ON NORMAL FORMS FOR LEVI-FLAT HYPERSURFACES

WITH AN ISOLATED LINE SINGULARITY

ARTURO FERNÁNDEZ PÉREZ

Abstract. We prove the existence of normal forms for some local real-analytic
Levi-flat hypersurfaces with an isolated line singularity. We also give sufficient
arXiv:1304.2669v2 [math.CV] 27 Jan 2014

conditions for that a Levi-flat hypersurface with a complex line as singularity

to be a pullback of a real-analytic curve in C via a holomorphic function.

1. Introduction
Let M ⊂ U ⊂ Cn be a real-analytic hypersurface, where U is an open set and
denote by M ∗ the regular part, that is, near each point p ∈ M ∗ , the variety M is
a manifold of real codimension one. For each p ∈ M ∗ , there is a unique complex
hyperplane Lp contained in the tangent space Tp M ∗ , and consequently defines a
real-analytic distribution p 7→ Lp of complex hyperplanes in Tp M ∗ , the so-called
Levi distribution. We say that M is Levi-flat, if the Levi distribution is integrable
in sense of Frobenius. The foliation defined by this distribution is called Levi-
foliation. The local structure near regular points is very well understood, according
to E. Cartan, around each p ∈ M ∗ we can find local holomorphic coordinates
z1 , . . . , zn such that M ∗ = {Re(zn ) = 0}, and consequently the leaves of Levi-
foliation are imaginary levels of zn . The singular case was studied by Burns-Gong
[2], The authors classified singular Levi-flat hypersurfaces in Cn with quadratic
singularities and also proved the existence of a normal form, in the case of generic
(Morse) singularities. In [4], Cerveau-Lins Neto have proved that a local real-
analytic Levi-flat hypersurface M with a sufficiently small singular set is given by
the zeros of the real part of a holomorphic function.
The aim of this paper is to prove the existence of some normal forms for local
real-analytic Levi-flat hypersurfaces defined by the vanishing of real part of holo-
morphic functions with an isolated line singularity (for short: ILS). In particular,
we establish an analogous result like in Singularity Theory for germs of holomorphic
functions.
The main motivation for this work is a result due to Dirk Siersma, who introduced
in [14] the class of germs of holomorphic functions with an ILS. More precisely, let
On+1 := {f : (Cn+1 , 0) → C} be the ring of germs of holomorphic functions and let
m be its maximal ideal. If (x, y) = (x, y1 , . . . , yn ) denote the coordinates in Cn+1
and consider the line L := {y1 = . . . = yn = 0}, let I := (y1 , . . . , yn ) ⊂ On+1 be its
ideal and denote by DI the group of local analytic isomorphisms ϕ : (Cn+1 , 0) →
(Cn+1 , 0) for which ϕ(L) = L. Then DI acts on I 2 and for f ∈ I 2 , the tangent
space of (the orbit of) f with respect to this action is the ideal defined by
∂f ∂f
τ (f ) := m. + I.
∂x ∂y

Date: January 28, 2014.

2010 Mathematics Subject Classification. Primary 32V40 - 32S65.
Key words and phrases. Levi-flat hypersurfaces - Holomorphic foliations.
1
2 ARTURO FERNÁNDEZ PÉREZ

and the codimension of (the orbit) of f is

I2
c(f ) := dimC .
τ (f )
A line singularity is a germ f ∈ I 2 . An ILS is a line singularity f such that
c(f ) < ∞. Geometrically, f ∈ I 2 is an ILS if and only if the singular locus of
f is L and for every x 6= 0, the germ of (a representative of) f at (x, 0) ∈ L is
equivalent to y12 + . . . + yn2 . In a certain sense ILS are the first generalization of
isolated singularities. D. Siersma proved the following result. (The topology on
On+1 is introduced as in [5, p. 145]).
Theorem 1.1. A germ f ∈ I 2 is DI -simple (i.e. c(f ) < ∞ and f has a neigh-
borhood in I 2 which intersects only a finite number of DI -orbits) if and only if f is
DI -equivalent to one the germs in the following table

