See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263913970

ON THE SINGULAR SCHEME OF CODIMENSION ONE HOLOMORPHIC

FOLIATIONS IN ℙ3

Article  in  International Journal of Mathematics · July 2010

DOI: 10.1142/S0129167X1000601X · Source: arXiv

CITATIONS READS
13 104

2 authors:

Luis Giraldo Antonio Jesus Pan-Collantes

Complutense University of Madrid IES San Juan de Dios
26 PUBLICATIONS   117 CITATIONS    4 PUBLICATIONS   13 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Antonio Jesus Pan-Collantes on 01 April 2019.

The user has requested enhancement of the downloaded file.

 ON THE SINGULAR SCHEME OF CODIMENSION ONE
HOLOMORPHIC FOLIATIONS IN P3
arXiv:0812.3369v1 [math.AG] 17 Dec 2008

LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

Abstract. In this work, we begin by showing that a holomorphic foliation

with singularities is reduced if and only if its normal sheaf is torsion free. In
addition, when the codimension of the singular locus is at least two, it is shown
that being reduced is equivalent to the reflexivity of the tangent sheaf. Our
main results state on one hand, that the tangent sheaf of a codimension one
foliation in P3 is locally free if and only the singular scheme is a curve, and
that it splits if and only if that curve is arithmetically Cohen-Macaulay. On
the other hand, we discuss when a split foliation in P3 is determined by its
singular scheme.

1. Introduction

The study of holomorphic foliations with singularities in complex manifolds has

attracted a lot of interest over the last 40 years. There are very good sources to
learn about the different aspects of the theory, and we just mention here the classical
paper [1] where a general theory in terms of coherent sheaves was developed, and
the recent book [19].

The tangent sheaf of the foliation, its normal sheaf and singular locus are the key
elements for this study. Needless to say, a very important problem is to analyze if
the tangent sheaf of the foliation is locally free as well as further properties (as the
splitting) when it is. These properties have important consequences, as for example
the ones recently noted in [3] and [6], where it has been proved that the fact that the
tangent sheaf splits, along with some properties of the singular locus, give stability
under deformations of the foliation and make it possible to characterize certain
components of the space that parameterizes holomorphic foliations.

The first section of this paper is devoted to the study of the tangent and normal
sheaves. We prove some key technical results for the main goals of our research that,
even when mentioned in some references, were not explicitly found in the literature

Date: September 2008.

1991 Mathematics Subject Classification. Primary 32S65, 37F75; secondary 14F05.
Key words and phrases. Holomorphic foliations, reflexive sheaves, split vector bundles.
Partially supported by Plan Nacional I+D grant no. MTM2004-07203-C02-02 and MTM2007-
61124, Spain. The second author also supported by the FPU grant no. AP2006-03911, Ministerio
de Educación y Ciencia, Spain.
1
2 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

on singular foliations. We think that they could be useful for other people working
in the field. Namely, we prove that a foliation is reduced if and only if its normal
sheaf is torsion free and that, under the assumption that the singular locus is of
codimension at least two, a foliation is reduced if and only if the tangent sheaf is
reflexive.

As an immediate application of the reflexivity of the tangent sheaf, we give an

affirmative answer to the question posed by Suwa in [19]: Is the tangent sheaf to a
1 dimensional reduced foliation locally free?

The main results contained in this paper establish a link between the above
mentioned properties of the tangent sheaf and some algebro-geometric properties
of the singular scheme. For codimension one foliations in P3 , using some results by
Roggero on reflexive sheaves (see [17]) we prove that the tangent sheaf is locally
free if and only if the singular scheme is a curve, i.e. a Cohen-Macaulay scheme
of pure dimension one (Theorem 3.2); moreover, this sheaf splits if and only if the
singular scheme is an arithmetically Cohen-Macaulay curve (Theorem 3.3).
We also prove (Theorem 3.5) an algebraic characterization of the splitting of
a locally free tangent sheaf. It is given by the fact that the homogeneous ideal
generated by the coefficients of a form defining the foliation, that defines the singular
scheme, is saturated.

An interesting problem is to decide when the singular scheme uniquely deter-

mines the foliation. That is, given a foliation defined by a 1-form ω, and having Z
as singular scheme, is it true that if the 1-form ω ′ defines a foliation with Z as its
singular scheme then we have ω = λω ′ for λ ∈ C∗ ?.
In dimension 2, the problem was completely solved recently by Campillo and
Olivares in [4]: for reduced foliations of degree d 6= 1 there are no two different
foliations with the same singular scheme.

With our approach, we get simpler proofs of their main results, and more impor-
tantly we prove some results in the P3 case. Namely, we prove unique determination
by the singular scheme for codimension one degree d foliations in P3 with split tan-
gent sheaf, when the splitting type is OP3 (a) ⊕ OP3 (b), with a, b ≤ −1.
We also show that if there is a subbundle of the tangent bundle of the foliation
of the form OP3 (1), then it is a linear pull-back of a foliation in a plane, and so is
determined by its singular scheme if d 6= 1.
When there is a OP3 subbundle of F , we just show some examples that illus-
trate the situation: there are foliations determined by the singular scheme (excep-
tional foliations, see [3]), and others that are not: the logarithmic foliations of type
L(1, 1, 1, 1).
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 3

2. Basic facts concerning the tangent sheaf

2.1. Preliminaries. In this section we recall the basics of the theory. The following
definition is taken from [1]:
Definition 2.1. Let M be a complex n-dimensional manifold, and let T M be its
tangent sheaf.

