holomorphic maps relative 485

Proposition 1. Suppose f is non-degenerate on Cp at p and let l ∈ Op be

a defining function for Cp . A germ H ∈ Op is transverse to (f, Cp ) if and
only if det(H  , l ) ≡ 0 mod I(Cp ).

Proof. Let f = I + lµ G, as in (2). We have det(H  , l ) ≡ 0 mod I(Cp ) if

and only if H  ≡ kl mod I(Cp ) for some k ∈ Op . Now if < H  , G >≡ 0
on Cp then, since G ≡ 0 on Cp , it follows that H  = kl for some k ∈ Op . On
the other hand, if H  ≡ kl mod I(Cp ) then < H  , G >≡ k < l , G >≡ 0
on Cp for f is non-degenerate on Cp .

Note that, if Cp is smooth at p, then the regular curves transverse (in the
usual sense) to Cp at p are actually transverse to (f, Cp ).
Since Cp is irreducible, it admits a local uniformization (see [5]). Namely,
there exists a homeomorphism ϕ : ∆ → Cp (where ∆ = {ζ ∈ C : |ζ| < 1})
such that ϕ(0) = p, ϕ is holomorphic on ∆ − {0} and ϕ (ζ) = 0 for all
ζ ∈ ∆ − {0} (ϕ (0) = 0 if and only if Cp is smooth at p). We denote by
Γ ∈ H1 (Cp − {p}, Z) the class of ϕ(∂∆), where ∂∆ is given by θ → ρeiθ
for θ ∈ [0, 2π] and a fixed 0 < ρ < 1.

Definition 3. Suppose f is non-degenerate on C at p, with Cp reduced and

irreducible at p. Let l ∈ Op be a defining function for Cp and τ ∈ Op be
transverse to (f, Cp ). The residual index of f with respect to Cp is

1 l◦f −l
Ind(f, C, p) := dτ.
2πi Γ l · (τ ◦ f − τ )
Remark 4. If Cp is smooth at p and f is non-degenerate on Cp , then
Ind(f, C, p) = ιp (f, Cp ), the Abate’s index. Indeed, choosing adapted local
coordinates such that Cp = V(y), we can take l = y, τ = x. The cycle Γ is
given by the curve y = 0, x = ρeiθ , for some ρ > 0 small and θ ∈ [0, 2π].
Therefore

1
Ind(f, C, p) = k(x)dx = ιp (f, Cp ).
2πi Γ
Lemma 1. The index Ind(f, C, p) is well-defined, i.e. it is independent of
the defining function l ∈ Op and the transverse τ ∈ Op .

Proof. Let f = I +lµ G, as in (2) and let ϕ : ∆ → Cp be a local uniformiza-

tion of Cp . Since f is non degenerate on Cp then < l , G >≡ 0 mod I(Cp ).
Namely < (l ◦ ϕ), (G ◦ ϕ) >≡ 0 on ∆. Together with < (l ◦ ϕ), ϕ >≡ 0
on ∆, and since we may assume l ◦ ϕ ≡ 0 on ∆ − {0}, this implies that
there exists γ ∈ O∗ (∆ − {0}) such that

G ◦ ϕ(ζ) = γ(ζ)ϕ (ζ) ∀ζ ∈ ∆ − {0}. (3)

486 F. Bracci, F. Tovena

Let h(ζ) := l◦f −l

lµ+1
◦ ϕ(ζ) for ζ ∈ ∆ and let k ∈ O∗p . Using (3) we get
  
(kl) ◦ f − kl l◦f k◦f −k l◦f −l
dτ = dτ + dτ
Γ kl(τ ◦ f − τ ) Γ l k(τ ◦ f − τ ) Γ l(τ ◦ f − τ )
 
< k , G > l◦f −l dτ
= 
dτ + µ+1
k < τ ,G > l < τ , G >
Γ Γ
< (k  ◦ ϕ), (G ◦ ϕ) >
= 
d(τ ◦ ϕ)
∂∆ (k ◦ ϕ) < (τ ◦ ϕ), (G ◦ ϕ) >
   
h dk h h
+ 
d(τ ◦ϕ) = + dζ = dζ,
∂∆ < (τ ◦ ϕ), (G ◦ ϕ) > Γ k ∂∆ γ ∂∆ γ
as wanted.
Example 1. Let Cp = V(l(0,0) ), with l(x, y) = x2 − y 3 , and let f (x, y) =
(x + 3x(x2 − y 3 ), y + 2y(x2 − y 3 )). Then Cp is a irreducible germ of a
curve with singularity at (0, 0), and it is easy to see that f is non-degenerate
on Cp . A local uniformization of Cp is given by ϕ(ζ) = (ζ 3 , ζ 2 ). Therefore
a straightforward calculation gives
 
