Degree of logarithmic foliations of type (1,1,1)

Mariano Chehebar1
Abstract
arXiv:2111.09134v1 [math.AG] 17 Nov 2021

The space of codimension one holomorphic foliations of degree 1 in a projective space has an irre-
ducible component whose general element is a logarithmic differential 1-form with simple poles in three
hyperplanes. We compute its projective degree by resolving its rational parametrization map through
succesive blow-ups with smooth centers.

1 Introduction
n
A codimension one holomorphic foliation Pn of degree d in the complex projective space P is defined by a
non-zero twisted differential 1-form ω = i=0 ai dxi such that ai are homogeneous
Pn polynomials of degree d+1
without non-constant common divisors, satisfying the descent condition i=0 ai xi = 0 and the Frobenius
integrability condition ω ∧ dω = 0. Multiplication by a non-zero constant does not change the invariant
hypersurfaces of ω. Thus, the moduli space F(n, d) is defined as the Zariski closure of the set of codimension
one holomorphic foliations of degree d in Pn inside the space PH 0 (Pn , Ω1Pn (d + 2)) of projectivized global
sections of twisted differential 1-forms. The geometry of the space of foliations in Pn for n ≥ 3 is very rich
and inspired a lot of research.
Many authors studied families and irreducible components of the space F(n, d). For general degree d,
there are a few known irreducible components such as pullback components (see [4]), rational components
(see [12] and [8]) and logarithmic components (studied in [2] and more recently in [7]). A codimension one
foliation belonging to a pullback component is defined by a 1-form which is the pullback by a rational map
of a 1-form in P2 . To P define logarithmic components, fix an integer partition d = (d1 , . . . , dm ) of d + 2 such
that m ≥ 2, d + 2Q= i di P and d1 ≥ · · · ≥ dm > 0. A logarithmic 1-form ω of type d in Pn is defined by
m m
the formula ω = ( i=1 Fi ) i=1 λi dF Fi where Fi is a homogeneous
i
polynomial of degree di in the variables
Pm
x0 , . . . , xn and λi ∈ C satisfy the descent condition i=1 λi di = 0. The Zariski closure of the space of
logarithmic 1-forms of type d is called logarithmic component of type d. We denote it by L (d). When the
partition is formed by two elements, i.e. m = 2, we call it rational component of type (d1 , d2 ). In [2], O.
Calvo Andrade proved for the first time that L (d) forms an irreducible component of the corresponding
moduli space of foliations. Recently in [7], the authors gave an algebraic proof of this fact.
A related problem is to find the decomposition of F(n, d) into irreducible components for small d. In
his book [13], Jouanolou exhausted these components for d = 0 and d = 1. In the case d = 0, the space
F(n, 0) is irreducible. For d = 1, the space F(n, 1) (whose elements are also called Jacobi equations) has two
irreducible components: rational of type (2, 1) and logarithmic of type (1, 1, 1). In their celebrated paper [3],
Cerveau and Lins-Neto showed that there are six irreducible components of F(n, 2), two of rational type, two
of logarithmic type, one linear pullback component and another exceptional component, see [3]. In the recent
article [6], the authors studied the case d = 3, showing that there are at least 24 irreducible components,
but not exhausting them all.
An important geometric invariant of F(n, d) is the degree of its irreducible components. There have been
several articles computing the degree of known irreducible components of the moduli space of foliations. In
[10], the authors gave formulas for the degree of some pullback components. The degree of the exceptional
component has been computed in [15]. In [8] and [14], the authors found the degree of several rational
components. The goal of this work is to compute the degree of the logarithmic component of type (1, 1, 1).
For this, we resolve its rational parametrization map ρ via three succesive blow-ups with smooth centers, see
Diagram 1 at the end of Section 3:
Keywords— Foliation, Algebraic Geometry, Intersection Theory.

1
Theorem. The map ρ′′′ : X ′′′ → PH 0 (Pn , Ω1Pn (3)) depicted in Diagram 1 is a resolution of the rational
parametrization map ρ : X 99K PH 0 (Pn , Ω1Pn (3)) of the logarithmic component L (1, 1, 1) of type (1, 1, 1).
We combine this Theorem with intersection theory techniques to get our main result:
Theorem. The degree of the logarithmic component L (1, 1, 1) of type (1, 1, 1) in P3 is 80.
This completes the degree computations for the irreducible components inside the moduli space F(n, 1)
of Jacobi equations in Pn :
Corollary. The moduli space F(3, 1) of Jacobi equations in P3 has two irreducible components L (2, 1) and
L (1, 1, 1) of degrees 55 and 80 respectively.
In Section 2, we introduce logarithmic components and recall some useful results about them. Moreover,
we describe the scheme structure of the base locus of the natural parametrization of the logarithmic com-
ponent L (1, 1, 1) and prove some techincal results for later use. In Section 3, we resolve the base locus of ρ
with a sequence of three succesive blow-ups with smooth centers. For this, we use local coordinates around
a point of the exceptional divisor and compute the set-theoretical base locus in each of the three steps of
the resolution. In Section 4, we use intersection theory techniques and the results of Sections 2 and 3 to
compute the degree of the logarithmic component of type (1, 1, 1).
Acknowledgements. The author would like to thank Fernando Cukierman, Javier Gargiulo Acea, César
Massri and Israel Vainsencher for their useful contributions, comments and suggestions.