Type Normal form Conditions

A∞ y12 + y22 + . . . + yn2
D∞ xy12 + y22 + . . . + yn2
Jk,∞ xk y12 + y13 + y22 + . . . + yn2 k≥2
T∞,k,2 x2 y12 + y1k + y22 + . . . + yn2 k≥4
Zk,∞ xy13 + xk+2 y12 + y22 + . . . + yn2 k≥1
W1,∞ x3 y12 + y14 + y22 + . . . + yn2
T∞,q,r xy1 y2 + y1q + y2r + y32 . . . + yn2 q≥r≥3
Qk,∞ xk y12 + y13 + xy22 + y32 . . . + yn2 k≥2
S1,∞ x2 y12 + y12 y2 + y32 + . . . + yn2
Table 1. Isolated Line singularities
The singularities in Theorem 1.1 are analogous of the A-D-E singularities due to
Arnold [1]. A new characterization of simple ILS have been proved by A. Zaharia
[15]. We prove the existence of normal forms for Levi-flat hypersurfaces with an
ILS.
Theorem 1. Let M = {F = 0} be a germ of an irreducible real-analytic hypersur-
face on (Cn+1 , 0), n ≥ 3. Suppose that
(1) F (x, y) = Re(P (x, y)) + H(x, y), where P (x, y) is one of the germs of the
Table 1.
(2) M = {F = 0} is Levi-flat.
(3) H(x, 0) = 0 for all x ∈ (C, 0) and j0k (H) = 0, for k = deg(P ).
Then there exists a biholomorphism ϕ : (Cn+1 , 0) → (Cn+1 , 0) preserving L such
that
ϕ(M ) = {Re(P (x, y)) = 0}.
This result is a Siersma’s type Theorem for singular Levi-flat hypersurfaces. We
remark that the function H is of course restricted by the assumption that M is
Levi flat. Now, if ϕ(M ) = {Re(P (x, y)) = 0}, where P is a germ with an ILS at
L then Sing(M ) = L. In other words, M is a Levi-flat hypersurface with an ILS at
L. If P (x, y) is the germ A∞ , we prove that Theorem 1 is true in the case n = 2.
Theorem 2. Let M = {F = 0} be a germ of an irreducible real-analytic Levi-flat
hypersurface on (C3 , 0). Suppose that F is defined by
F (x, y) = Re(y12 + y22 ) + H(x, y),
where H is a germ of real-analytic function such that H(x, 0) = 0 and j0k (H) = 0
for k = 2. Then there exists a biholomorphism ϕ : (C3 , 0) → (C3 , 0) preserving L
such that ϕ(M ) = {Re(y12 + y22 ) = 0}.
 LEVI-FLAT HYPERSURFACES WITH AN ISOLATED LINE SINGULARITY 3

The above result should be compared to [2, Theorem 1.1]. This result can
be viewed as a Morse’s Lemma for Levi-flat hypersurfaces with an ILS at L. The
problem of normal forms of Levi-flat hypersurfaces in C3 with an ILS seems difficult
in the other cases. To prove these results we use techniques of holomorphic foliations
developed in [4] and [6]. Another normal forms of singular Levi-flat hypersurfaces
have been obtained in [2], [7] and [9].
This paper is organized as follows: In Section 2, we recall some definitions and
known results about Levi-flat and holomorphic foliations. Section 3 is devoted to
prove Theorem 1. In Section 4, we prove Theorem 2. Finally, in Section 5, using
holomorphic foliations, we give sufficient conditions for that a Levi-flat hypersurface
with a complex line as singularity to be a pullback of a real-analytic curve in C via
a holomorphic function, (see Theorem 5.7).

