DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS

PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ,

AND HERNÁN NECIOSUP
arXiv:1703.10749v2 [math.DS] 6 Apr 2017

Dedicated to Felipe Cano, on the occasion of his first 60 years of mathematical life

Abstract. We study in this paper several properties concerning singularities

of foliations in (C3 , 0) that are pull-back of dicritical foliations in (C2 , 0).
Particularly, we will investigate the existence of first integrals (holomorphic
and meromorphic) and the dicriticalness of such a foliation. In the study of
meromorphic first integrals we follow the same method used by R. Meziani
and P. Sad in dimension two. While the foliations we study are pull-back of
foliations in (C2 , 0), the adaptations are not straightforward.

1. Introduction
In the paper [FM], first and second authors studied the analytic classification
of a generic class of quasi-ordinary singularities of codimension one, holomorphic
foliations in dimension three. More precisely, we considered foliations F that can
be generated by a holomorphic 1-form
0 0 0 0
Ω = d(z 2 + xp y q ) + (xp y q )k αU (xp y q )dz,
where p, q ∈ N\{0}, d = gcd(p, q), p = dp0 , q = dq 0 , U (t) ∈ C{t}, U (0) = 1, α ∈ C∗ .
In that paper, we focused in foliations of generalized surface type, condition that
implies certain relations between k and d. We supposed that one of the two following
conditions is satisfied:
(i) Either 2k > d,
(ii) Or 2k = d and a certain arithmetical condition on α is satisfied.
The classifying object turns out to be the projective holonomy of a certain com-
ponent of the exceptional divisor that appears in a particular reduction of the
singularities, that follows the idea of Jung’s method [G, Co]. This reproduces the
scheme established in dimension two by D. Cerveau and R. Moussu [CeM], and R.
Meziani [Me].
We are now interested in the dicritical case, and the problem of knowing if such
a foliation has a meromorphic first integral. As we will see, in this situation we
must have that d = 2k. Under this last condition, several possibilities may appear.
For generic values of α the analytic classification of such foliations is essentially the

Date: February 12, 2018.

Key words and phrases. Holomorphic foliations, dicritical foliations.
This work was funded by the Dirección de Gestión de la Investigación at the PUCP through
grant DGI-2015-1-0045, and by the Ministerio de Economı́a y Competitividad from Spain, under
Projects “Álgebra y Geometrı́a en Dinámica Real y Compleja III” (Ref.: MTM2013-46337-C2-1-P)
and “Álgebra y geometrı́a en sistemas dinámicos y foliaciones singulares” (Ref: MTM2016-77642-
C2-1-P).
1
2 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

same that in the case d < 2k, in the sense that projective holonomy classifies. We
will suppose that we are in a non-generic case, and that the foliation is dicritical.
The objective of the paper is to study the existence, in this case, of first integrals,
both of holomorphic or meromorphic type, and the dicriticalness of such a foliation.
In particular, in order to investigate the existence of meromorphic first integrals
we will follow the ideas developed in dimension two by R. Meziani and P. Sad in
[MeSa]. Let us observe that the case we are studying is a pull-back of a two-
dimensional foliation. More precisely, consider the map

ρ : C3 −→ C2
0 0
(x, y, z) 7−→ (xp y q , z).

Then,
Ω = ρ∗ (d(z 2 + t2k ) + tk αU (t)dz).
It is tempting to try to use this map in order to transfer the properties of Fk,α,U
(foliation in (C2 , 0) generated by ωk,α,U = d(z 2 + t2k ) + tk αU (t)dz) to properties
of F. In fact, if there is a meromorphic first integral F (t, z) for Fk,α,U it is clear
that Ω = ρ∗ Fk,α,U has ρ∗ F = F ◦ ρ as meromorphic first integral. But in general
it is not true that the existence of a first integral for a pull-back implies that the
original foliation has also a first integral. Consider, e.g., any foliation generated by
a 1-form ω in (C2 , 0), and the restriction of its reduction of singularities to a small
neighbourhood of a regular point. This restriction has always a first integral, while
the original foliation may not. For the same reason, it is not straightforward that,
if a foliation is dicritical, a pull-back of the foliation is also dicritical. The same
example applies. Nevertheless, in the case under consideration, the condition for
the dicriticalness and for the existence of a meromorphic first integral is essentially
the same as in dimension two. We will explore carefully this approach in this paper.
Let us briefly recall the criteria for the existence of a meromorphic first integral
for nilpotent foliations in (C2 , 0). First of all, we have to remind that there is not a
topological characterization, as Suzuki example shows (see [CeMa]). In our case,
consider a foliation generated by

ω = d(z 2 + t2p ) + αtp U (t)dz,

such that, after p blow-ups, we arrive at a situation as described in [MeSa]: there

are two singularities, not corners, in the last component Dp of the divisor, call them
mi , m−i , such that mi is imposed to be linearizable and dicritical, and m−i simple
resonant. Take a transversal Σ on Dp , where the holonomy of Dp is computed. Ac-
cording to [MeSa], the existence of a meromorphic first integral for ω is equivalent
to the existence of a rational function r(T ) on the transversal Σ invariant by the
holonomy group of Dp . Besides this, Camacho [Ca] characterizes in full generality
the existence of a meromorphic first integral in terms of the singular holonomy
of the foliation. Let us briefly recall this notion and make the link between both
approaches, for the sake of completeness. The result shown by Camacho is the
following one:
Theorem [Ca]. The foliation F admits a meromorphic first integral if and only
if:
(i) The singular holonomy of the nondicritical branches is finite.
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 3