2 Preliminaries
In this section, we recall some basic definitions and facts about logarithmic components, see for example
[2] and [7]. Also, we study the scheme structure of the base locus of the natural parametrization of the
logarithmic component of type (1, 1, 1) and prove some technical results that we will use in the following
sections.
Definition 1. Fix integer Pnumbers d ≥ 0, n ≥ 3, m ≥ 2 and an integer partition d = (d1 , . . . , dm ) of
m
d + 2 such that d + 2 = i=1 di and d1 ≥ · · · ≥ dm > 0. A logarithmicQdifferential Pm1-form i of type d in
m
Pn is defined by a twisted differential 1-form that can be written as ω = ( i=1 Fi ) i=1 λi dF where Fi is
Pm Fi
a homogeneous polynomial of degree di in the variables x0 , . . . , xn and λi ∈ C satisfy i=1 λi di = 0. The
logarithmic component of type d in Pn is the Zariski closure of the set of logarithmic differential 1-forms of
type d in Pn . We denote it by L (d).
In other words, there is a multilinear map
m
Y
µd : Λ(d) × Sn (di ) → H 0 (Pn , Ω1Pn (d + 2))
i=1
Pm Qm
defined by µd (λ, F) = F i=1 λi dF
Fi where Sn (di )
i
= H 0 (Pn , O(di )), F = (F1 , . . . , Fm ) ∈ i=1 Sn (di ),
Qm P m
F = i=1 Fi and Λ(d) = {λ = (λ1 , . . . , λm ) ∈ Cm : i=1 λi di = 0}. This multilinear map induces a rational
map
m
Y
ρd : Xd := P(Λ(d)) × PSn (di ) 99K PH 0 (Pn , Ω1Pn (d + 2)).
i=1

The logarithmic component L (d) of type d is the Zariski closure of the image of the map ρd . Notice that
the target space of ρd is a projective space. Thus, ρd is defined by a list of global sections of the line bundle
OXd (1, . . . , 1). To avoid overloading notation, wePwill usually omit projective classes for elements in the
m
image of ρd , writing for example ρd ([(λ, F)]) = F i=1 λi dF Fi .
i

Pm
Remark 2. Notice that every logarithmic differential 1-form ω = F i=1 λi dF n
Fi of type d in P defines a
i

codimension
Pm one foliation of degree d. Indeed, straightforward calculations yield the two equalities iR (ω) =
( i=1 λi di ) F and dω = dF F ∧ ω. Therefore iR (ω) = 0 and ω ∧ dω = 0.

2
 As its name already suggests, the varieties L (d) are irreducible components of the space of codimension
one foliations.
Theorem 3 ([2] and [7]). The variety L (d) ⊆ F(n, d) of logarithmic forms of type d is an irreducible
component of F(n, d). Moreover, the scheme F(n, d) is generically reduced along L (d).
Even though the map ρd is not injective in general, it is generically finite whenever m ≥ 3.
Remark 4. Notice that ρd cannot be generically injective in general. Indeed, fix an integer partition d
as before and define Ae = {i : di = e}. Let S(e) = {σQ∈ Sm : σ(j) = j for all j ∈ / Ae } be a subgroup of the
group of permutations of m elements and let S(d) = e S(e) ⊆ Sm . Notice that S(d) acts in the domain of
µd via σ · (λ, F) = (σ · λ, σ · F) in the natural way. Thus, the map µd is constant in the orbits of this action.
Qm
We say that ρd is injective up to order if the induced map with domain (Λ(d) × i=1 Sn (di )) /S(d) is
injective.
Proposition 5 ([7]). Let d = (d1 , . . . , dm ) be an integer partition with d1 ≥ · · · ≥ dm > 0 and m ≥ 3. The
rational map ρd is generically injective up to order.
For computing the degree of a logarithmic component, it is essential to understand its base locus. The
set theoretical base locus of the natural parametrization ρd has been described in [7, Section 5]. Let us recall
′
briefly that description. Let d be as above and take a decomposition ϕ = (e, d′ ) ∈ Zm×m × (Zm
′
~
≥0 \ 0) such
≥0
Pm′
that di = j=1 eij d′j for every 1 ≤ i ≤ m. For each one of those decompositions, define the Segre-Veronese
map
′
m
Y m
Y
νϕ : Sn (d′j ) 99K Sn (dj )
j=1 j=1

Qm′ e Qm′ e
given by νϕ (G1 , . . . , Gm′ ) = ( j=1 Gj 1j , . . . , j=1 Gj mj ). Also, define sets Bϕ := Λ(ϕ) × im(νϕ ) where
P m
Λ(ϕ) = {λ = (λ1 , . . . , λm ) ∈ Cm : i=1 λi eij = 0 for all j}. Notice that the sets Bϕ are contained in the
base locus of µd and therefore their projectivizations are included in the base locus of ρd . Also, for every pair
of decompositions ϕ(1) = (e(1) , d′(1) ), ϕ(2) = (e(2) , d′(2) ), define a relation in the following way: ϕ(1) ≤ ϕ(2) if
rank(e(1) ) = rank(e(2) ) and there exists a matrix e(3) such that e(1) = e(2) · e(3) . This realtion is not a partial
order. Indeed, if two matrices e(1) , e(2) differ by a permutation of columns, the corresponding decompositions
ϕ(1) , ϕ(2) satisfy ϕ(1) ≤ ϕ(2) and ϕ(1) ≥ ϕ(2) at the same time. Therefore, we will consider decompositions
up to permutation of columns so that ≤ is an order relation. The irreducible components of this base locus
are in one-to-one correspondence with vector partitions that are maximal with respect to that order.
Theorem 6 ([7]). The irreducible components of the base locus of ρd are the projectivization of the sets Bϕ
where ϕ = (e, d′ ) is a maximal decomposition with respect to the order ≤ and rk(e) < m.
As we are going to focus in the case d = 1, m = 3 and d = (1, 1, 1), let us set the notation ρ := ρ(1,1,1) ,
Λ := Λ(1, 1, 1) and X := X(1,1,1) ≃ P1 × (Pn )3 . Also, let Bρd be the base locus scheme of ρd , let Bρ be the
base locus scheme of ρ and set µ(1) = (0, 1, −1), µ(2) = (1, 0, −1), µ(3) = (1, −1, 0) ∈ Λ. A direct application
of the former result to our special case yields the decomposition of the set-theoretical base locus of ρ into
the following four irreducible components.
Corollary 7. The irreducible components of the base locus of ρ are given set theoretically by the following
smooth varieties:

(B0 )red = {[(λ, F)] ∈ X : F1 = F2 = F3 }

n o
B1 = [(µ(1) , F)] ∈ X : F2 = F3
n o
B2 = [(µ(2) , F)] ∈ X : F1 = F3
n o
B3 = [(µ(3) , F)] ∈ X : F1 = F2 .