2. Levi-flat hypersurfaces and Foliations

In this section we works with germs at 0 ∈ Cn+1 of irreducible real-analytic
hypersurfaces and of codimension one holomorphic foliations. Let M = {F = 0},
where F : (Cn+1 , 0) → (R, 0) is a germ of an irreducible real-analytic function, and
M ∗ := {F = 0}\{dF = 0}. Let us define the singular set of M (or “set of critical
points” of M ) by
Sing(M ) := {F = 0} ∩ {dF = 0}. (2.1)
Note that Sing(M ) contains all points q ∈ M such that M is smooth at q, but the
codimension of M at q is at least two. In general the singular set of a real-analytic
subvariety M in a complex manifold is defined as the set of points near which M is
not a real-analytic submanifold (of any dimension) and “in general” has structure
of a semianalytic set; see for instance, [11]. In this paper, we work with Sing(M )
as defined in (2.1). We recall that (in this case) the Levi distribution L on M ∗ is
defined by
Lp := ker(∂F (p)) ⊂ Tp M ∗ = ker(dF (p)), for any p ∈ M ∗ . (2.2)
∗
Let us suppose that M is Levi-flat, this implies that M is foliated by complex
codimension one holomorphic submanifolds immersed on M ∗ .
Note that the Levi distribution L on M ∗ can be defined by the real-analytic
1-form η = i(∂F − ∂F¯ ), which is called the Levi 1-form of F . It is well known that
¯ ) ∧ ∂ ∂F
the integrability condition of L is equivalent to equation (∂F − ∂F ¯ |M ∗ = 0.
n+1
Let us consider the series Taylor of F at 0 ∈ C ,
X
F (x, y) = Fiµjν xi y µ x̄j ȳ ν
i,µ,j,ν

where F̄iµjν = Fjνiµ ; i, j ∈ N, µ = (µ1 , . . . , µn ), ν = (ν1 , . . . , νn ), (x, y) ∈ C × Cn ,

y µ = y1µ1 . . . ynµn and ȳ ν = ȳ1ν1 . . . ȳnνn . The complexification FC ∈ O2n+2 of F is
defined by the serie
X
FC (x, y, z, w) = Fiµjν xi y µ z j wν ,
i,µ,j,ν

where z ∈ C, w = (w1 , . . . , wn ) ∈ C and wν = w1ν1 . . . wnνn . Notice that F (x, y) =

n

FC (x, y, x̄, ȳ). The complexification MC of M is defined as MC := {FC = 0} and

defines a complex subvariety in C2n+2 , its regular part is MC∗ := MC \{dFC = 0}.
Now, assume that M is Levi-flat. Then the integrability condition of
¯ )|M ∗
η = i(∂F − ∂F
implies that ηC |MC∗ is integrable, where
ηC := i[(∂x FC + ∂y FC ) − (∂z FC + ∂w FC )].
4 ARTURO FERNÁNDEZ PÉREZ

Therefore ηC |MC∗ defines a codimension one holomorphic foliation LC on MC∗ that

will be called the complexification of L.
Let W := MC∗ \Sing(ηC |MC∗ ) and denote by Lζ the leaf of LC through ζ, where
ζ ∈ W . The next results will be used several times along of the paper.
Lemma 2.1 (Cerveau-Lins Neto [4]). For any ζ ∈ W , the leaf Lζ of LC through ζ
is closed in MC∗ .
Definition 2.2. The algebraic dimension of Sing(M ) is the complex dimension of
the singular set of MC .
The following result will be used enunciated in the context of Levi-flat hyper-
surfaces in Cn+1 .
Theorem 2.3 (Cerveau-Lins Neto [4]). Let M = {F = 0} be a germ of an ir-
reducible analytic Levi-flat hypersurface at 0 ∈ Cn+1 , n ≥ 2, with Levi 1-form
η = i(∂F − ∂F ¯ ). Assume that the algebraic dimension of Sing(M ) ≤ 2n − 2. Then
there exists a unique germ at 0 ∈ Cn+1 of holomorphic codimension one foliation
FM tangent to M , if one of the following conditions is fulfilled:
(1) n ≥ 3 and codMC∗ (Sing(ηC |MC∗ )) ≥ 3.
(2) n ≥ 2, codMC∗ (Sing(ηC |MC∗ )) ≥ 2 and LC admits a non-constant holomorphic
first integral.
Moreover, in both cases the foliation FM admits a non-constant holomorphic first
integral f such that M = {Re(f ) = 0}.