1 6 − 3y 3 1 6
Ind(f, C, (0, 0)) = dx = dζ = 6.
2πi Γ 3x 2πi ∂∆ ζ

3 The residual index in the reducible case

We let C = ∪Nα=1 Cα be a germ of a reduced curve at p, Cα irreducible in

Op for α = 1, . . . , N and let f be a germ of holomorphic map at p fixing
C.
Definition 4. We say that f is non-degenerate on C at p if it is non-degenerate
on Cα for α = 1, . . . , N .
Let l = l1 · · · lN be a defining function of C with V(lα ) = Cα . Let Γα be
the cycle for Cα given by the local uniformization.
Definition 5. If f is non-degenerate on C then we define the residual index
of f on C with respect to Cα as

1 l◦f −l
Ind(f, Cα , C, p) := dτ,
2πi Γα l · (τ ◦ f − τ )
where τ ∈ Op is transverse to (f, Cα ).
Recall that one of many equivalent definitions of the (local) intersection
number at p of Cα and Cβ for α = β is

1 dlβ
(Cα · Cβ )p = .
2πi Γα lβ
Residual indices of holomorphic maps relative 487

The following relation between Ind(f, Cα , p) and Ind(f, Cα , C, p)—from

which it follows that Ind(f, Cα , C, p) is well defined—holds (cfr. Proposi-
tion (1.4) of [7]):

Proposition 2. If f is non-degenerate on C at p then


Ind(f, Cα , C, p) = Ind(f, Cα , p) + (Cα · Cβ )p (4)
β=α

Proof. Let f = Id + lµ G as in (2). Up to reordering, we can assume α = 1.

Then
l◦f −l l1 ◦ f lN −1 ◦ f lN ◦ f − lN
= ··· ·
l l1 lN −1 lN
l1 ◦ f lN −2 ◦ f lN −1 ◦ f − lN −1 l1 ◦ f − l1
+ ··· · + ... + .
l1 lN −2 lN −1 l1
lj ◦f
Let τ ∈ Op be transverse to (f, C1 ). Since lj = 1 on C1 , we get

 N
 1 
1 l◦f −l lj ◦ f − lj
dτ = dτ +
2πi Γ1 l · (τ ◦ f − τ ) 2πi Γ1 lj · (τ ◦ f − τ )
j=2

1 l1 ◦ f − l1
dτ.
2πi Γ1 l1 · (τ ◦ f − τ )

By definition the last term is Ind(f, Cα , p). As for the other terms, note first
that det(lj , l1 ) ≡ 0 on C1 for j = 1, and therefore lj is transverse to (f, C1 )
for j = 1 by Proposition 1. Thus < lj , G >≡ 0 on C1 . Arguing as in the
 k◦f −k
proof of Lemma 1 for the calculation of the integral Γ k(τ ◦f −τ ) dτ , we find
 
1 lj ◦ f − lj 1 dlj
dτ = = (C1 · Cj )p .
2πi Γ1 lj · (τ ◦ f − τ ) 2πi Γ1 lj

From this we get the formula.

N
Definition 6. We let Ind(f, C, p) := α=1 Ind(f, Cα , C, p).

4 The index formula

Let f : M → M be a holomorphic map of a two dimensional complex

manifold M . Let C ⊂ M be a connected compact reduced and globally
irreducible curve such that f|C = id|C . We say that f is non-degenerate on
C if f is non-degenerate on every p ∈ C.
488 F. Bracci, F. Tovena

Lemma 2. Let M be a two dimensional complex manifold and C a con-

nected and locally irreducible curve in M . If f : M → M is a holomorphic
map such that f|C = id|C then f is non-degenerate on C (i.e. at any point
of C) if and only if it is non-degenerate at just one point.
Proof. Suppose f is non-degenerate on C at q ∈ C. Let U be a (small)
neighborhood of q and l ∈ O(U ) be a defining function for C on U , i.e.
I(C)p = (l)p for any p ∈ U ∩ C (this is possible for the sheaf of ideals of
C is coherent, see [4]). Suppose f = Id + lµ G in U , with µ, G as in (2).
By hypothesis < l , G >≡ 0 mod I(C)q , thus < l , G >≡ 0 mod I(C)p
for any p ∈ C ∩ U by the identity principle. Therefore f is non-degenerate
on C at p for any p ∈ C ∩ U . Since C is connected, it follows that f is
non-degenerate on all of C.
Since C globally irreducible means that C −Sing(C) is pathwise connected,
the previous lemma implies:
Proposition 3. Let M be a two dimensional complex manifold, f : M →
M a holomorphic map and C ⊂ M a connected, reduced, globally ir-
reducible curve. Then f is non-degenerate on C if and only if it is non-
degenerate at just one point of C.
Now we can state our main theorem.
Theorem 2. Let M be a two dimensional complex manifold, f : M → M
a holomorphic map and C ⊂ M a compact, reduced, globally irreducible
curve of fixed points of f . If f is non-degenerate on C then

Ind(f, C, p) = C · C ,
p∈C

where C · C is the self-intersection number of C.