3
Remark 8. The components B1 , B2 and B3 are smooth of dimension 2n and (B0 )red is also smooth and
its dimension is n + 1. Additionally, the sets B1 , B2 and B3 are pairwise disjoint and the intersections
(B0 )red ∩ Bi = [(µ(i) , F)] ∈ X : F1 = F2 = F3 are smooth of dimension n for i = 1, 2, 3.
We will show that the base locus scheme Bρ of ρ is reduced along B1 , B2 , B3 and non-reduced along
(B0 )red . For this, let us describe the Zariski tangent space of the base locus scheme Bρ along its irreducible
components. Let W be a vector space and for every w ∈ W let [w] be its class in PW . There is a natural
identification of the tangent space T[w] PW of PW at [w] with W/[w]. From now on, let (λ′ , F′ ) be an element
of the vector space Λ × (Sn (1))3 and let [(λ′ , F′ )]x be its class inside Tx X for some x ∈ X. We will omit the
subindex x whenever it is clear from the context. In order to compute the tangent space of Bρ , we prove the
following techincal result.
Lemma 9.Q Let d = (d1 , . . . , dm ) be a partition of an integer, let x = (λ, F), v = (λ′ , F′ ) be elements
m
of Λ(d) × i=1 Sn (di ) such that the class of x belongs to the base locus Bρd of ρd and let ε be a formal
parameter. The following formula holds:

ρ([x + εv]) = εH1 (x, v) + ε2 H2 (x, v) mod(ε3 ), (1)

where the terms in the formula are defined by

m m
!!
X dFi X F′
H1 (x, v) = F λ′i +d λi i
i=1
Fi i=1
Fi
! !
m
X Xm λi ′ 2
F′ i λ′i Fi′ Fi − 2 Fi
H2 (x, v) = H1 (x, v) + F d .
i=1
Fi i=1
Fi 2
  2 
1 1 1 F′ Fi′
Proof. Notice that Fi +εFi′ = F′
≡ Fi 1 − ε Fii + ε2 Fi mod(ε3 ). Therefore, we replace in the
Fi (1+ε( Fi ))
i
formula of ρ to get
m
! m  2 !!
Y X d(Fi + εFi′ ) F′ Fi′
ρ([x + εv]) ≡ (Fi + εFi′ ) (λi + ελ′i ) 1 − ε i + ε2 mod(ε3 ).
i=1 i=1
Fi Fi Fi
Pm
From a direct expansion of the product on the left and using the hypothesis i=1 λi dF
Fi = 0, we get
i

m
X m
X  ! m
X
! !
dFi dFi′ F ′ dFi F′ 2
ρ([x + εv]) ≡ F λ′i + λi − i 2 ε+ i
ε +
i=1
Fi i=1
Fi Fi i=1
Fi
m   !!
X 2
dFi′ F ′ dFi Fi′ dFi′ dFi Fi′
+F λ′i − i 2 − λi − ε2 mod(ε3 ).
i=1
Fi Fi Fi2 Fi3
   
dFi′ Fi′ dFi Fi′ Fi′ dFi′ dFi Fi′ 2 1 Fi′ 2
We replace the equalities Fi − Fi2
=d Fi and Fi2
− Fi3
= 2d Fi2
in the last formula. The
result follows.

Remark 10. Notice that the expression H1 (x, v) in (1) does not depend on the representative of a class in
Tx X, as the point x is in the base locus of ρ. Therefore, the expression H1 coincides with the derivative of
the parametrization dµd in the closed points of the base locus, see [7, Section 6].
We use the previous Lemma to get information about the scheme structure of the base locus of ρ.
Proposition 11. The base locus scheme Bρ of ρ is reduced along B1 , B2 and B3 and non-reduced along
(B0 )red .

Proof. Let us compute the Zariski tangent space of Bρ at a general closed point of each irreducible component.
Q3
Take a closed point [(λ, F)] ∈ Bρ and a tangent vector [(λ′ , F′ )] ∈ Tx X ≃ (Λ/hλ)i × i=1 (Sn (1)/hFi i). The

4
vector [(λ′ , F′ )] is tangent to Bρ at [(λ, F)] if and only if ρ((λ, F) + ε(λ′ , F′ )) = 0 mod(ε2 ). Using Lemma 9
and dividing by F , we get the condition
3 3
!
H1 ((λ, F), (λ′ , F′ )) X ′ dFi X Fi′
= λi +d λi = 0. (2)
F i=1
Fi i=1
Fi
P 
3 F′
If we assume F1 = F2 = F3 holds, equality (2) becomes d i=1 λi Fi1 = 0. Thus, the tangent space of
Bρ along a point [(λ, F)] ∈ (B0 )red is defined by
3
X
λi Fi′ = 0 mod(F1 ) (3)
i=1

and therefore has codimension n inside T[(λ,F)] X. We deduce that Bρ is non-reduced along (B0 )red . 
F1′ −F2′
For the case F1 = F2 6= F3 and λ = (1, −1, 0), equality (2) becomes (λ′1 +λ′2 ) dF1 ′ dF3
F1 +λ3 F3 +d F1 = 0.
As λ′1 + λ′2 = −λ′3 , the tangent space of Bρ along a point (λ, F) ∈ (B0 )red is defined by the two conditions

F1′ = F2′ mod(F1 ) λ′3 = 0.

Therefore Bρ is reduced along B3 as its tangent space has codimension n + 1 inside X. The argument for
reducedness along B1 and B2 is analogous.
The reader should note that we computed the Zariski tangent space of Bρ along the non-reduced com-
ponent supported in (B0 )red in the proof of the last proposition, see equation (3).
Definition 12. We define B0 to be the scheme-theoretical (non-reduced) irreducible component of Bρ
supported in (B0 )red .
To finish the section, we point out a final remark regarding the tangent space of the scheme B0 .
Remark 13. From the proof of Proposition 11, we obtain that T B0 |(B0 )red is a vector bundle of rank 2n + 1
and Tx B0 = Tx Bi + Tx (B0 )red for every x ∈ Bi ∩ (B0 )red , i = 1, 2, 3. Also, using the obvious isomorphism
(B0 )red ≃ PΛ × PSn (1) and the description of the Zariski tangent space of Bρ , we deduce the following exact
sequence of vector bundles on (B0 )red :

0 → T B0 |(B0 )red → T X|(B0 )red → T PSn (1) ⊗ OPΛ (1) → 0.