3. Proof of Theorem 1
We write
F (x, y) = Re(P (x, y1 , . . . , yn )) + H(x, y1 , . . . , yn ),
where P (x, y1 , . . . , yn ) is one of the polynomials of the Table 1, H : (Cn+1 , 0) →
(R, 0) is a germ of real-analytic function such that H(x, 0) = 0 for all x ∈ (C, 0)
and j0k (H) = 0, for k = deg(P ). The complexification of F is given by
1 1
FC (x, y, z, w) = P (x, y) + P (z, w) + HC (x, y, z, w),
2 2
thus MC = {FC (x, y, z, w) = 0} ⊂ (C2n+2 , 0), where z ∈ C and w = (w1 , . . . , wn ) ∈
Cn .
Since P (x, y) has an ILS at L, we get Sing(MC ) = {y = w = 0} ≃ C2 . In
particular, the algebraic dimension of Sing(M ) is 2. On the other hand, the com-
plexification of η = i(∂F − ∂F ¯ ) is
ηC := i[(∂x FC + ∂y FC ) − (∂z FC + ∂w FC )].
Recall that η|M ∗ and ηC |MC∗ define L and LC respectively. Now we compute
Sing(ηC |MC∗ ). We can write dFC = α + β, with
Xn n
∂FC ∂FC 1 ∂P 1 X ∂P
α := dx + dyj = (x, y)dx + (x, y)dyj + θ1
∂x j=1
∂yj 2 ∂x 2 j=1 ∂yj

and
Xn n
∂FC ∂FC 1 ∂P 1 X ∂P
β := dz + dwj = (z, w)dz + (z, w)dwj + θ2
∂z j=1
∂wj 2 ∂z 2 j=1 ∂wj
Pn ∂HC Pn ∂HC
where θ1 = ∂H
∂x dx +
C ∂HC
j=1 ∂zj dzj and θ2 = ∂z dz + j=1 ∂wj dwj .
Note that ηC = i(α − β), and so
ηC |MC∗ = (ηC + idFC )|MC∗ = 2iα|MC∗ = −2iβ|MC∗ . (3.1)
 LEVI-FLAT HYPERSURFACES WITH AN ISOLATED LINE SINGULARITY 5

In particular, α|MC∗ and β|MC∗ define LC . Therefore Sing(ηC |MC∗ ) can be split in two
parts. In fact, let M1 := {(x, y, z, w) ∈ MC | ∂F ∂FC
∂z 6= 0 or ∂wj 6= 0 for some j =
C

1, . . . , n} and M2 := {(x, y, z, w) ∈ MC | ∂F
∂x 6= 0 or
C ∂FC
∂zj 6= 0 for some j = 1, . . . , n},
∂HC ∂HC
then MC = M1 ∪ M2 . If we denote by A0 = ∂x , Aj = ∂zj for all 1 ≤ j ≤ n and
∂HC ∂HC
by B0 = ∂z , Bj = ∂wj for all 1 ≤ j ≤ n, we obtain that Sing(ηC |MC∗ ) = X1 ∪ X2 ,
where
∂P ∂P ∂P
X1 := M1 ∩ { (x, y) + A0 = (x, y) + A1 = . . . = (x, y) + An = 0}
∂x ∂y1 ∂yn
and
∂P ∂P ∂P
X2 := M2 ∩ { (z, w) + B0 = (z, w) + B1 = . . . = (z, w) + Bn = 0}.
∂z ∂w1 ∂wn
Since P is a polynomial with an ILS at L = {y = 0}, we conclude that
codMC∗ Sing(ηC |MC∗ ) = n.
By hypothesis n ≥ 3, then it follows from Theorem 2.3, part (1) that there exists
a germ f ∈ On+1 such that the holomorphic foliation F defined by df = 0 is tangent
to M . Moreover M = {Re(f ) = 0}. Note that if M = {Re(f ) = 0} = {F = 0},
with F an irreducible germ, we must have that Re(f ) = U · F , where U is a germ
of real-analytic function with U (0) 6= 0. Without loss of generality, we can assume
that U (0) = 1. In particular, Re(f ) = U · F implies that f = P + h.o.t. According
to Theorem 1.1, there exists a biholomorphism ϕ : (Cn+1 , 0) → (Cn+1 , 0) preserving
L such that f ◦ ϕ−1 = P , (f is DI -equivalent to P , because f is a germ with ILS
at L). Therefore, ϕ(M ) = {Re(P ) = 0} and the proof ends.