(ii) For each dicritical component Di of the resolution of F, there is a meromor-

phic function fi : Di → C such that it is invariant by the singular holonomy
of the nondicritical branches intersecting Di .
(iii) Given any two dicritical components Di and Dj , and a branch Bij between
them, i.e., qi = Bij ∩ Di 6= ∅ 6= Bij ∩ Dj = qj , the germs of fi at Di , fj at
Dj are compatible, in the sense that they take the same values ove the same
leaves L near Bij .
Consider a corner of the divisor obtained after reduction of singularities, where
the foliation is resonant linearizable, locally defined by ωpq = pydx + qxdy, (p, q) =
1, and where Di is the component x = 0, and Dj is y = 0. The holonomy hi of
Di at this singularity is hi (t) = e−2πip/q t, and analogously, hj (t) = e−2πiq/p t turns
out to be the holonomy map of Dj . Dulac’s map is, in these coordinates, a map
D : Σi → Σj such that if (t, 1) ∈ Σi , (1, D(t)) ∈ Σj (Σi , Σj are transversals through
t = 1 to Di , Dj , respectively), and (t, 1), (1, D(t)) are on the same leaf. So, we have

tp = D(t)q and then D(t) = tp/q .

Define the adjunction map by D◦hi = hD D
i ◦D, hi : Di → Dj . A simple computation
D q
shows that hi (t) = ξt, where ξ = 1. The singular holonomy group is the maximal
subgroup obtained by iteration of the adjunction process made at the corners of a
branch. In our case, as e−2πip/q is a primitive q-root of unity, ξ = (e−2πip/q )m , and
so, hD m D
i = hj , for some m ≥ 1. So, the adjunction of hi to the holonomy group in
this case does not increase it.
Consider now the situation studied by [MeSa]. After reduction of singularities,
removing the only dicritical component, there are two chains of divisors and all the
corners are resonant linearizable. Take a transversal Σ on Dp , the component where
the separatrix lies. Iterating the above construction, we have that the singular
holonomy group agrees with the usual one of Dp . In particular, it is finite. If
P ∈ Σ, the leaf through P cuts the dicritical component D. Assume that there
exists a rational function f on D, invariant by the singular holonomy, as in (ii) of
the Theorem above. The previous construction defines a value r(P ) on P ∈ Σ, and
so, a rational function r(T ) such that r ◦ h = r, for h an element of the holonomy
group of Dp . This is the condition of [MeSa].
Conversely, if such a r(T ) exists, by previous considerations, r ◦ h = r for every h
in the singular holonomy group, and a rational function on D can be defined. These
considerations relate the conditions stated by Camacho and the ones by Meziani
and Sad for the existence of a meromorphic first integral. We will try to reproduce
the last one in our three dimensional, quasi-ordinary case.
The plan of the paper is as follows. In Section 2, following [CC, C] we will
recall the main notions concerning pre-simple singularities of foliations in dimension
three, focusing in the existence of a meromorphic first integral. In Sections 3 and
4 we will investigate the existence of holomorphic and meromorphic first integrals,
respectively, for our three-dimensional foliations, relating the existence of these
objects with the existence of the corresponding objects in dimension two. Finally,
in Section 5 we will study the dicriticalness of such a foliation.

Acknowledgments. The authors want to thank Felipe Cano Torres, from the
Universidad de Valladolid, for his continuous support and encouragement, and for
fruitful conversations during the preparation of this work. First and third author
4 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

wants to thank Universidad de Valladolid, and second author wants to thank Pon-
tificia Universidad Católica del Perú for the hospitality during several visits to both
institutions.

2. Remarks on pre-simple singularities

In this section we will recall the notion of pre-simple and simple singularities of
holomorphic foliations, as defined in [CC] and [C]. We will establish some results
concerning their rigidity and the existence of meromorphic first integrals.
Let F be a germ of holomorphic foliation in (Cn , 0), generated by an integrable
1-form ω, adapted to a normal
Q crossings divisor E. Choosing a coordinate system
such that E is written as i∈A xi = 0, a meromorphic generator of F is
X dxi X
ω= ai + ai dxi , (a1 , a2 , . . . , an ) = 1.
xi
i∈A i∈A
/
The component xi = 0 of the divisor will be invariant by F if xi does not divide
ai . A component xi = 0 is called dicritical if it is not invariant. Denote
A∗ = {i ∈ A; xi - ai }
the set of indexes corresponding to non-dicritical components.
The dimensional type of F, denoted τ , is the minimum number of coordinates
needed to write a generator of F. Equivalently, it is the codimension of the vector
space
X (F) = {X(0); X ∈ X (Cn , 0), ω(X) = 0},
(where X (Cn , 0) denotes the space of germs of holomorphic vector fields at the
origin in (C2 , 0)), according to the rectification theorem for regular vector fields.
The adapted order of F at the origin is defined as
ν(F, E) := min{ν(ai ); 1 ≤ i ≤ n},
where ν(ai ) denotes the (usual) order of ai as formal power series. The adapted
multiplicity of F at 0 is
µ(F, E) = min{ν(ai ); i ∈ A} ∪ {ν(ai ) + 1; i ∈
/ A}.
The resonance invariant of (F, E), Rs(F, E) is:
• Rs(F, E) = 1 if ν(F, E) = µ(F, E) and F is radially dicritical, i.e., the
new component of the exceptional divisor that appears after blowing-up the
origin is dicritical.
• Rs(F, E) = 2 if ν(F, E) = µ(F, E), Rs(F, E) 6= 1 and, in some local coor-
dinates, there is a map Φ : A → Z>0 such that
!
X
ν Φ(i)ai > ν(F, E).
i∈A

• Rs(F, E) = 0 in other case.