Here the surjective

P map from the right is given on each fiber over (λ, (F0 , F0 , F0 )) ∈ (B0 )red by the assignment
[(λ′ , F′ )] 7→ [ i λi Fi′ ]. We observe that [(λ′ , 0)] represents a tangent vector at [(λ, (F0 , F0 , F0 ))] to (B0 )red
and Bρ as well.

3 Resolution of the rational parametrization map

In this section, we resolve the indeterminacy locus of ρ through a sequence of three succesive blow-
ups with smooth centers. In order to do this, we use standard local coordinates of a blow-up in an open
neighbourhood of a point of the exceptional divisor. We compute the support of the base locus in each of the
three steps of the resolution. There is a commutative diagram summarizing the three steps of the resolution
of ρ at the end of the section, see Diagram 1. From now on, the reader is advised to use it as an outline for
organizing the whole argument.
Let us start with the first step of the resolution of ρ.
Definition 14. Denote the blow-up of X along (B0 )red by X ′ , its natural projection map by π : X ′ → X,
its exceptional divisor by E ′ ≃ PN(B0 )red X, the strict transforms of Bi for i = 1, 2, 3 by Bi′ and the extension
of the rational map ρ by ρ′ : X ′ 99K PH 0 (Pn , Ω1Pn (3)). The map ρ′ is well defined on an open subset of
E ′ and coincides with the rational map ρ ◦ π on X ′ − E ′ . Let Bρ′ be its indeterminacy scheme and let B0′
be the closed subscheme of Bρ′ formed by all the scheme-theoretical (not necessarily reduced) irreducible
components of Bρ′ supported in E ′ .

5
 Now, we describe suitable coordinates in an open neighbourhood around a point in the exceptional divisor
of X ′ .
Remark 15. We can construct local coordinates (ε, x, v) for X ′ around a point x′ ∈ E ′ such that:
i. ε is a local single equation defining the Cartier divisor E ′ ⊂ X ′ in an open neighborhood of x′ .
ii. (0, x, v) represents local coordinates for E ′ around x′ such that x moves along the base (B0 )red and v
moves along the fiber Ex′ over x of the projective bundle E ′ ≃ PN(B0 )red X.
iii. the coordinate x represents the projective class of an element also denoted x ∈ Λ × (Sn (1))3 and the
coordinate v represents the projective class of a vector inside T[x] X ≃ Λ × (Sn (1))3 /hxi that lifts to a
vector also denoted v ∈ Λ × (Sn (1))3 . Then for every ε 6= 0 in a small open neighbourhood of 0 we have
the equality π(ε, x, v) = [x + εv]. Notice that the expression [x + εv] ∈ X does not depend on the choice
of the elements x, v ∈ Λ × (Sn (1))3 .
With this in mind, we compute the set-theoretical base locus of ρ′ .
Proposition 16. The scheme B0′ is supported on the subvariety PN(B0 )red B0 ⊂ E ′ . In other words, the
reduced scheme (B0′ )red is equal to PN(B0 )red B0 .
Proof. Let us consider x′ ∈ E ′ and local coordinates (ε, x, v) around x′ as in Remark 15. Using Lemma 9,
for every ε 6= 0 in a small open neighborhood of x′ we can compute

ρ(π(ε, x, v)) = ρ([x + εv]) = εH1 (x, v) + ε2 H2 (x, v) + O(ε3 ).

Since H1 (x, v) does not vanish generically on E ′ , the generic multiplicity of ρ([x + εv]) along E ′ = (ε = 0)
is one. As a consequence, we define ρ′ locally using the coordinates of Remark 15:

ρ′ (ε, x, v) := 1ε ρ([x + εv]). (4)

We deduce that ρ′ (0, x, v) = H1 (x, v) provided the expression is not equal to zero. Therefore (B0′ )red consists
of the points with coordinates (0, x, v) such that H1 (x, v) = 0, i.e. such that v represents an element of
PN(B0 )red ,x B0 as claimed.
We proceed with the second step of the resolution of ρ.
Definition 17. Let X ′′ be the blow-up of X ′ along (B0′ )red , let π ′ : X ′′ → X ′ be its corresponding projection
map and let E ′′ be its exceptional divisor. Also, let ρ′′ : X ′′ 99K PH 0 (Pn , Ω1Pn (3)) be the extension of the
map ρ′ , let Bρ′′ be its indeterminacy scheme and let Bi ′′ be the double strict transforms of Bi for i = 1, 2, 3.
As in the first step of the resolution of ρ, our goal is to calculate the support of the base locus Bρ′′ .
For this, we will prove a couple of technical results. But first, we describe suitable coordinates in an open
neighbourhood around a point in the exceptional divisor of X ′′ .
Remark 18. In the same line of Remark 15, we construct local coordinates (ε′ , x′ , v ′ ) in X ′′ around a point
x′′ ∈ E ′′ . Firstly, ε′ is a local equation defining E ′′ ⊆ X ′′ . Also, (0, x′ , v ′ ) are local coordinates for E ′′ such
that x′ moves along the center of the blow-up (B0′ )red and v ′ moves along the fiber Ex′′′ over x′ of the projective
bundle E ′′ ≃ PN(B0′ )red X ′ . This means that x′ represents a point in X ′ of coordinates (0, x, v) such that
H1 (x, v) = 0 and v ′ represents the projective class of a normal vector inside N(B0′ )red ,x′ X ′ ≃ Tx′ X ′ /Tx′ (B0′ )red .
Because of Proposition 16, there are n linearly independent directions of the form [(0, 0, e v)] inside
N(B0′ )red ,x′ X ′ . Adding a direction of the form [(e ε, 0, 0)], we get n + 1 linearly independent directions in
N(B0′ )red ,x′ X ′ . We also know that rk(N(B0′ )red X ′ ) = n+1, as dim(X ′′ ) = 3n+1 and dim((B0′ )red ) = 2n. There-
fore, the coordinate v ′ moves along an element of the form [(e v)] ∈ X ′ satisfying H1 (x, ve) 6= 0. Moreover,
ε, 0, e
for every ε′ 6= 0 in a small open neighbourhood of 0 we have the equality π ′ (ε′ , x′ , v ′ ) = (0, x, v) + ε′ (e ε, 0, e
v ).
The following Proposition gives an explicit formula for ρ′′ over the points of the exceptional divisor E ′′
of X ′′ .