• A codimension k holomorphic foliation with singularities in M is an injective

morphism of sheaves ϕ : F −→ T M , such that ϕ(F ) is an integrable coherent
subsheaf of T M , of rank n − k. The integrability means that for each point x ∈ M ,
ϕ(F )x is closed under the bracket operation for vector fields.
Notation: we will denote by Fϕ the foliation given by ϕ : F −→ T M .
• The sheaves F and the NFϕ := T M/ϕ(F ) are, respectively, the tangent and
normal sheaves of the foliation.
• Sing(Fϕ ) = {x ∈ M | (NFϕ )x is not a free OM,x − module} is the singular set of
the foliation Fϕ .
The following facts can be found, for instance, in [16]:
(1) The normal sheaf is coherent.
(2) By definition, Sing(Fϕ) is the singular set of the sheaf NFϕ . It is a closed
analytic subvariety of M of codimension at least one.
(3) There is a rank n − k holomorphic vector bundle F on M \ Sing(Fϕ) whose
sheaf of sections is ϕ(F )|M\Sing(Fϕ ) .
Note also that being a subsheaf of a locally free sheaf, F is torsion free and
therefore its singular set S(F ) = {x ∈ M | Fx is not a free OM,x − module} is of
codimension ≥ 2.
Moreover, it is also clear that S(F ) ⊂ Sing(Fϕ ). Indeed, if x 6∈ Sing(Fϕ ) there
is a neighbourhood V of x such that the sequence
0 −→ F (V ) −→ T M (V ) −→ NFϕ (V ) −→ 0
is an exact sequence of OM (V ) modules, and the second and third modules are
free. Hence (see [8], Appendix 3) the sequence splits since NFϕ (V ) is projective
and F (V ) is a direct summand of a free module, and so it is projective. We conclude
that, for every p ∈ V , Fp is a free OM,p module. In particular, x 6∈ S(F ).

In the study of singular holomorphic foliations, it is usual to deal with reduced

ones. Let us recall their definition ([1]):
Definition 2.2. Let Fϕ be a foliation:
• ϕ(F ) is full if given any open set U ⊂ M , and γ a holomorphic section of T M |U
such that γ(x) ∈ ϕ(F )x for each x ∈ U ∩ (M \ Sing(Fϕ)), then at each point
p ∈ U ∩ Sing(Fϕ) the germ of the holomorphic vector field γ is in ϕ(F )p .
• Fϕ is reduced if ϕ(F ) is full.
4 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

2.2. Some remarks. For the sake of completeness and lack of reference, we now
state and prove some properties of the tangent and normal sheaves of a holomorphic
foliation. Recall that a coherent sheaf F is reflexive if and only if it is isomorphic
to its bidual F ∗∗ ([16], [12]).

Remark 2.3. The foliation Fϕ is reduced if and only if NFϕ is torsion free. If the
singular locus is of codimension at least 2, the foliation is reduced if and only if F
is a reflexive sheaf.

Indeed, assume that Fϕ is reduced, hence ϕ(F ) is full. Torsion freeness is a

local property and locally free sheaves are torsion free, so it is enough to show that
(NFϕ )x is a torsion free OM,x -module for x ∈ Sing(Fϕ).
Let us suppose, on the contrary, that it is not. Then, there is a nonzero element
m ∈ (NFϕ )x and a nonzero element a ∈ OM,x such that am = 0 in (NFϕ )x . Let U
be a small enough neighborhood of x in M , so that we can take α ∈ OM (U ) and
γ ∈ T M (U ) such that αx = a, πx (γx ) = m.
NFϕ is locally free in V = U ∩ (M \ Sing(Fϕ)), hence torsion free and we have
that
γ(U ∩ (M \ Sing(Fϕ )) ∈ ϕ(F )(U ∩ (M \ Sing(Fϕ)).
As ϕ(F ) is full, we have that γ(U ) ∈ ϕ(F )(U ), which gives m = 0 in (NFϕ )x , a
contradiction.

Conversely, suppose that NFϕ is torsion free. If U ⊂ M is an open subset and

γ is a section of T M (U ) such that γx ∈ (ϕ(F ))x for each x ∈ U ∩ (M \ Sing(Fϕ )),
then for a nonzero element f ∈ OM (U ) vanishing on the analytic set Sing(Fϕ),
and for each p ∈ U ∩ Sing(Fϕ ), (f γ)p = fp γp gives the zero element in (NFϕ )p .
Hence, γp ∈ (ϕ(F ))p . Thus, ϕ(F ) is full and the foliation is reduced.

The second part of the statement follows from the first, using that a torsion free
sheaf F is reflexive if and only if there is a locally free sheaf E such that F ⊂ E and
E/F is torsion free (see [15, p.61]).