Definition 1. The singularity at the origin is pre-simple for (F, E) if E is non-
dicritical and one of the following holds:
1. Either ν(F, E) = 0.
2. Or ν(F, E) = µ(F, E) = 1, Rs(F, E) = 0, and the directrix, i.e., the vector
space defined by {In1 (ai ) = 0; i ∈ A} (where In1 denotes the first order
terms) has codimension one and normal crossings with E.
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 5

According to [C, Prop. 46], there are formal coordinates (x1 , x2 , . . . , xn ) such
that, any pre-simple singularity of dimensional type τ is given by one of the following
formal meromorphic types:
Pτ Qτ
A. ω = i=1 λi dx xi , with
i
i=1 λi 6= 0.
Pk Pτ
B. ω = i=1 pi xi + ψ(x1 · · · xpkk ) i=2 αi dx
dxi p1
xi , where pi are natural numbers,
i

p1 6= 0, ψ(0) = 0, αi 6= 0 for i = k + 1, . . . , τ . This will be called the resonant

case. Pk p2 pk Pτ
C. dx1 − x1 i=2 pi dx xi + x2 · · · xk
i dxi
i=1 αi xi , with (p2 , . . . , pk ) ∈ N
k−1
\
{(0, . . . , 0)}, αi ∈ C. This is the Dulac case.
We will focus here in the three dimensional case: singularities of holomorphic fo-
liations are of dimensional type τ = 2 or 3. According to the previous classification,
pre-simple singularities are:
1. If τ = 2, pre-simple is equivalent to the non-nilpotence of the linear part of
the dual vector field.
2. If τ = 3, they correspond to one of the formal meromorphic models A, B or
C, as above.
According to [CC, C], any germ of foliation in dimension three can be trans-
formed into another one having only pre-simple singularities, using an appropriate
sequence of blow-ups with permissible centers. If the foliation has a meromorphic
first integral, all the pre-simple singularities that appear after this process must also
have meromorphic first integral, and as a consequence, the leaves of the foliation
must be closed. This fact imposes some restrictions that we will explore here.
P3
Case A: Consider a pre-simple singularity, formally equivalent to i=1 λi dx̂ x̂i . It
i

λ1 λ2 λ3
has a multiform first integral x̂1 x̂2 x̂3 . According to [CeMa, Cor. 1.2., pg.
115] it is semi-divergent and consequently the singularity has three conver-
gent separatrices, tangent to xi = 0. Consequently, in appropriate analytic
coordinates the foliation is defined by
3
X dxi
λi (1 + bi (x)) ,
i=1
xi

with bi (0) = 0. Consider the separatrix x1 = 0. Its holonomy group is

generated by two diffeomorphisms h2 (x1 ), h3 (x1 ), such that
λk
hk (x1 ) = e− λ1 2πi x1 + · · · .
If the leaves of the foliation are closed, the holonomy is periodical, i.e.,
λk pk
= , pi ∈ Z.
λ1 p1
So, it is linearizable and then, the foliation also is, i.e. a generator is
3
X dxk
ω= pk ,
xk
k=1
3
with (p1 , p2 , p3 ) ∈ (Z \ {0}) . So, existence of a meromorphic first integral
for a pre-simple singularity of type A, and dimensional type τ = 3 implies
linearizability.
6 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