6
Proposition 19. Using the local coordinates of Remark 18, we have the equality

ρ′′ (ε′ , x′ , v ′ ) = H1 (x, ve) + εeH2 (x, v) + O(ε′ ),

provided the expression is not equal to zero.

Proof. First, following Proposition 16, we compute
1
ρ′ ◦ π ′ (ε′ , x′ , v ′ ) = ρ′ ((0, x, v) + ε′ (e
ε, 0, e
v )) = eρ([x
ε′ ε + ε′ εe(v + ε′ ve)]).

Using Lemma 9, we get

′ 2
ρ′ ◦ π ′ (ε′ , x′ , v ′ ) = 1
e ((ε ) ε
ε′ ε eH1 (x, ve) + (ε′ εe)2 H2 (x, v) + O((ε′ εe)3 )) =
= ε′ (H1 (x, ve) + εeH2 (x, v)) + O((ε′ )2 ).

As the expression H1 (x, ve) + εeH2 (x, v) does not vanish identically on E ′′ , the generic multiplicity of ρ′ ◦ π ′
along E ′′ = (ε′ = 0) is equal to 1. Consequently we define
1 ′
ρ′′ (ε′ , x′ , v ′ ) = (ρ ◦ π ′ (ε′ , x′ , v ′ )) (5)
ε′
and we get our claim.
In particular, ρ′′ is well defined over a closed point in the exceptional divisor of local coordinates
(0, (0, x, v), [(e
ε, 0, e
v )]) if and only if H1 (x, ve) + εeH2 (x, v) does not vanish. Let us show that it does not
vanish in the fiber of a generic point of (B0′ )red . For this, we will use the following Lemma.

Lemma 20. Let x = [(λ, (F0 , F0 , F0 ))] ∈ (B0 )red be a closed point and let v = [(λ′ , F′ )], ve = [(λe′ , f
F′ )] ∈ Tx X
′ f′ 3
with F , F ∈ (Sn (1)) . Assume that H1 (x, v) = H1 (x, ve) + αH2 (x, v) = 0 for some α ∈ C \ 0. Then, either
F1′ ≡ F2′ ≡ F3′ mod(F0 ) or there exists i 6= j such that Fi′ ≡ Fj′ mod(F0 ) and λi + λj = 0.
 P3 ′
 P3
i=1 λi Fi
Proof. From the equation H1 (x, v) = d F0 = 0, we deduce that i=1 λi Fi′ ≡ 0 mod(F0 ). Addition-
ally, the equality H1 (x, ve) + αH2 (x, v) = 0 and the fact that H1 (x, v) = 0 yields
P3 !
f′ ′ ′ λi ′ 2
i=1 F0 (λi Fi + αλi Fi ) − α 2 Fi
d = 0.
F02

Thus the expression inside the differential is constant. By multiplying it by F02 and reducing it modulo F0 ,
P3 2 P3
we get i=1 λi Fi′ ≡ 0 mod(F0 ). Adding the descent condition i=1 λi = 0, we get the following system
   
1 1 1 λ1
 F1′ F2′ F3′  · λ2  ≡ 0 mod(F0 ).
2 2 2
F1′ F2′ F3′ λ3

The result follows directly from a standard determinant and kernel computation of Vandermonde matrices.

We use these last two results to compute the support of Bρ′′ .

F3
Proposition 21. The scheme Bρ′′ is supported in the disjoint union i=1 Bi ′′ .
S3
Proof. Observe that ρ′ is well-defined outside (B0′ )red ∪ i=1 Bi′ and the generic point of the Bi ′′ belongs
to the base locus Bρ′′ . Hence, it is enough to show that the only closed points of Bρ′′ intersecting the
exceptional divisor E ′′ belong to one of the Bi ′′ . Let us take a closed point x′′ ∈ E ′′ such that ρ′′ (x′′ ) = 0.
The local coordinates (0, (0, x, v), [(e v )]) for x′′ of Remark 18, satisfy the equalities
ε, 0, e

ρ(x) = 0 H1 (x, v) = 0 H1 (x, ve) + εeH2 (x, v) = 0. (6)

7
 From these conditions, we deduce that εe 6= 0. Otherwise, we would get H1 (x, ve) = 0 and therefore
(e v ) would represent the zero vector inside N(B0′ )red ,x′ X ′ . This is a contradiction as v ′ is the projective
ε, 0, e
class of (e ε, 0, e
v ). Notice that v cannot be the class of a vector tangent to (B0 )red as its projective class
inside N(B0 )red ,x X should be well defined. Thus, applying Lemma 20, using the same notation for x, v, ve
and assuming without losing generality that i = 1, j = 2 we get that F1′ = F2′ 6= F3′ , λ1 + λ2 = λ3 = 0.
Consequently, we write x = [((1, −1, 0), (F0 , F0 , F0 ))]. Also, we can assume that λ′ = ~0 and F2′ = 0 as
v = [(λ′ , F′ )] represents a vector of N(B0 )red ,x X. We therefore get that H2 (x, v) = 0. Also, from equation
(6) we deduce that H1 (x, ve) = 0, i.e. v ′ is represented by [(e ε, 0, ve)] = [(1, 0, 0)] inside PN(B0′ )red ,x′ X ′ .
Finally, let us prove that x = (0, (0, x, v), [(1, 0, 0)]) is a closed point of B3 ′′ . For this, we construct
′′

locally a curve x′′ (t) such that limt→0 x′′ (t) = x′′ . Indeed, the curve given in local coordinates of Remark 18
by x′′ (t) = (t, (0, x, v), [(1, 0, 0)]) pushes via the two blow-up projections to the curve

π ◦ π ′ (x′′ (t)) = π((0, x, v) + t(1, 0, 0)) = [x + tv] = [((1, −1, 0), (F0 , F0 , F0 + tF3′ ))].