We can make the following further remarks:

• The singular locus of a reflexive sheaf is of codimension ≥ 3 (see [12], [16]).
Hence, the same is true for the singular set of the tangent sheaf of a reduced
holomorphic foliation.
• The tangent sheaf of a foliation F is torsion free and we have ([16]), that
F ⊂ F ∗∗ . Hence, if the singular locus of the foliation is of codimension ≥ 2
we can consider the reduced foliation given by the double dual of F .

Let us finally obtain a consequence of Remark 2.3. Suwa, in [19], made a dis-
tinction between 1-dimensional singular holomorphic foliations and “foliations by
curves”, a concept that was defined in [9] where it is shown that they correspond
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 5

to holomorphic foliations with singularities with locally free rank 1 tangent sheaf.
We now answer the question that he posed in [19], Remark 1.9, page 179 :
The tangent sheaf of a dimension one reduced foliation on a complex n-dimensional
manifold is a line bundle

It is a consequence of our Theorem 2.3, and of the fact that a reflexive sheaf of
rank one is a line bundle (see Lemma 1.1.15, page 154, in [16]).
We obtain an immediate proof of Proposition 1.7 in [19] (with no use of [14] or
[18]):

Let Fϕ be a holomorphic foliation with singularities.

(1) If it is reduced, the codimension of its singular locus is ≥ 2;
(2) if its tangent sheaf F is locally free, and the codimension of the singular
locus of Fϕ is at least 2, then the foliation is reduced.
The first assertion follows from Proposition 2.3, and so does the second after
noting that locally free sheaves are reflexive.

3. Tangent sheaf vs. singular scheme

From now on, all the foliations that we will consider will be defined in projective
space, of codimension one, reduced, and with singular set of codimension ≥ 2. In
this section we characterize when the tangent sheaf is locally free or split (i.e.,
direct sum of line bundles), in terms of the geometry of the singular scheme of the
foliation.
Recall that a degree d codimension one holomorphic foliation with singularities
is defined by a global section ω ∈ H 0 (Pn , Ω1Pn (d + 2)). The form ω satisfies the
integrability condition ω ∧dω = 0. The degree is the number of tangencies (counted
P
with multiplicities) of a generic line with the foliation. We can write ω = Fj dzj ,
where the Fj are homogeneous polynomials of the same degree deg Fj = d + 1
Pn
satisfying j=0 zj Fj = 0.
The singular set of the foliation is given by F0 = · · · = Fn = 0, and it has
a natural structure of closed subscheme of Pn , given by the homogeneous ideal
(F0 , . . . , Fn ). Recall ([11]) that two homogeneous ideal I, J define the same pro-
jective scheme X if and only if they have the same saturation, i.e., I sat = J sat ,
where
[ M
I sat = (I : (z0 , . . . , zn )l ) = H 0 (Pn , IX (n)),
l≥0 n

IX being the ideal sheaf of the subscheme X.

The singular scheme can be also obtained (see [7]) as the closed subscheme of
Pn whose ideal sheaf is the image of the co-section

ω ∗ : (ΩPn (d + 2))∗ −→ O Pn
6 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

To simplify the notation, let us write Z := Sing(F ϕ ), and I Z = Im(ω ∗ ) to denote

the ideal sheaf defining the singular subscheme in Pn . The induced map

T Pn −→ I Z ⊗ O Pn (d + 2)

is surjective, and moreover, it makes the following sequence exact:

ϕ
(1) 0 −→ F −→ T Pn −→ I Z ⊗ OPn (d + 2) −→ 0.

Observe that NF = I Z ⊗ OPn (d + 2).

We will focus on the case n = 3. As we have pointed out, the tangent sheaf F is
reflexive. Its first Chern class can be computed from (1) and equals 2 − d.
We will make use of a Theorem in [17], that we state in the particular case that
will be of use for us. Let ǫ = d2 − 1 when d even, and ǫ = d−12 − 1 for d odd. Let
hi (F (ℓ)) := dimC H i (P3 , F (ℓ)).

Theorem 3.1 ([17]). The tangent sheaf F is locally free if and only if h2 (F (p)) = 0,
for some p ≤ ε − 2.
If h2 (F (p)) = 0 for p = ε − 3 when d is even, or for p ∈ {ε − 4, ε − 3, ε − 2} when
d is odd, then F splits.

We now obtain some results relating algebro-geometric properties of the singular

scheme to the fact that the tangent sheaf is locally free, and also to its being split.
For us, a curve will be an equidimensional, locally Cohen-Macaulay, dimension one
subscheme of Pn . Given a curve C in P3 , the Hartshorne-Rao module is defined to
be
X
H 1 (P3 , I C (k)).
k∈Z
It is well known that this module is finite dimensional as a C-vector space, and that
is trivial if and only if C is arithmetically Cohen-Macaulay.
First, we characterize those foliations having locally free tangent sheaf.

Theorem 3.2. The tangent sheaf F is locally free if and only if the singular scheme
Z is a curve.