Case B: If k = 2, the foliation is a saddle-node. Indeed, assume p3 = 0, and

take a transversal section x1 = ε, for ε 6= 0 small enough. The resulting
two-dimensional foliation is
x3 (p2 + α2 ψ(εp1 xp22 ))dx2 + α3 x2 ψ(εp2 xp22 )dx3 ,
which is a saddle-node. Consequently, the leaves are not closed, and a mero-
morphic first integral cannot exist.
If k = 1, and the residual spectrum of the singularity [C, Rem. 20] is
resonant, i.e., α2 /α3 ∈ Q<0 , after blowing-up the x-axis a certain number
of times we arrive to a saddle-node again. If this residual spectrum is non
resonant, or k = 3, the singularity is simple. In this case, the holonomy group
of x1 = 0 has been computed in [CeMo, Section 4]. Under the assumption
that the leaves are closed, this group turns out to be linearizable, and the
foliation has a holomorphic first integral such that, in appropriate coordinates
can be written as xp11 xp22 xp33 , so, in fact, it is of the previous type A.
Case C: An analytic generator of the foliation is
dx2 dx3
ω = dx1 − (p2 x1 + · · · ) − (p3 x1 + · · · ) .
x2 x3
Consider the following commuting vector fields, tangent to ω:
∂ ∂
X = (p2 x1 + · · · ) + x2
∂x1 ∂x2
∂ ∂
Y = (p3 x1 + · · · ) + x3 .
∂x1 ∂x3
Take b > 0 and define Xb = X + bY . The linear part of Xb is
∂ ∂ ∂
Xb,1 = (p2 + bp3 )x1 + x2 + bx3 .
∂x1 ∂x2 ∂x3
The triple (q1 , q2 , q3 ) ∈ N3 is a resonance for Xb if
q1 (p2 + bp3 ) + q2 + q3 b = p2 + bp3 .
Assume b ∈
/ Q. The previous equality means that
q1 p2 + q2 = p2 ; q1 p3 + q3 = p3 ,
and so, q1 ∈ {0, 1}. If q1 + q2 + q3 ≥ 2, (q1 , q2 , q3 ) = (0, p2 , p3 ), and by
Poincaré normalization theorem, Xb is analytically equivalent to
∂ ∂ ∂
Ab = ((p2 + bp3 )x1 + a1 (xp22 xp33 )) + x2 + bx3 .
∂x1 ∂x2 ∂x3
Consider b0 6= b, b0 ∈
/ Q, and construct analogously Ab0 . The 1-form
ω̃ = ιAb ιAb0 (dx1 ∧ dx2 ∧ dx3 )
is a generator of the foliation and implies that the normal form in this case
is, in fact, convergent. x2 = 0, x3 = 0 are separatrices. The holonomy
generators h1 (x3 ), h2 (x3 ) of x3 = 0 verify
1 1
− p
+ α3 log(h2 (x3 )) + α2 = − p3 + α3 log(x3 ),
h1 (x3 ) 3 x3
1 1
− α3 log(h2 (x3 )) = p3 − α3 log(x3 ).
h2 (x3 )p3 x3
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 7

The leaves being closed, α3 = 0. This implies that

h2 (x3 )p3 = xp33 , i.e. h2 (x3 ) = ξx3 , ξ p3 = 1.
We also have that
1 1
− p
+ α2 = − p3 ,
h1 (x3 ) 3 x3
and so,
xp33
h1 (x3 )p3 = .
1 + α2 xp33
So, h1 (x3 ) = ηx3 + ∗ · xp33 + · · · , with η p3 = 1. Periodicity of h1 implies that
α2 = 0. This forbids singularities of type C with (α2 , α3 ) 6= (0, 0) to have a
meromorphic first integral.

3. Holomorphic first integrals

Let us study the problem of the existence of holomorphic first integrals. Accord-
ing to [MM], a singular holomorphic foliation has a first integral if and only if all
the leaves are closed and only a finite number of them contain the singularity in
their closure. So, it is a property of topological nature.
Let us consider a nilpotent singularity on (Cm+1 , 0), i.e., a singularity of folia-
tion of the type zdz + · · · . Under a coordinate change, using Loray’s Preparation
Theorem [L], this foliation can be generated by a 1-form
Ω := Ωf,n,p,α,U = d(z 2 + f n ) + αf p U (f )dz,
where f (x) : (Cm , 0) → (C, 0), α ∈ C∗ , U (t) ∈ C{t}, U (0) = 1. Write f =
f1r1 · · · fkrk , gcd(r1 , . . . , rk ) = 1. Consider the map
Φ : (Cm+1 , 0) −→ (C2 , 0)
(x, z) 7−→ (f (x), z),
so, Ω = Φ∗ ωn,p,α,U , where
ωn,p,α,U = d(z 2 + tn ) + αtp U (t)dz.
It is clear that, if h(t, z) is a holomorphic first integral for ωn,p,α,U , h ◦ Φ is a
holomorphic first integral for Ω. Let us show the converse in this case.
The map Φ is a surjective submersion on (Cm+1 , 0) \ S, where
S := (x, z) ∈ (Cm+1 , 0); ∇f (x) = 0 ⊆ S 0 := (x, z) ∈ (Cm+1 , 0); f (x) = 0 .
 

Proposition 2. If L is a closed leaf for Ω, Φ(L) is a closed leaf for ωn,p,α .

Proof. L \ S 0 is closed on (Cm+1 , 0) \ S 0 , and Φ(Cm+1 \ S 0 ) = C∗ × C (at least as a
germ around the origin, we won’t precise the neighbourhood in which this equality
is valid). By [CLN, Thm. 3, Ch. 2], L \ S 0 is a connected component of Φ−1 (F 0 ),
F 0 = Φ(L \ S 0 ) being a leaf of Fωn,p,α,U in C∗ × C.
Consider a trivializing chart U for the foliation defined by ωn,p,α,U around P ∈
F 0 , where the foliation is defined by the levels of a submersion g : U → C. For every
point (t0 , z0 ) ∈ U , the inverse image by Φ is the variety f (x) = t0 , z = z0 . This
variety has a finite number of connected components, so, reducing U if necessary,
Φ−1 (U ) consists in a finite number of connected trivializing sets. The leaf L \ S 0
being closed, in each of these open sets there are a finite number of plaques of L\S 0 ,
so F has a finite number of components in U and then, it is closed.
8 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

z
z

φ(x, y, z) = (f (x, y), z)

z0
L z0
F = φ(L)

t
y

Figure 1. The image of a closed leaf is closed

So, Φ(L \ S 0 ) is closed on C∗ × C. As

Ω ∧ df = (2z + αf p U (f ))dz ∧ df,

S 0 is transversal to FΩ outside z = 0. Consider now the leaf F , such that L is a

connected component of Φ−1 (F ). In fact, F = Φ(L). This leaf cuts the axis t = 0
transversally outside z = 0. As L is closed, the “vertical” axis z cuts transversally
L in a discrete number of points, so F does also. Then F is closed outside the
origin. Apply finally Remmert-Stein Theorem in order to extend F up to the
origin, concluding that Φ(L) is closed.