But π ◦ π ′ (x′′ (t)) is contained in B3 for every t 6= 0. The proposition follows.

We arrive at the third and final step of the resolution of ρ.
F3
Definition 22. Let X ′′′ be the blow-up of X ′′ along i=1 Bi ′′ , let π ′′ : X ′′′ → X ′′ be its projection map
and let ρ′′′ : X ′′′ 99K PH 0 (Pn , Ω1Pn (3)) be the extension of ρ′′ . Also, let Bρ′′′ be its base locus scheme, let
F3
Ei′′′ = π ′′−1 (Bi ′′ ) be the fiber of Bi ′′ via π ′′ and let E ′′′ = i=1 Ei′′′ be the exceptional divisor.
Before showing that the map ρ′′′ is regular, let us describe suitable coordinates in an open neighbourhood
around a point in the exceptional divisor of X ′′ .
Remark 23. Using Remarks 15 and 18, we construct local coordinates (ε′′ , x′′ , v ′′ ) in X ′′ around a point
x′′′ ∈ E ′′′ in a
similar fashion. Here ε′′ is a local equation for
E ′′′ ⊆ X ′′′ . As Bρ′′ is smooth along
Bi ′′ \ Bi ′′ ∩ E ′′ , we will focus on the points x′′′ ∈ π ′′−1 E ′′′ ∩ Bi ′′ and assume that i = 3.
The coordinate x′′ represents a point [(0, x′ , v ′ )] ∈ X ′′ . Here x′ is of the form [(0, x, v)] ∈ X ′ such that
3
H1 (x, v) = 0, x = [((1, −1, 0), (F0 , F0 , F0 ))] ∈ (B0 )red , v = (~0, (0, 0, F3′ )) ∈ Λ × (Sn (1)) and v ′ = [(1, 0, 0)].
For the coordinate v ′′ , we find a basis of n+ 1 vectors of NB3′′ ,x′′ X ′′ . We select n of them with coordinates
(0, 0, e ve) ∈ Tx′ X ′ . Because of Remark 13, e
v ′ ) ∈ Tx′′ X ′′ , where ve′ = (0, 0, e ve represents a direction normal to
B3 such that H1 (x, eve) = 0 and thus it lifts to a vector of the form (~0, (Ff f′ , 0, 0)) ∈ Λ × (S (1))3 . We
1 n
′′ ′ ′′ ′
add a direction normal to B3 of the form (0, x e , 0) ∈ Tx′′ X where x e has coordinates (0, x e, 0) ∈ Tx′ X ′
3
and x e = ((0, 1, −1), ~0) ∈ Λ × (Sn (1)) . Therefore, a general vector has the form v ′′ = (~0, x e′ , ve′ ) where
x′
e = (0, αe x, 0) ∈ X for some α ∈ C and ve = (0, 0, e
′ ′ ′
ve) ∈ Tx′ X as before.

To finish the section, let us show that the induced map ρ′′′ : X ′′′ → PH 0 (Pn , Ω1Pn (3)) with base locus
scheme Bρ′′′ is indeed regular.
Theorem 24. The map ρ′′′ : X ′′′ → PH 0 (Pn , Ω1Pn (3)) depicted in Diagram 1 is a resolution of the rational
parametrization map ρ : X 99K PH 0 (Pn , Ω1Pn (3)) of the logarithmic component L (1, 1, 1) of type (1, 1, 1).

Proof. We only need to prove that the induced map ρ′′′ is regular. Without loss of generality, it is enough
to prove that ρ′′′ is well defined in the fiber of a point x′′ ∈ E ′′ ∩ B3 ′′ . Take local coordinates (ε′′ , x′′ , v ′′ )
around x′′′ as in Remark 23. Now, we use Lemma 9 and the definition of ρ′′ (equation (5)) to get a formula
for ρ′′′ . We get

ρ′′ ◦ π ′′ (ε′′ , x′′ , v ′′ ) = ρ′′ (0, x′ + ε′′ xe′ , v ′ + ε′′ ve′ ) = H1 (x + ε′′ αe ve) + H2 (x + ε′′ αe
x, ε′′ e x, v) =
′′
(7)
ε (H1 (x, eve) + H2 (e xα,F0 , v)) + O(ε′′2 )

where xeα,F0 = ((0, α, −α), (F0 , F0 , F0 )). Cancelling out ε′′ in (7) we get the formula for ρ′′′ . In particular,
inside E ′′′ it is given by the formula

ρ′′′ (0, x′′ , v ′′ ) = H1 (x, eve) + H2 (e

xα,F0 , v). (8)

8
Now, suppose there exists a point of coordinates (0, x′′ , v ′′ ) that belongs to the base locus Bρ′′′ of ρ′′′ . Then,
the expression for ρ′′′ in equation (8) vanishes. Dividing by F03 we get
!
f
f′
F1
d = 0.
F0

f
f′ ≡ 0 mod(F0 ) and therefore e
This yields F ve is zero. This is a contradiction because e
ve cannot vanish, as v ′′
1
is the projective class of a normal vector.
The following commutative diagram summarizes this section:
F3
E ′′′ = i=1 Ei′′′ X ′′′
π ′′
F3
E ′′ i=1 Bi ′′ X ′′
ρ′′′
π′
ρ′′
(B0′ )red E′ X′
ρ′
π
ρ
(B0 )red X PH 0 (Pn , Ω1Pn (3))

Diagram 1: The sequence of blow-ups resolving the rational parametrization map ρ.