Proof. Suppose that the tangent sheaf is locally free, then Z has no isolated points.
The reason is that, in this case, the singular scheme is given by the vanishing of the
2 × 2 minors of the 3 × 2 matrix corresponding to the local expression of ϕ. Hence,
Z is locally determinantal and Cohen-Macaulay [8]. As the codimension of Z is at
least two, we are done. Observe that the singular scheme has no embedded points
and no isolated points.

To prove the converse, suppose that Z satisfies the properties in the statement.
We will make use of Theorem 3.1.
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 7

By considering the short exact sequence (1), after tensoring with O P3 (−q), and
taking the long exact sequence of cohomology we get:
· · · −→ H 1 (P3 , F (−q)) −→ H 1 (P3 , T P3(−q)) −→ H 1 (P3 , I Z (d + 2 − q)) −→
−→ H 2 (P3 , F (−q)) −→ H 2 (P3 , T P3(−q)) −→ H 2 (P3 , I Z (d + 2 − q)) −→ · · ·
Observe that, from Bott’s formula (see [16]), if we take q > 4, then

h1 (P3 , T P3 (−q)) = h2 (P3 , T P3 (−q)) = 0 ,

so
H 2 (P3 , F (−q)) ≃ H 1 (P3 , I Z (d + 2 − q)).
Now, since the Harshorne-Rao module has finite dimension, we have that

H 1 (P3 , I Z (d + 2 − q)) = 0

for q large enough, and hence for some −q < ε − 2. Hence, F is locally free.

Now, we characterize foliations whose tangent sheaf splits.

Theorem 3.3. Suppose d > 1. F splits if and only if Z is an arithmetically

Cohen-Macaulay curve.

Proof. Suppose F splits. In particular, it is locally free, and hence Z is a curve.

Consider the exact sequence
ϕ
0 −→ F (p) −→ T P3 (p) −→ IZ (d + 2 + p) −→ 0

and the cohomology exact sequence:

(2) · · · → H 1 (P3 , T P3 (p)) → H 1 (P3 , I Z (d + 2 + p)) → H 2 (P3 , F (p)) → . . .

Bott’s formula tells us that H 1 (P3 , T P3(p)) = 0 for every p. In addition, since
F splits, H 2 (P3 , F (p)) = 0 for every p, too. Then the Hartshorne-Rao module is
trivial.

Conversely, suppose Z is an arithmetically Cohen-Macaulay curve. We use

Proposition 3.1. Take p = ε − 3 if d is even, and p = ε − 2 if d is odd. It will
be sufficient to show that
H 2 (P3 , F (p)) = 0.
Consider the following piece of the long exact cohomology sequence (2):

· · · → H 1 (P3 , I Z (d + 2 + p)) → H 2 (P3 , F (p)) → H 2 (P3 , T P3 (p)) → . . .

and note that since the Hartshorne-Rao module is trivial, we just need to check
that
H 2 (P3 , T P3 (p)) = 0.
8 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

Suppose first that d is even; then, ε−3 = d2 −4 and by Bott’s formula, H 2 (P3 , T P3 ( d2 −
4)) = 0 for d > 0. If d is odd, ε − 2 = d−7 2 3 3 d−7
2 , and so, H (P , T P ( 2 )) = 0 for
d > −1.

It is well known that every arithmetically Cohen-Macaulay curve is connected,

so we have:

Corollary 3.4. If the tangent sheaf F splits, then Z is connected.

We now obtain another characterization for the splitting of the tangent sheaf of
a foliation:

Theorem 3.5. F splits if and only if it is locally free and the ideal I = (F0 , F1 , F2 , F3 )
is saturated.

Proof. From the exact sequence (1) and Euler’s sequence

0 −→ OP3 −→ O ⊕4 3
P3 (1) −→ T P −→ 0

we get the exact sequence:

(3) 0 −→ F ⊕ O P3 −→ O⊕4
P3 (1) −→ I Z ⊗ O P3 (d + 2) −→ 0.

If F splits, H 1 (P3 , F (q) ⊕ OP3 (q)) = 0 for every q, and so we have a surjective
map:

(4) H 0 (P3 , O⊕4 0 3

P3 (1 + q)) −→ H (P , I Z ⊗ O P (d + 2 + q))
3

By simple inspection of the map inducing this surjection we conclude that

H 0 (P3 , I Z ⊗ OP3 (d + 2 + q)) = Id+2+q

for every q ∈ Z, and therefore I is saturated.

Conversely, if the ideal I is saturated the map (4) is surjective. Since

H 1 (P3 , O⊕4
P3 (1)) = 0,

we get that H 1 (P3 , F (q) ⊕ O P3 (q)) = 0, and hence H 1 (P3 , F (q)) = 0, for every q.
In order to apply Horrocks’ criterion for the splitting of the vector bundle F
(see [16], p. 39), we just need to prove also that H 2 (P3 , F (q)) = 0 for every q. By
Serre’s duality

H 2 (P3 , F (q)) = H 1 (P3 , F ∗ (−q − 4)) for every q.

But (this can be found in Hirzebruch [13])

F ∗ ≃ det F ∗ ⊗ F ,

and so the vanishing follows from the previous calculation. 