Consequently, if Ω has a holomorphic first integral, as Φ is surjective, the leaves

of the foliation generated by ωn,p,α,U are closed. If there exists infinitely many
separatrices, their inverse images would be separatrices of Ω containg the analytic
set Φ−1 (0), which contradicts the fact that Ω has a holomorphic first integral. When
a singular holomorphic foliation has a first integral, [MM] allows to conclude the
following result:

Theorem 3. The holomorphic foliation FΩf,n,p,α,U in (Cm+1 , 0) has a holomorphic

first integral if and only if Fωn,p,α,U has a holomorphic first integral.

In [CeM], criteria are given for the existence of a holomorphic first integral in
terms of the projective holonomy. The generators of this group must be of finite
order, and also the group, which implies its abelianity. Combining these criteria
with the results in [Me], we have:

Corollary 4. A nilpotent singular foliation FΩ defined by Ω = Ωf,n,p,α,U = 0 has

a holomorphic first integral if and only if one of the following conditions occurs:
1. n < 2p, the projective holonomy of Fωn,p,α,U is abelian, and its generators
have finite order.
(16 + r)2
2. n = 2p, α2 6= , ∀ r ∈ Q≥0 , and also the projective holonomy of
16 + 2r
Fωn,p,α,U is abelian, and its generators have finite order.
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 9

Note that the conditions on n, p are necessary to guarantee that no dicriti-

cal components and no saddle-nodes appear in the reduction of singularities of
Fωn,p,α,U .
Let us mention that in [M], an example of a nilpotent foliation without a holo-
morphic first integral is presented. The projective holonomy of the foliation defined
by the 1-form ω = d(y 2 + x3 ) + x(2xdt − 3ydx) has two generators h1 , h2 , with
h21 = h32 = id but they don’t commute, so the projective holonomy is not finite.

4. Meromorphic first integrals in the quasi-ordinary case

Let us focus now in cuspidal, quasi-ordinary singularities of holomorphic folia-
tions in dimension three. A generator in appropriate coordinates is
Ω = d(z 2 + (xp y q )k ) + α(xp y q )n U (xp y q )dz,
where α ∈ C∗ , (p, q) = 1, U (t) ∈ C{t}, U (0) = 1 [FM, FMN]. In this section
we are interested in the existence of pure meromorphic first integrals (i.e., non
holomorphic, as defined in [CeMa]). The following result restricts the values of n
and k that we have to study:
Proposition 5. If a quasi-ordinary singularity of foliation, generated by a 1-form
d(z 2 + (xp y q )k ) + α(xp y q )n U (xp y q )dz,
has a pure meromorphic first integral, then k = 2n.
Proof. If 2n > k, it is a generalized surface. If 2n < k, blow-up pn times the y-axis,
and qn times the x-axis. The equations of this transform are, in the main chart,
z = txpn y qn . The strict transform of ω is
Ω̃ = (2nt2 + k(xp y q )k−2n + αntU )(pydx + qxdy) + xy(2t + αU )dt.
The origin turns out to be a simple singularity of type B, resonant, and there
is another singularity corresponding to the point t = − α2 . This singularity is a
saddle-node, so, the foliation cannot have a first integral.
Consequently, 2n = k. 

From now on, we will assume that k = 2n. After blowing up qn times the x-axis,
and pn times the y-axis, the resulting foliation turns out to be, after eliminating
common factors,
Ω̃ = (2z 2 + αzU + 2)(pnydx + qnxdy) + xy(2z + αU )dz.
The only singular points away from the intersections of the components of the
exceptional divisor are those defined by (x = 0, 2z 2 + αz + 2 = 0). Assume that
α 6= ±4, so this equation has two different solutions α1 , α2 . Consider the point
corresponding z = α1 . After a coordinate change z − α1 = z̃, this point P1 is the
origin, and in the new coordinates the 1-form that generates the foliation is
 
ω = 2z̃(z̃ + α1 − α2 ) + α(z̃ + α1 )Ũ (pnydx + qnxdy) + xy(2z̃ − 2α2 + αŨ )dz̃,

where U = 1 + Ũ . The terms of lower order of Ω are, in these coordinates,

 
dx dy α2 dz̃
ΩL,1 = 2xyz̃(α1 − α2 ) pn + qn + .
x y α2 − α1 z̃
10 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