4 Degree Calculations
In this section, we use the resolution of the rational parametrization map ρ of Section 3 compute the
degree of the logarithmic component of type (1, 1, 1). For this, we use standard intersection theory techniques.
Let h be the hyperplane class of PH 0 (Pn , Ω1Pn (3)). We know that the degree of L (1, 1, 1) is defined
as the degree of the Chow 0-cycle (h|L (1,1,1) )3n+1 inside L (1, 1, 1). Recall that the map ρ is generically
injective up to order, see Proposition 5. Thus, the degree of the map ρ is the same as the order of the group
S(1, 1, 1) ≃ S3 defined in Remark 4. We deduce that the map ρ′′′ is generically 6 to 1 and
Z Z
3n+1
′′′∗
ρ (h) =6 (h|L (1,1,1) )3n+1 .
X ′′′ X ′′′

Therefore, to compute the degree of L (1, 1, 1), we pullback the hyperplane class h through ρ′′′ . Recall that
the rational map ρ was defined by sections of the line bundle OX (1, 1, 1, 1). Using equations (4), (5) and (7)
and the local calculations in the previous section, we know that ρ′′′ is defined by sections of the line bundle

OX (1, 1, 1, 1) ⊗ OX ′ (−E ′ ) ⊗ OX ′′ (−E ′′ ) ⊗ ⊕3i=1 OX ′′′ (−Ei′′′ ).

We conclude that the pullback of the hyperplane class h of the target space of ρ′′′ is
4
X 3
X
ρ′′′∗ (h) = hi − e 1 − e 2 − e3,i .
i=1 i=1

Here h1 , h2 , h3 , h4 are generators of X such that h1 is the (pullback of the) hyperplane class of PΛ ≃ P1
and h2 , h3 , h4 are the (pullback of the) hyperplane class of the three different PSn (1) ≃ Pn . Also, e1 = [E ′ ],
e2 = [E ′′ ] and e3,i = [Ei′′′ ]. Consequently, the degree of L (1, 1, 1) is computed as

Z 4 3
!3n+1
1 X X
hi − e 1 − e 2 − e3,i .
6 X ′′′ i=1 i=1

9
 For computing this number, we use standard intersection theory techniques. The main reference for this
theory is [11]. Using the fact that the three blow-up projections are birational, we will push forward the zero
cycle ρ′′′∗ (h)3n+1 to X in three steps to get a zero cycle inside X with the same degree. The computation
inside X is easier as it is a product of projective spaces.
For pushing forward this cycle, we need to understand the pushforwards of powers of the cycles e1 , e2 , e3,i
through their corresponding blow-up projections. These are exactly the Segre classes of the normal bundles
of the respective centers of blow-up in the ambient space, see [11, Corollary 4.2.2]. Eventhough Segre classes
can be defined generally for every cone inside a scheme, in the case of vector bundles they can be defined
as the graded parts of a formal inverse of the corresponding total Chern class inside the Chow ring, see [11,
Proposition 4.1]. The sum of all the Segre classes of a vector bundle is the total Segre Class and it is additive
in exact sequences.
As a first step, we collect the coefficients of e3,i and use our knowledge of the normal bundle NBi ′′ X ′′ to
push it forward to X ′′ . Indeed, we know that

π∗′′ (ej3,i ) = (−1)j−1 sj ′ (NBi ′′ X ′′ ) ∩ [Bi ′′ ]

where j ′ = j − codim(Bi ′′ ) = j − n − 1.
For computing the class [Bi ′′ ] we use twice the Blow-Up Formula, see [11, Theorem 6.7]:

[Bi ′′ ] = [Bi ] − j1 ∗ c(E1 ) ∩ π ∗ (s(NBi ∩(B0 )red Bi ) n+1

−j2 ∗ c(E2 ) ∩ π ′∗ (s(NBi′ ∩(B0′ )red Bi′ )) n+1 .

Here j1 and j2 are the inclusions of E ′ and E ′′ in X ′ and X ′′ and the bundles E1 ≃ π ∗ (N(B0 )red X)/NE ′ X ′
and E2 ≃ π ′∗ (N(B0′ )red X ′ )/NE ′′ X ′′ are the excess bundles. The Segre class s(NBi ∩(B0 )red Bi ) is deduced from
the exact sequence

0 → T (Bi ∩ (B0 )red ) → T Bi |Bi ∩(B0 )red → NBi ∩(B0 )red Bi → 0.

On the other hand, we deduce the equality Bi′ ∩ (B0′ )red = Bi′ ∩ E ′ using the description of the tangent space
TB0 in equation (3). Therefore, we get NBi′ ∩(B0′ )red Bi′ ≃ OE ′ ′ (−1) where EB′
′ is the exceptional divisor of
B i
i
Bi′ . For the computation of c(E1 ) we use the isomorphism NE ′ X ′ ≃ OE ′ (−1) and the exact sequence

0 → T (B0 )red → T X|(B0 )red → N(B0 )red X → 0. (9)

To calculate c(E2 ) we use the two isomorphisms N(B0′ )red E ′ ≃ OE ′ (1) ⊗ π ′∗ (N(B0 )red X/N(B0 )red B0 ) and
NE ′′ X ′′ ≃ OE ′′ (−1), the exact sequence

0 → N(B0′ )red E ′ → N(B0′ )red X ′ → NE ′ X ′ |(B0′ )red → 0 (10)

and the formula for the Chern class of a tensor product with a line bundle , see [9, Proposition 9.13] and
[9, Proposition 5.17]. The Chern classes of the bundle N(B0 )red X/N(B0 )red B0 can be deduced from the exact
sequences of Remark 13, the exact sequences

0 → T (B0 )red → T X|(B0)red → N(B0 )red X → 0

0 → T (B0 )red → T B0 |(B0 )red → N(B0 )red B0 → 0

and our knowledge of the tangent bundles of (B0 )red and X.

For the calculation of the Segre classes of the normal bundles NBi ′′ X ′′ , consider the auxiliary algebraic
subsets of X of dimension 2n + 1 containing both (B0 )red and the corresponding Bi


Z1 = {[(λ, (A, B, B))] : A, B ∈ Sn (1)} ⊆ X
Z2 = {[(λ, (B, A, B))] : A, B ∈ Sn (1)} ⊆ X


Z3 = {[(λ, (B, B, A))] : A, B ∈ Sn (1)} ⊆ X.