 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 9

Let us observe that Theorem 3.5 translates into an algebraic setting the splitting
of the tangent sheaf of the foliation. Thus, it could be an effective tool in trying to
solve the following important open problem, posed in [3]: Given a foliation in P3
with locally free tangent sheaf, does it split?
P3
The question above can be restated as follows: let ω = i=0 Fi dzi define a
3
degree d foliation in P , with locally free tangent sheaf (i.e. the scheme defined by
F0 = · · · = F3 = 0 is a curve), is the ideal (F0 , F1 , F2 , F3 ) saturated? Furthermore,
is this ideal saturated for every reduced foliation?

P3
We note that given a reduced foliation defined by ω = i=0 Fi dzi , using for ex-
ample the algorithm nsatiety( ) implemented in the Singular library “noether.lib”
[10], one immediately checks whether the ideal (F0 , F1 , F2 , F3 ) is saturated or not.

Let us point out that Theorem 3.3 in the paper of Campillo and Olivares ([4])
can be restated as follows: for any foliation in P2 defined by a 1-form
2
X
ω= Fi dxi
i=0

with singular set of codimension at least two, the ideal (F0 , F1 , F2 ) is saturated.
Our approach gives a much simpler proof of that result, since the tangent sheaf of
such a foliation is a line bundle.

4. On the determination of a split foliation by the singular scheme

Now we face the problem of deciding when a codimension one foliation in P3 is

determined by its singular scheme, as stated in the first Introduction. In P2 , the
problem was completely solved in [4], Theorem 3.5: a degree r = 0, or r ≥ 2 foliation
is uniquely determined by its singular subscheme; for degree r = 1 they construct
a 1-dimensional family of distinct foliations with the same singular subscheme.

In P3 , we study this problem just for foliations whose tangent sheaf splits (recall
that this is trivially true for foliations in the plane). We develop a method to
deal with the problem, and obtain affirmative results for certain splitting types
(Theorem 4.3, Theorem 4.5). We also show examples for the remaining splitting
types, showing that the answer depends on the particular singular scheme Z.

We begin by noting the following

Remark 4.1. Two split foliations defined by F0 dz0 + F1 dz1 + F2 dz2 + F3 dz3 and
G0 dz0 + G1 dz1 + G2 dz2 + G3 dz3 , with the same singular scheme Z are of the same
degree d. Indeed, it is enough to observe that by Theorem 3.5, the homogenous ideal
(F0 , . . . , F3 ) = (G0 , . . . , G3 ) is saturated.
10 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

We now prove a lemma that contains a basic idea for the problem. Recall (see,
for example [6]) that a codimension one holomorphic distribution is defined just in
the same way as a foliation, removing the integrability condition.

Lemma 4.2. Let Fϕ be a foliation defined by a projective integrable 1-form ω =

P3 ′
i=0 Fi dzi with split tangent sheaf, and singular scheme Z. Any distribution Fϕ′
with singular scheme Z is split, and induces a linear syzygy
ℓ0 F0 + ℓ1 F1 + ℓ2 F2 + ℓ3 F3 = 0 , ℓi ∈ C[z0 , z1 , z2 , z3 ], deg ℓi = 1 , i = 0, . . . , 3.

Proof. First note that Theorem 3.3 applies for distributions, as the integrability
does not enter in the proof; thus we conclude that F ′ splits. Then
F ′ = OP3 (a′ ) ⊕ OP3 (b′ ) , with a′ + b′ = 2 − d, and a′ ≤ b′ ≤ 1.
Note that a′ ≤ b′ ≤ 1 follows from the stability of T P3 (see [16]). Furthermore,
a = a′ and b = b′ , as can be deduced by considering the long cohomology sequences
obtained from (1) and
ϕ′
(5) 0 −→ F ′ −→ T P3 −→ I Z ⊗ O Pn (d + 2) −→ 0,
as we have that h0 (P3 , IZ (ℓ)) equals
h0 (P3 , T P3 (ℓ − d − 2)) − h0 (P3 , F (ℓ − d − 2)) ,
and also
h0 (P3 , T P3 (ℓ − d − 2)) − h0 (P3 , F ′ (ℓ − d − 2))
for every ℓ ∈ Z.
The singular subscheme of the distribution Fϕ′ ′ is Z, and it is defined by the
ideal (F0′ , F1′ , F2′ , F3′ ), where the Fi′ are degree d + 1 homogeneous polynomials such
that ω ′ = 3i=0 Fi′ dzi defines Fϕ′ ′ .
P

Hence, there is a 4 × 4 matrix M = (aij ) ∈ GL (4, C) such that

   ′ 
F0 F0
 F   F′ 
 1   1 
(6) M · = .
 F2   F2′ 
F3 F3′
P3 ′
Now, the Euler condition j=0 zj Fj = 0, gives
 
F0
 F 
 1
(z0 z1 z2 z3 ) · M ·   = 0,

 F2 
F3
which produces a linear syzygy:
ℓ0 F0 + ℓ1 F1 + ℓ2 F2 + ℓ3 F3 = 0.
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 11