Similarly, in the coordinates centered around the point P2 corresponding to z = α2 ,

the lower order terms are
 
dx dy α1 dz̃
ΩL,2 = 2xyz̃(α2 − α1 ) pn + qn + .
x y α1 − α2 z̃
In the paper [FM], the authors studied the generic case where
(16 + r)2
α2 6= ,
16 + 2r
for every r ∈ Q≥0 , where the singularities are already reduced. We are now inter-
ested in the complementary case
(16 + r)2
α2 = ,
16 + 2r
for some r ∈ Q>0 . In this case, one of the numbers α2α−α 2
1
, α1α−α
1
2
is a negative
rational, and the other one, a positive rational, this last point being already reduced.
Assume α2α−α 2
1
∈ Q<0 , and then, P2 is simple resonant. The singularity at P1 may
fall in one of the following two cases:
1. Either it is linearizable, and then, it is dicritical.
2. Or it is of Dulac type.
In the second case, as we have seen in Section 2, the foliation cannot have a mero-
morphic first integral. So, we will suppose in the sequel that we are in the first case,
and then, the foliation is dicritical. Locally in a neighbourhood of P1 , the foliation
is defined by
 
dx dy dz̃
Ω1 = xyz pn + qn −s ,
x y z̃
where s ∈ Q>0 .
In order to linearize Ω around P1 , let us consider the following vector fields:
∂ 1 ∂
X1 = g(x, y, z̃) − x
∂ z̃ pn ∂x
∂ 1 ∂
X2 = g(x, y, z̃) − y ,
∂ z̃ qn ∂y
where
2z̃(z̃ + α1 − α2 ) + α(z̃ + α1 )Ũ α2 − α1
g(x, y, z̃) = = z + h.o.t.
2z̃ − 2α2 + αŨ α2
They are commuting and, in the considered case, α2α−α 2
1
∈ Q<0 . The linearization is
obtained after an analytic transformation in the z̃ variable: z̃ = z̃(x, y, z), and then,
a separatrix z̃ = 0, that perhaps is not the only smooth separatrix corresponding
to this point. If this is the case, we choose one of the separatrices. Anyway, there
is an infinite number of separatrices through this point, defined by the equation
z s1 = Cxpns2 y qns2 , for C ∈ C,
where s = ss21 . Assume that a meromorphic first integral exists, F (x, y, z). In the
coordinates around P1 , it is F̃ (x, y, z̃). Let us follow the same scheme of [MeSa]:
take a transversal Σ to the component of the divisor defined by y = 0, i.e., the
special component in the notation of [FM]. For simplicity, assume that Σ ≡ (x =
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 11

z̃ = 1). f˜(t) = F̃ (1, t, 1) is the restriction of F̃ to Σ, defined for small |t|. If t ∈ C

is small enough, the leaf through (1, t, 1) is
1
z 2 = qn xpn y qn ,
t
and so, it approaches any neighbourhood of the origin. Defining
f˜(t) = F (x, y, (t−qn xpn y qn )1/s ),
f˜(t) can be extended to a function on C̄, necessarily rational. Let h0 , h1 be the
generators of the holonomy of this component, calculated in the transversal Σ.
The fact that f˜(t) is the restriction of a meromorphic first integral implies that
f˜ ◦ h1 = h1 , f˜ ◦ h2 = h2 , i.e., any element h(t) of the holonomy group of the special
component is fixed by the rational function f (t).
Conversely, assume that a rational function f (t) exists on a transversal Σ ∼ = C̄,
invariant by the holonomy. Every leaf cuts Σ so, following the leaves a first integral
may be defined in a neighbourhood of this special component of the exceptional
divisor. The other singularities, by [MM], have holomorphic first integrals, so this
meromorphic first integral extends to a neighbourhood of the origin, and defines a
meromorphic first integral for the foliation. So, we have proved:
Theorem 6. A meromorphic first integral exists for the quasi-ordinary cuspidal
foliation generated by Ω = d(z 2 + (xp y q )k ) + α(xp y q )n U (xp y q )dz if and only if
k = 2n and a rational function exists invariant by the projective holonomy of the
special component of the divisor that appears after reduction of the singularities as
in [FM].

5. Dicriticalness
Finally, we shall investigate the relation about the dicriticalness of a nilpotent
foliation in (C2 , 0) with the dicriticalness of its pull-back via a map
Φ : (C3 , 0) −→ (C2 , 0)
(x, y, z) 7−→ (f (x, y), z).
So, we will consider cuspidal, nilpotent singularities generated by a 1-form
Ω = d(z 2 + f (x, y)n ) + αf (x, y)p U (f )dz,
where U (t) is a unit, U (0) = 1.
For, let us recall, following [C], that a germ of singular holomorphic foliation
F in (Cn , 0), generated by a 1-form ω, is dicritical if there exists an analytic map
ϕ : (C2 , 0) → (Cn , 0), not invariant by F (i.e., ϕ∗ ω 6= 0), such that ϕ∗ F is the
foliation (dx = 0), and ϕ(y = 0) is invariant by F.
With this characterization, if Ω defines a dicritical foliation on (C3 , 0), also
ωn,p,α,U defines a dicritical foliation in (C2 , 0). Indeed, if ϕ : (C2 , 0) → (C3 , 0)
exists, in the conditions of the previous characterization, and Φ◦ϕ = ρ, ρ∗ ωn,p,α,U =
ϕ∗ Φ∗ ωn,p,α,U = ϕ∗ Ω 6= 0, ρ∗ ωn,p,α,U ≡ (dx = 0), and ρ(y = 0) = Φ(ϕ(y = 0)) is
invariant.
Φ
It is not true, in general, that given a map (Cn , 0) −→ (Cm , 0), if ω defines a
∗
dicritical foliation in (C , 0), then Φ ω also defines a dicritical foliation on (Cn , 0).
m