10
The key fact is that (B0 )red and Bi intersect properly inside Zi . This follows from the dimensions and set
theoretical descriptions of the sets involved. Thus, we get s(NBi ′′ X ′′ ) = s(NBi′ Zi′ )s(π ′∗ (NZi′ X ′ ) ⊗ OE ′′ (1)),
where Zi′ is the strict transform of Zi , see [1, Section 4.3]. Notice that Bi and (B0 )red intersect properly
inside Zi and therefore s(NBi′ Zi′ ) = s(π ∗ NBi Zi ). From the exact sequence

0 → T Bi → T Zi |Bi → NBi Zi → 0,
1
we get that s(π ∗ NBi Zi ) = 1+h 1
|Bi = 1. For computing s(π ′∗ (NZi′ X ′ ) ⊗ OE ′′ (1)) we use the isomorphism
′
NZi′ X ≃ NZi X ⊗ OE ′ (1) (see [11, Appendix B.6.10]), the formula for the Segre class of a tensor product
with a line bundle ([11, Example 3.1.1]) and the exact sequence

0 → T Zi → T X|Zi → NZi X → 0.

As a second step, we collect the terms of coefficient ej2 . We know that

π∗′ (ej2 ) = (−1)j−1 sj ′ (N(B0′ )red X ′ ) ∩ [(B0′ )red ]

where j ′ = j − codim((B0′ )red ) = j − n − 1. The computation of the class [(B0′ )red ] is deduced from the
formula for the class of a projective bundle in terms of the hyperplane class of the exceptional divisor and
the already mentioned Chern classes of the bundle π ′∗ (N(B0 )red X/N(B0 )red B0 ), see [9, Proposition 9.13]. For
the Segre classes sj ′ (N(B0′ )red X ′ ), we use the isomorphisms N(B0′ )red E ′ ≃ OE ′ (1) ⊗ π ′∗ (N(B0 )red X/N(B0 )red B0 )
and NE ′ X ′ ≃ OE ′ (−1), the exact sequence (10) and the formula for the Segre class of a tensor product with
a line bundle, see [11, Example 3.1.1].
Finally, we collect the terms of coefficient ej1 . We know that

π∗ (ej1 ) = (−1)j−1 sj ′ (N(B0 )red X) ∩ [(B0 )red ],

where j ′ = j − codim((B0′ )red ) = j P

− 2n. As (B0 )red is the image of a diagonal embedding, its class inside
the Chow ring of X is [(B0 )red ] = 0≤i,j≤n hi2 hj3 h2n−i−j
4 . For the Segre classes sj ′ (N(B0 )red X) we use the
exact sequence (9).
To implement the actual calculations, it is best to use a computer program. A script using Macaulay2
can be found in [5]. The following table shows the degree of the logarithmic components L (1, 1, 1) of type
(1, 1, 1) in Pn for small n.

n Degree
3 80
4 4035
5 165984
6 6091960
7 208063680
8 6766823415

Table 1: Degree of the logarithmic components L (1, 1, 1) of type (1, 1, 1) in Pn for small n.

In particular, we have the following

Theorem 25. The degree of the logarithmic component L (1, 1, 1) of type (1, 1, 1) in P3 is 80.
The degree of the component L (2, 1) inside F(3, 1) is 55. It was computed in [8, Section 5.3.1]. Therefore,
this completes the calculation of the degree of the two irreducible components of the space of Jacobi equations
in P3 .
Corollary 26. The moduli space F(3, 1) of Jacobi equations in P3 has two irreducible components L (2, 1)
and L (1, 1, 1) of degrees 55 and 80 respectively.

11
References
[1] P. Aluffi. Chern classes of blow-ups. Mathematical Proceedings of the Cambridge Philosophical Society
148 (2) (2010), 227-242.

[2] O. Calvo-Andrade. Irreducible components of the space of foliations. Mathematische Annalen 299 (1994).
[3] D. Cerveau and A. Lins-Neto. Irreducible components of the space of holomorphic foliations of degree
two in CP(n). Annals of Mathematics 143 (1996).
[4] D. Cerveau, S. J. Edixhoven and A. Lins Neto. Pull-back components of the space of holomorphic
foliations in CP(n), n ≥ 3. J. Algebraic Geom. 10(4) (2001), 695–711.
[5] M. Chehebar. https://github.com/marianocheh/Degree111.
[6] R. Constant da Costa, R. Lizarbe and J. V. Pereira. Codimension one foliations of degree three on
projective spaces. arXiv: 2102.10608 (2021).
[7] F. Cukierman, J. N. Gargiullo Acea and C. Massri. Stability of logarithmic differential one-forms.
Transactions of the American Mathematical Society 371.9 (2019), 6289-6308.
[8] F. Cukierman, J. V. Pereira and I. Vainsencher. Stability of foliations induced by rational maps. Annales
de la Faculté des sciences de Toulouse: Mathématiques - 18(4) (2009), 685-715.
[9] D. Eisenbud and J. Harris. 3264 and all that: a second course in algebraic geometry. Cambridge Uni-
versity Press (2016).
[10] V. Ferrer and I. Vainsencher. Linear pullback components of the space of codimension one foliations.
Bulletin of the Brazilian Mathematical Society (2020).
[11] W. Fulton. Intersection Theory. Springer-Verlag, New York (1985).
[12] X. Gomez-Mont and A. Lins-Neto. Structural stability of singular holomorphic foliations having a mero-
morphic first integral. Topology, Vol. 30 Issue 3 (1991), 315-334.
[13] J. P. Jouanoulou. Équations de Pfaff Algébriques. Springer Lecture Notes in Mathematics 163 (1970).
[14] D. Leite and I. Vainsencher. Degrees of spaces of holomorphic foliations of codimension one in Pn . J.
Pure Appl. Algebra (2017).
[15] A. Rossini and I. Vainsencher. Degree of the exceptional component of foliations in P3 . RAC-SAM
(2019).

1
Departamento de Matemática - IMAS, FCEyN, Universidad de Buenos Aires, Buenos Aires, Argentina.
The author was fully supported by CONICET.
E-mail address: mchehebar@dm.uba.ar

12