4. Proof of Theorem 2
The idea is to use Theorem 2.3, part (2). In order to prove our result in the case
n = 2, we are going to prove that LC has a non-constant holomorphic first integral.
We begin by a blow-up along C := {y1 = y2 = w1 = w2 = 0} ≃ C2 ⊂ C6 . Let
F (x, y1 , y2 ) = Re(y12 + y22 ) + H and M = {F = 0} Levi-flat. Its complexification
can be written as
1 1
FC (x, y1 , y2 , z, w1 , w2 ) = (y12 + y22 ) + (w12 + w22 ) + HC (x, y1 , y2 , z, w1 , w2 ).
2 2
Note that
Sing(MC ) = {y = w = 0} = C.
Let E be the exceptional divisor of the blow-up π : C̃6 → C6 along C. Denote by
M̃C := π −1 (MC \ {C}) ⊂ C̃6 the strict transform of MC via π and by F̃ := π ∗ (LC )
the foliation on M̃C .
Now, we consider an especial situation. Suppose that M̃C is smooth and set
C̃ := M̃C ∩ E. Moreover, assume that C̃ is invariant by F̃ . Take S = C̃ \ SingF̃ ,
P
then S is a smooth leaf of F̃ . Pick p0 ∈ S and a transverse section through p0 .
P P
Let G ⊂ Diff( , p0 ) be the holonomy
P group of the leaf S of F̃ . Since dim( ) = 1,
we can assume that G ⊂ Diff( , 0). We state a fundamental lemma.
Lemma 4.1 (Fernández-Pérez [9]). In the above situation, suppose that the follow-
ing properties are verified:
(1) For any p ∈ S\Sing(F̃ ) the leaf Lp of F̃ through p is closed in S.
(2) g ′ (0) is a primitive root of unity, for all g ∈ G\{id}.
Then LC admits a non-constant holomorphic first integral.
6 ARTURO FERNÁNDEZ PÉREZ