In particular, if m = 2, dicritical foliations are exactly foliations with an infinite

number of separatrices, but in higher dimensions, it is not even true that they have
a separatrix [J]. So, for a general map Φ, it is not always possible to study the
12 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

dicriticalness of the pull-back from the point of view of the existence of infinitely
many separatrices. We will see that in our case the situation is different, and much
simpler. If ωn,p,α,U defines a dicritical foliation on (C2 , 0), then we must have,
according to [Me, MeSa]
1. n = 2p.
2. α 6= ±4.
3. Under (i) and (ii), after p blowing-ups, the quotient of the eigenvalues
2 −α1 )
at the singular points (different from corners) are, respectively, 2p(α
2α1 +α ,
2p(α1 −α2 )
2α2 +α , where α1 , α2 are the roots of the equation 2y 2 + αy + 2 = 0. It
is necessary that one of these numbers is a positive rational. In this case, if
it is not an integer or the inverse of an integer, then it is dicritical. In the
remaining situation, either it is dicritical or of Dulac type, depending on the
coefficients of U (x).
Denote, as in the Introduction, ωp,α,U = d(z 2 + t2p ) + αtp U (t)dz. Let us see
that, in our case, if ωp,α,U is dicritical, then Φ∗ ωp,α,U also is. For, we shall use the
characterization from [C] explicitely. In order to locate a dicritical component in
the reduction of singularities of ωp,α,U (in fact, the only one), we must follow the
following steps:
1. Blow-up the origin p times. In an appropriate chart, this is represented by
the map (t, z) → (t, tp z).
2. Make a translation z 7→ z + αi , with previous notations.
3. Blow up the origin a certain number of times, depending on the quotient
of the eigenvalues at the singular point. This is done via a map (t, z) 7→
(tn1 z m1 , tn2 z m2 ), with n1 m2 − m1 n2 = 1.
After these transformations, we arrive at a corner where one of the axis is a dicritical
component of the exceptional divisor. The composition of the transformations 1.,
2. and 3. is
E(t, z) = (tn1 z m1 , tpn1 z pm1 (tn2 z m2 + α1 )) .
Assume, for instance, that the horizontal axis z = 0 is transversal to the foliation
E ∗ F defined by the saturated of E ∗ ωp,α,U . In this case, this foliation is generated
by dt + B(t, z)dz. This foliation can be rectified by a diffeomorphism S(t, z) =
(tS1 (t, z), z), with S1 (0, 0) 6= 0. This means that S ∗ (dt + B(t, z)dz) ∧ dx = 0.
Assume, first, that we are in the easier case where f (x, y) = xp1 y p2 , i.e., the foliation
Φ∗ ωp,α,U is of quasi-ordinary type. We will call this map E ◦ S a dicriticalness
section, i.e., a map verifying the dicriticalness condition of [C] as given above. It
is our objective to lift this map to (C3 , 0) in order to show that Ω := Φ∗ ωp,α,U is
dicritical. The map E ◦ S should be lifted to (C3 , 0) in a map σ = (σ1 , σ2 , σ3 ) such
that Φ◦σ = E◦S, but this is not always possible, as in particular it would imply that
σ1p1 σ2p2 = tn1 S1 (t, z)n1 z m1 , which is not possible if p1 , p2 are big enough. But, let
us observe that, if we define ϕ = (ϕ1 , ϕ2 ) : (C2 , 0) → (C2 , 0) by ϕ(t, z) = (tr1 , z r2 ),
the composition E ◦ S ◦ ϕ is again a dicriticalness section for ωp,α,U . Choosing
r1 = p1 , r2 = p2 , this map can be lifted to (C3 , 0) defining
σ1 (t, z) = tn1 S1 (tp1 , z q1 )n1 /p1 ,
σ2 (t, z) = z m1 ,
σ3 (t, z) = tpn1 p1 S1 (tp1 , z q1 )pn1 z pm1 q1 (tn2 p1 S1 (tp1 , z q1 )n2 z m2 q1 + α1 ).
So, the foliation Φ∗ ωn,p,α is also dicritical.
 DICRITICAL NILPOTENT HOLOMORPHIC FOLIATIONS 13