Now then, H 0 (P3 , F ⊕ OP3 ) is the vector space of linear syzygies because of (3).
So such a distribution as in the Lemma above gives a 1-dimensional linear subspace
of the vector space of linear syzygies H 0 (P3 , F ⊕ OP3 ). Conversely, given a linear
subspace of H 0 (P3 , F ⊕ OP3 ) of dimension one, we can take a generator and express
it in the form
(z0 z1 z2 z3 ) · M .
If M ∈ GL(4, C), then equation (6) gives coefficients for a homogeneous 1-form
defining a distribution with the same singular scheme. Therefore, we can assure
that the family of distributions with the same singular scheme as the foliation
F ϕ is parameterized by a Zariski open subset DF ϕ of P(H 0 (P3 , F ⊕ O P3 )) ⊂
P(H 0 (P3 , OP3 (1)⊕4 )), obtained after removing the algebraic subset corresponding
to non-invertible matrices.
Therefore, foliations sharing the singular scheme Z correspond to an algebraic
subset of DF ϕ , defined by the equations expressing the integrability condition.

Let us just point out that if we find a basis of H 0 (P3 , F ⊕ OP3 ) it is easy to write
explicit equations for the integrability condition.
Now we present our first result on the characterization of a foliation by its sin-
gular scheme:

Theorem 4.3. Let Fϕ be a degree d reduced foliation in P3 , with F a rank two

split vector bundle, defined by a projective integrable one form ω = 3i=0 Fi dzi . Let
P

Z be its singular subscheme. Suppose that

F = OP3 (a) ⊕ OP3 (b) , with a ≤ b ≤ −1, a + b = 2 − d.

Then, if Fϕ′ ′ is another foliation in P3 with the same singular subscheme Z, defined
by the form ω ′ = 3i=0 Fi′ dzi , we have ω = λω ′ for λ ∈ C∗ .
P

Proof. By Lemma 4.2, we get a linear syzygy

ℓ0 F0 + ℓ1 F1 + ℓ2 F2 + ℓ3 F3 = 0.

From the sequence (3), we conclude that this syzygy is a multiple of the Euler
P3
relation j=0 zj Fj = 0, so there is a λ ∈ C∗ with M = λ Id and also ω ′ = λω.

We analyze now the other possible splitting types. To start with we prove,

Proposition 4.4. Let Fϕ be a degree d reduced foliation in P3 . Fϕ is the linear

pull-back of a degree d reduced foliation in P2 if and only if F = OP3 (1)⊕OP3 (1−d) .

Proof. When Fϕ is a linear pull-back it is known (see [6]) that the tangent sheaf
splits, and that the splitting type is as in the statement of the theorem. Indeed,
note that OP3 (1) is the tangent sheaf of the foliation whose leaves are the fibers of
the projection, and the splitting is given by the projection.
12 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

For the converse, suppose that the tangent sheaf of Fϕ splits as

F = OP3 (1) ⊕ OP3 (1 − d) .

Then, by first tensoring (3) with OP3 (−1) and considering the exact cohomology
sequence, we get four complex numbers a0 , . . . , a3 (not all of them equal to zero)
such that a0 F0 + a1 F1 + a2 F2 + a3 F3 = 0. Suppose a0 6= 0.
If we change coordinates:

z0 = a0 z0′



 z = z′ + a z′

1 1 1 0


 z2 = z2 + a2 z0′
′

z3 = z3′ + a3 z0′


we get a new expression for the form defining Fϕ :

η = G1 dz1′ + G2 dz2′ + G3 dz3′

where G1 , G2 and G3 are homogeneous in the new variables.

Let us prove that Gi ∈ C[z1′ , z2′ , z3′ ]. To do that, we go to the affine open z3′ = 1,
and get a 1-form

η3 = G1 (z0′ , z1′ , z2′ , 1)dz1′ + G2 (z0′ , z1′ , z2′ , 1)dz2′

that defines the foliation in that affine chart. To simplify, we write:

f (z0′ , z1′ , z2′ ) = G1 (z0′ , z1′ , z2′ , 1), g(z0′ , z1′ , z2′ ) = G2 (z0′ , z1′ , z2′ , 1).

As the singular set of the foliation is of codimension two, f and g have no common
irreducible factor.
Note that η3 is integrable (i.e. η3 ∧ dη3 = 0), and hence we have:
∂f ∂g
′ g= ′ f.
∂z0 ∂z0
From the equality above, and as f and g have no common irreducible factor, it
∂f ∂f ∂g
follows that f divides ∂z ′ and hence
∂z0′ = 0. Analogously, ∂z0′ = 0. Thus f and g
0
do not depend on z0′ . Proceeding in a analogous way in another affine open set, we
can deduce that G0 , G1 and G2 are in C[z1′ , z2′ , z3′ ].
We can finally conclude that Fϕ is a linear pull-back of the foliation given by

η = G1 dz1′ + G2 dz2′ + G3 dz3′

in the projective plane z0′ = 0.

Now, we come back to the problem of deciding whether the singular scheme
characterizes split foliations and prove:
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 13

Theorem 4.5. Let Fϕ be a degree d 6= 1 reduced foliation in P3 , with

F = OP3 (1) ⊕ OP3 (1 − d) ,

and let Z be its singular scheme. There is no other foliation with singular scheme
Z.