Proof. Let G′ = {g ′ (0)/g ∈ G} and consider the homomorphism φ : G → G′ defined

by φ(g) = g ′ (0). We claim that φ is injective. In fact, assume that φ(g) = 1 and
suppose by contradiction that g 6= id. In this case g(z) = z + az r+1 + . . ., where
a 6= 0. According to [12], the pseudo-orbits of this transformation accumulate at
P
0 ∈ ( , 0), contradicting the fact that the leaves of F̃ are closed and so the assertion
is proved. Now, it suffices to prove that any element g ∈ G has finite order (cf.
[13]). In fact, φ(g) = g ′ (0) is a root of unity thus g has finite order because φ is
injective. Hence, all transformations of G have finite orderP and G is linearizable.
This implies that there is a coordinate system w on ( , 0) such that G = hw →
λwi, where λ is a dth -primitive root of unity (cf. [13]). In particular, ψ(w) = wd is
a first integral of G, that is ψ ◦ g = ψ for any g ∈ G.
Let Γ be the union of the separatrices of LC through 0 ∈ C6 and Γ̃ be its
strict transform under π. The first integral ψ can be extended to a first integral
ϕ : M̃C \Γ̃ → C by setting X
ϕ(q) = ψ(L̃q ∩ ),
where L̃p denotes the
P leaf of F̃ through q. Since ψ is bounded (in a compact
neighborhood of 0 ∈ ), so is ϕ. It follows from Riemann extension theorem that
ϕ can be extended holomorphically to Γ̃ with ϕ(Γ̃) = 0. This provides the first
integral of LC . 
The rest of the proof is devoted to prove that we are indeed in the conditions
of Lemma 4.1. It is follows from Lemma 2.1 that the leaves of LC are closed.
Therefore, we need to prove that each generator of the holonomy group G of F̃
with respect to S has finite order.
Consider for instance the chart (U1 , (x, t, s, z, u, v)) of C̃6 where
π(x, t, s, z, u, v) = (x, tu, su, z, u, vu) = (x, y1 , y2 , z, w1 , w2 ).
We have
M̃C ∩ U1 = {(x, t, s, z, u, v) ∈ U1 |1 + t2 + s2 + v 2 + uH1 (x, t, s, z, u, v) = 0},
where H1 = H(x, ut, us, z, u, uv)/u3 and this fact imply that
E ∩ M̃C ∩ U1 = {(x, t, s, z, u, v) ∈ U1 |1 + t2 + s2 + v 2 = u = 0}.
It is not difficult to see that these complex subvarieties are smooth. Now, let us
describe the foliation F̃ on U1 . In fact, note that the foliation LC is defined by
α|MC∗ = 0, where
X2
1 ∂P 1 ∂P 1 ∂P ∂HC ∂HC
α= dx + dy1 + dy2 + dx + dyj .
2 ∂x 2 ∂y1 2 ∂y2 ∂x j=1
∂yj
P2 ∂HC
It follows that α = y1 dy1 + y2 dy2 + ∂H
∂x dx +
C
j=1 ∂yj dyj , then F̃ |U1 is defined by
α̃|M̃C ∩U1 = 0, where
α̃ = (t2 + s2 )du + utdt + usds + uθ̃, (4.1)
and P2
π ∗ ( ∂H
∂x dx +
C ∂HC
j=1 ∂yj dyj )
θ̃ = .
u2
Therefore, the singular set of F̃ |U1 is given by
SingF̃ |U1 = {u = t + is = 0} ∪ {u = t − is = 0}.
On the other hand, note that the exceptional divisor E is invariant by F̃ and the
intersection with SingFe is
SingF̃ |U1 ∩ E = {u = t + is = v 2 + 1 = 0} ∪ {u = t − is = v 2 + 1 = 0}.
 LEVI-FLAT HYPERSURFACES WITH AN ISOLATED LINE SINGULARITY 7

In particular, S := (E ∩ M̃C )\SingLeC is a leaf of Fe. We calculate the generators

of the holonomy group G of the leaf S. We work in the chart U1 , because of the
symmetry of the variables in the definition of the variety P M̃C .
Pick p0 = (0, 1, 0, 0, 0, 0) ∈ S ∩ U1 and a transversal = {(0, 1, 0, 0, λ, 0)|λ ∈ C}
parameterized by λ at p0 . We have that
SingF̃ |U1 ∩ E = {u = t + is = v 2 + 1 = 0} ∪ {u = t − is = v 2 + 1 = 0}.
For each j = 1, 2; let ρj be a 2td -primitive root of −1. The fundamental group
π1 (S, p0 ) can be written in terms of generators as
π1 (S, p0 ) = hγj , δj i1≤j≤2 ,
where for each j = 1, 2; γj are loops that turn around {u = t+is = v−ρj = 0} and δj
are loops that turns around {u = t−is = v −ρj = 0}. Therefore, G = hfj , gj i1≤j≤2 ,
where fj and gj correspond to [γj ] and [δj ], respectively. We get from (4.1) that
fj′ (0) = e−πi and gj′ (0) = e−πi for all 1 ≤ j ≤ 2. The proof of the theorem is
complete.