Let us consider now the general case, i.e., where Φ(x, y, z) = (f (x, y), z), for a
general f (x, y). In this case, consider a local monomialization of f (x, y), i.e., a local
map ρ(x, y) : (C2 , 0) → (C2 , 0) such that f (ρ(x, y)) = xa y b V (x, y), V being a unit,
a, b 6= 0. As before, modify the dicriticalness section E ◦ S defining ϕ1 (t, z) = ta .
This implies that we must choose σ1 (t, z) = tn1 , σ2 (t, z) = z m1 . In this case, the
equality f (ρ(σ1 , σ2 )) = ϕ1 (t, z)n1 S1 (ϕ1 (t, z), ϕ2 (t, z))n1 ϕ2 (t, z)m1 turns out to be
z b V (tn1 , y m1 )1/m1 = S1 (ta , ϕ2 (t, z))n1 /m1 ϕ2 (t, z),
which has a unique solution ϕ2 (x, y) by Implicit Function Theorem. Now, σ3 (x, y)
can be defined accordingly.
This ends the construction. We sum up previous considerations in the following
theorem:
Theorem 7. The foliation in (C3 , 0) defined by the 1-form Ω = d(z 2 + f (x, y)2p ) +
αf (x, y)p U (f )dz is dicritical if and only if ω = d(z 2 + t2p ) + αtp U (t)dz defines a
dicritical foliation in (C2 , 0).
Remark 8. Previous result also holds in dimension greater than three, as all the
arguments are valid there. The differences in the proof for this extended case consist
only in technical points, and we shall not reproduce it here.
Dicritical foliations in dimension two are also characterized for the existence
of infinitely many separatrices at the origin. This is not true in higher dimension,
where a dicritical foliation may not have any separatrix. Nevertheless, this condition
is still sufficient: a foliation with infinitely many separatrices is always dicritical. In
the cuspidal case we study in this paper, this situation cannot happen: any nilpotent
dicritical foliation generated by a 1-form Ω = d(z 2 + f (x, y)2p ) + αf (x, y)p U (f )dz
has infinitely many separatrices. In fact, the foliation in (C2 , 0) generated by ωp,α,U ,
in the dicritical case, has separatrices that can be parameterized by Puiseux series
y = α1 xp + · · · , series with rational exponents. These series, by composition with
Φ, give infinitely many separatrices for Ω.

References
[Ca] C. Camacho. Dicritical Singularities of Holomorphic Vector Fields. Contemporary
Mathematics vol. 269 (2001), 39–45.
[CLN] C. Camacho, A. Lins Neto. Teoria geométrica das folheações. IMPA, Projeto Euclides
(1979).
[C] F. Cano. Reduction of the singularities of codimension one holomorphic foliations in
dimension three. Annals of Math. 160 (2004), 907–1011.
[CC] F. Cano, D. Cerveau. Desingularization of non-dicritical holomorphic foliations and
existence of separatrices. Acta Math., 169 (1992), 1–103.
[CeMa] D. Cerveau and J.-F. Mattei. Formes intégrables holomorphes singulières. Astérisque
97 (1982).
[CeM] D. Cerveau, R. Moussu. Groupes d’automorphismes de (C, 0) et équations
différentielles ydy + · · · = 0, Bull. Soc. Math. France 116 (1988) 459-488.
[CeMo] D. Cerveau, J. Mozo Fernández. Classification analytique des feuilletages singuliers
réduits de codimension 1 en dimension n ≥ 3. Erg. Th. and Dyn. Systems 22 (2002),
1041–1060.
[Co] V. Cossart. Desingularization in dimension 2, in Resolution of surface singularities.
Lecture Notes in Mathematics, 1101, 79–98. Springer-Verlag (1984).
[FM] P. Fernández Sánchez, J. Mozo Fernández. Quasi-ordinary cuspidal foliations in
(C3 , 0). Journal of Differential Equations 226 (2006), 250–268.
14 PERCY FERNÁNDEZ-SÁNCHEZ, JORGE MOZO-FERNÁNDEZ, AND HERNÁN NECIOSUP

[FMN] P. Fernández Sánchez, J. Mozo Fernández, H. Neciosup. On codimension one

foliations with prescribed cuspidal separatriz. J. Differential Equations 256 (2014) 1702–
1717.
[G] J. Giraud. Desingularization in low dimension, in Resolution of surface singularities.
Lecture Notes in Mathematics, 1101, 50–78. Springer-Verlag (1984).
[J] J.P. Jouanolou. Équations de Pfaff algébriques. Lecture Notes in Mathematics, 708.
Springer, 1979.
[L] F. Loray, A preparation theorem for codimension one foliations, Ann. of Math. 163
(2006), 709–722.
[MM] J.F. Mattei, R. Moussu, Holonomie et intégrales premiéres, Ann. Sci. École Normale
Sup. 13 (1980) 469–523.
[Me] R. Meziani. Classification analytique d’équations différentielles ydy + · · · = 0 et espaces
de modules. Bol. Soc. Brasil Mat. 27 (1996), 23–53.
[MeSa] R. Meziani, P. Sad. Singularités nilpotentes et intégrales premières. Publ. Mat. 51
(2007), 143–161.
[M] R. Moussu. Holonomie évanescente des équations différentielles dégénérées trans-
verses, in Singularities and Dynamical systems. North-Holland (1985), 151–173.

(Percy Fernández) Dpto. Ciencias - Sección Matemáticas, Pontificia Universidad Católica

del Perú, Av. Universitaria 1801, San Miguel, Lima 32, Peru
E-mail address: [email protected]

(Jorge Mozo-Fernández) Dpto. Álgebra, Análisis Matemático, Geometrı́a y Topologı́a,

Facultad de Ciencias, Universidad de Valladolid, Campus Miguel Delibes, Paseo de
Belén, 7, 47011 Valladolid - Spain
E-mail address: [email protected]

(Hernán Neciosup) Dpto. Ciencias - Sección Matemáticas, Pontificia Universidad Católica

del Perú, Av. Universitaria 1801, San Miguel, Lima 32, Peru
E-mail address: [email protected]