Proof. Suppose there is another foliation Gψ with singular scheme Z. As Z is

an arithmetically Cohen-Macaulay curve, its tangent sheaf G splits, with the same
splitting type as that of F . Hence, from the previous Proposition 4.4 both Fϕ and
Gψ are linear pull-backs of two foliations of a projective plane in P3 .
As Z is a cone, taking any plane H ⊂ P3 not passing through the vertex p of Z,
we obtain both Fϕ and Gψ as linear pull-backs from p of two degree d foliations in
H (see for example Section 2 in [5]). Now, as d 6= 1, by Theorem 3.5 in [4] these
two foliations are the same, as their common singular subscheme is H ∩ Z.

Finally, after Theorems 4.3 and 4.5, there is just one splitting type for each
degree d that remains to be considered:

F = OP3 ⊕ OP3 (2 − d) .

We can make the following remarks:

• If d = 1, a foliation with splitting type F = OP3 ⊕ OP3 (1) is a linear pull-back
(by Proposition 4.4). The intersection with a general plane H gives a degree 1
foliation in H whose singular scheme determines Z. However (see Remark 3.6 in
[4]), there is a one dimensional family of different foliations in the plane H with
the same singular scheme, giving different foliations in P3 by pull-back with the
same singular scheme as Fϕ . In fact, for any degree one foliation in P2 , using the
immediate translation for dimension 2 of Lemma 4.2 we conclude that the family
of plane foliations with the same singular scheme is of dimension one.

• If d = 2, for a foliation Gψ with split tangent sheaf G = OP3 ⊕ OP3 there is a pair
of linear vector fields X, and Y in ψ(G) generating a Lie algebra. According to the
classification, this can be abelian or isomorphic to the affine Lie algebra.
In the abelian case, we can diagonalize simultaneously both vector fields, and the
resulting foliation is logarithmic of type L(1, 1, 1, 1), as can be seen in [6]. Note that
the other possible type of degree 2 logarithmic foliations L(1, 1, 2) have a tangent
sheaf which is not locally free, as it contains isolated points (see [7]).
A foliation in L(1, 1, 1, 1) is defined by a form (see [2]):
3
X dℓi
ω = ℓ0 ℓ1 ℓ2 ℓ3 λi ,
i=0
ℓi
14 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

where the ℓi are linear forms in general position, and the scalars satisfy
3
X
(7) λi ∈ C∗ , λi = 0.
i=0

Its singular scheme Z is the given by six lines giving the edges of a tetrahedron
obtaining by intersecting any two of the ℓi .
Now, the same linear forms ℓi with different choices of scalars satisfying (7)
provide distinct foliations with the same singular scheme, as these determine the
holonomy of the foliation.

If the Lie algebra generated by the linear vector fields is isomorphic to the affine
Lie algebra, after a linear change of coordinates we can choose generators X ′ and
Y ′ with
[X ′ , Y ′ ] = Y ′ .
Thus, we are in the setting of the exceptional component of degree 2 foliations in
P3 , introduced by Cerveau and Lins Neto (see [5]): a general member is given by
a split foliation ([3]), with tangent sheaf F = OP3 ⊕ OP3 , and they prove that this
foliation Fϕ is rigid (i.e. an open dense subset of the component is described by
the action of the group P GL(4, C) on that foliation).
The singular scheme Z has three irreducible components: a line ℓ, a conic C
tangent to ℓ at a point p, and a twisted cubic with the line ℓ as an inflection line
at p.
In this setting, the reader can check that the equations expressing the integrabil-
ity, obtained from the basis of H 0 (P3 , F ⊕ OP3 ) given by X ′ , Y ′ (see their explicit
expression in [5]) and the radial vector field R have a unique solution: the one
corresponding to the foliation F ϕ itself.

• If d > 2, we have that dim H 0 (P3 , F ⊕ OP3 ) = 2. In [3], it is proven that there
is an exceptional component of the space of (codimension one) degree d foliations
in P3 , whose general member is a foliation F ϕ with split tangent sheaf of the form
OP3 ⊕ OP3 (2 − d).
F ϕ is associated to a representation of the affine Lie algebra, in such a way that
in an affine open set, F ϕ is defined by a one form ω = iS iX (d V ol) , where S is a
linear vector field and X is quasi-homogeneous:
∂ ∂ ∂
S = (1 + d + d2 )z1 + (1 + d)z2 + z3 ,
∂z1 ∂z2 ∂z3
∂ ∂ ∂
X = (1 + d + d2 )z2d + (1 + d)z3d + .
∂z1 ∂z2 ∂z3
Thus, we can take R and S as a basis of the vector space H 0 (P3 , F ⊕ OP3 ).
The vector field S corresponds to a linear syzygy that can be interpreted as
a matrix M ∈ GL(4, C) acting on the coefficients of the projective form ω =
 ON THE SINGULAR SCHEME OF PROJECTIVE FOLIATIONS IN P3 15

P3
i=0 Fi dzi , extending

ω = (1+d)(z2 −z3d+1 )dz1 −(1+d+d2 )(z1 −z2dz3 )dz2 −(1+2d+2d2+d3 )(z2d+1 −z1 z3d )dz3

to projective space, and defining F ϕ . By direct computation, we note that the form
ω1 with coefficients given by the equation (6) is not integrable.
Hence, any distribution with the same singular scheme Z as the foliation F ϕ (see
[3] for its explicit geometric description) can be defined by a form α0 ω0 + α1 ω1 ,
where ω0 = ω. The integrability condition is

α0 α1 ω0 ∧ dω1 + α0 α1 ω1 ∧ dω0 + α21 ω1 ∧ dω1 .