Note that p∈C Ind(f, C, p) is actually a finite sum since C is compact.
In order to prove the theorem we need some local analysis. Suppose U is
a (small) open neighborhood of a singular point p ∈ C and C is irreducible
at p. Let V ⊆ U be a open neighborhood of p such that f (V ) ⊆ U . Let
π : Ũ → U be a quadratic transformation of the point p. That is Ũ is a
two dimensional complex manifold, π : Ũ → U is a proper holomorphic
map, D := π −1 (p) is a projective line and π : Ũ − D → U − {p} is
a biholomorphism. We let C̃ := π −1 (C − {p}) be the strict transform of
C. Thus C̃ ∩ D = {p̃} and the total transform of C under π is given by
π −1 (p) = C̃ + mD, for m = (C̃ · D)p̃ , the multiplicity of C at p. The
map f naturally induces a holomorphic map f˜ : π −1 (V ) → Ũ such that
f˜|D = Id|D and π ◦ f˜ = f ◦ π (see [1]). In particular f˜|C̃ = Id|C̃ . Note that
since f is non-degenerate on C ∩ U and π is a biholomorphism out of D
then f˜ is non-degenerate on C̃.
Residual indices of holomorphic maps relative 489

Lemma 3. In the above situation we have

Ind(f, C, p) = Ind(f˜, C̃, p̃) + m2 . (5)
Proof. Let {l = 0} be a local defining function for C at p. We can choose
local coordinates (x, y) on U and (u, v) on Ũ such that l(0, y) ≡ 0, p =
(0, 0), p̃ = (0, 0) and π(u, v) = (u, uv). Then ˜l := l ◦ π is a defining
function of the total transform π −1 (C) and ˜l(u, v) = l(u, uv) = um t(u, v),
where t(0, 0) = 0, t(0, v) ≡ 0 and t is a defining function of C̃. In particular
t(u, v) = u−m (l ◦ π). Note that u is transverse to (f˜, C̃) at p̃ and that if Γ̃
is the cycle for C̃ at p̃ given by the local uniformization then π∗ (Γ̃ ) = Γ ,
the cycle for C at p. We are now ready to calculate Ind(f˜, C̃, p̃):

t ◦ f˜ − t
2πi Ind(f˜, C̃, p̃) = du
Γ̃ t(u ◦ f˜ − u)

(l ◦ f ◦ π) u−m ◦ f˜ − u−m
= du
Γ̃ (l ◦ π) u−m (u ◦ f˜ − u)
  
−1 ∗ l ◦ π ◦ f˜ − l ◦ π
+ (π ) du
π∗ (Γ̃ ) l ◦ π(u ◦ f˜ − u)
 
du l◦f −l
= −m + dx = 2πi(−m2 + Ind(f, C, p)).
Γ̃ u Γ l(x ◦ f − x)
Now we are in the position to prove Theorem 2.
Proof of Theorem 2. Take a global resolution π : M̃ → M of the singularities
of C. The total transform of C can be written as
N

C̃ + mj Dj ,
j=1

where (1) C̃ is a compact connected non-singular curve and π|C̃ : C̃ → C

is a resolution of singularities of C, (2) each Dj is a projective line and
mj a positive integer, (3) π is biholomorphic out of the exceptional divisor
D := ∪N j=1 Dj , (4) C̃ intersects D at a finite number of points which are non-
singular points of D and each intersection is transverse. The map f induces
a holomorphic map f˜ on M̃ with f˜|C̃ = id|C̃ and f˜ non-degenerate on C̃.
Using Proposition 2 and Lemma 3 it is easy to see that if p ∈ C is a singularity
of C, U a (small) open neighborhood of p such that C ∩ U = ∪N α=1 Cα with
Cα irreducible, then
M

Ind(f, Cα , C, p) = Ind(f˜, C˜α , qα ) + mj (Dj · C̃α )qα , (6)
j=1
490 F. Bracci, F. Tovena

where C̃α is the strict transform of Cα , and (Dj · C̃α )qα = 0 or 1 according
whether Dj intersects C̃α at the point qα ∈ π −1 (p) or not. By (6) the sum
  ˜ 
p∈C Ind(f, C, p) equals p̃∈C̃ Ind(f , C̃, p̃)+ mj (Dj ·C̃). By Theorem
1 and the projection formula we get the result.

Acknowledgements. The authors warmly thank Marco Abate for introducing them into this
subject, and the referee for many useful comments improving the aspect of the paper.

References

1. M. Abate: The residual index and the dynamics of holomorphic maps tangent to the
identity. Duke Math. J. (1) 107 (2001), 173–207
2. C. Camacho, P. Sad: Invariant varieties through singularities of holomorphic vector
fields. Ann. of Math. (2) 115 (1982), 579–595
3. W. Fulton: Intersection Theory, 2nd edn. Springer, 1998
4. R. Gunning, H. Rossi: Analytic functions of several complex variables. Prentice-Hall,
1965
5. H. B. Laufer: Normal two dimensional singularities. Ann. Math. Stud. 71 (1971)
6. A. Lins Neto: Algebraic solutions of polynomial differential equations and foliations in
dimension two. Holomorphic Dynamics (Mexico 1986), Lecture Notes in Math., vol.
1345, pp. 192–232. Springer, 1988
7. T. Suwa: Indices of holomorphic vector fields relative to invariant curves on surfaces.
Proc. Amer. Math. Soc. 123 (1995), 2989–2997