An explicit computation shows that the only solution is α1 = 0, and so the only
foliation with singular scheme Z is F ϕ .

In principle, we do not know whether there are other irreducible components

of the space of foliations with the same split tangent sheaf. In fact, despite the
recent important results in the literature, the knowledge of these spaces is still
quite uncomplete.

Remark 4.6. Observe that the proofs of the results in Sections 2 and 3 except
those of Proposition 4.4 and Theorem 4.5, could be immediately adapted to deal
with singular distributions, since no use is made of the integrability condition.
Note also that we can extend our approach to foliations in Pn , and get immediate
results: Theorem 3.5, Lemma 4.2 and Theorem 4.3 are valid in that setting.

References
[1] P. Baum, R. Bott, Singularities of Holomorphic Foliations, Jour. of Diff. Geom. 7, (1972),
279–342.
[2] O. Calvo-Andrade, Irreducible components of the space of holomorphic foliations, Math. Ann.
vol 29, no. 1 (1994), 751–767.
[3] O. Calvo-Andrade, D. Cerveau, L. Giraldo, A. Lins-Neto, Irreducible components of the space
of foliations associated to the affine Lie algebra, Ergodic Theory Dynam. Systems, vol. 24
(2004), no. 4, 987–1014.
[4] A. Campillo, J. Olivares, Polarity with respecto to a foliation and Cayley-Bacharach Theo-
rems, J. reine angew. Math. 534 (2001), 95-118.
[5] D. Cerveau, A. Lins-Neto, Irreducible components of the space of holomorphic foliations of
degree two in CP(n), n ≥ 3, Ann. of Math. 143 (1996), 577–612.
[6] F. Cukierman, J.V. Pereira, Stability of holomorphic foliations with split tangent sheaf, Amer.
J. Math. vol 130, no. 2 (2008), 413–439.
[7] F. Cukierman, M. G. Soares, I. Vainsencher, Singularities of logarithmic foliations, Compo-
sitio Math. 142 (2006), 131–142.
[8] D. Eisenbud, Commutative Algebra with a view toward Algebraic Geometry, Graduate Texts
in Mathematics, vol. 150 (1995), Springer Verlag.
[9] X. Gómez-Mont, Universal families of foliations by curves, Astérisque 150–151, Société
Mathématique de France, 1987, 109–129.
 16 LUIS GIRALDO AND ANTONIO J. PAN-COLLANTES

[10] G. -M. Greuel, G. Pfister. H. Schönemann, Singular 3.0. A Computer Algebra System for
Polynomial Computations. Center for Computer Algebra, University of Kaiserslautern (2005).
http://www.singular.uni-kl.de.
[11] R. Hartshorne, Algebraic Geometry, Springer, Graduate Text in Mathematics: 52 (1977).
[12] R. Hartshorne, Stable Reflexive Sheaves, Math. Annalen, vol. 254 (1980), 121–176.
[13] F. Hirzebruch, Topological Methods in Algebraic Geometry, Classics in Mathematics.
Springer-Verlag, Berlin (1995)
[14] Y. Mitera, J. Yoshizaki, The local analytic triviality of a complex analytic singular foliation,
Hokkaido Math. J., vol. 33 (2004), no. 2, 275–297.
[15] Y. Miyaoka, T. Peternell, Geometry of Higher Dimensional Algebraic Varieties, DMV-
Seminar, vol. 26, Birkhäuser Verlag (1997).
[16] C. Okonek, M. Schneider, H. Spindler, Vector bundles on complex projective spaces, Progress
in Mathematics, 3. Birkhaüser (1980).
[17] M. Roggero, On the rank 2 reflexive sheaves and the subcanonical curves in P3 , Comm
Algebra, 16(9) (1988), 1779–1790.
[18] T. Suwa, Unfoldings of complex analytic foliations with singularities, Japanese J. Math. vol
9 (1983), 181–206.
[19] T. Suwa, Indices of Vector Fields and Residues of Singular Holomorphic Foliations, Actu-
alités Mathématiques, Hermann Éditeurs, Paris, 1998.

Departamento de Geometrı́a y Topologı́a, Facultad de Ciencias Matemáticas, Uni-

versidad Complutense de Madrid. Plaza de Ciencias 3, Ciudad Universitaria. 28040
Madrid, SPAIN
E-mail address: luis.giraldo@mat.ucm.es

Departamento de Matemáticas, Facultad de Ciencias, Universidad de Cádiz. Polı́gono

Rı́o San Pedro. 11510 Puerto Real (Cádiz), SPAIN.
E-mail address: antonio.pan@uca.es

View publication stats