5. Levi-flat hypersurfaces with a complex line as singularity

In this section, we work with the system of coordinates (z1 , . . . , zn ) ∈ Cn . The
canonical local models examples of Levi-flat hypersurfaces M in C3 such that
Sing(M ) = L = {z1 = z2 = 0} are {Re(z12 + z22 ) = 0} and {z1 z̄2 − z̄1 z2 = 0}.
Recently, Burns and Gong [2] classified, up to local biholomorphism, all germs
of quadratic Levi-flat hypersurfaces. Namely, up to biholomorphism, there is only
five models:
Type Normal form Singular set
Q0,2k Re(z12 + z22 + . . . + zk2 ) Cn−k
2 2 2
Q1,1 z1 + 2z1 z̄1 + z1 empty
Qλ1,2 z12 + 2λz12 z̄1 + z12 Cn−1
2
Q2,2 (z1 + z̄1 )(z2 + z̄2 ) R × Cn−2
Q2,4 z1 z̄2 − z̄1 z2 Cn−2
Table 2. Levi-flat quadrics
We address the problem of provide conditions to characterize singular Levi-flat
hypersurfaces with a complex line as singularity. Using the classification due to
Burns and Gong [2], it is not hard to prove the following proposition.
Proposition 5.1. Suppose that M is a quadratic real-analytic Levi-flat hypersur-
face in Cn , n ≥ 3 such that Sing(M ) = {z1 = z2 = . . . = zn−1 = 0}. Then
(1) If n = 3, M is biholomorphically equivalent to Q0,2 or Q2,4 .
(2) If n ≥ 4, M is biholomorphically equivalent to Q0,2(n−1) .
Proof. To prove part (1), observe that only there are two models of M which admits
Sing(M ) = {z1 = z2 = 0} as singularity, Q0,2 or Q2,4 . Now to prove part (2), note
that if n ≥ 4, the real hypersurface {z1 z̄2 − z̄1 z2 = 0} has a complex subvariety of
dimension n − 2 as singularity. It is follows that M is biholomorphically equivalent
to Q0,2(n−1) . 
In order to obtain a characterization, we define the Segre varieties associated
to real-analytic hypersurfaces. Let M be a real-analytic hypersurface defined by
{F = 0}. Fix p ∈ M , the Segre variety associated to M at p is the complex variety
in (Cn , p) defined by
Qp := {z ∈ (Cn , p) : FC (z, p̄) = 0}. (5.1)
Now assume that M is Levi-flat and denote by Lp the leaf of L through p ∈ M ∗ .
We denote by Q′p the union of all branches of Qp which are contained in M . Observe
8 ARTURO FERNÁNDEZ PÉREZ

that Q′p could be the empty set when p ∈ Sing(M ). Otherwise, it is a complex
variety of pure dimension n − 1.
The following result is classical, we proved it here for completeness.
Proposition 5.2. In above situation, Lp is an irreducible component of (Qp , p)
and Q′p = Lp .
Proof. Since p ∈ M ∗ , E. Cartan’s theorem assures that there exists a holomorphic
coordinate system such that near of p, M is given by {Re(zn ) = 0} and p is the
origin. In this coordinates system the foliation L is defined by dzn |M ∗ = 0. In
particular, L0 = {zn = 0} and obviously {zn = 0} is a branch of Q0 . Furthermore,
L0 is the unique germ of complex variety of pure dimension n − 1 at 0 which is
contained in M . Hence Q′0 = L0 . 
Let p ∈ Sing(M ), we say that p is a Segre degenerate singularity if Qp has di-
mension n, that is, Qp = (Cn , p). Otherwise, we say that p is a Segre nondegenerate
singularity.
Suppose that M is defined by {F = 0} in a neighborhood of p, observe that p is
a degenerate singularity of M if z 7−→ FC (z, p̄) is identically zero.
Remark 5.3. If V is a germ of complex variety of dimension n − 1 contained in
M then for p ∈ V , we have (V, p) ⊂ (Qp , p). In particular, if there exists distinct
infinitely many complex varieties of dimension n − 1 through p ∈ M then p is a
Segre degenerate singularity.
To continuation, we consider a germ at 0 ∈ Cn of a codimension one singular
holomorphic foliation F .
Definition 5.4. We say that F and M are tangent, if the leaves of the Levi foliation
L on M are also leaves of F .
Definition 5.5. A meromorphic (holomorphic) function h is called a meromorphic
(holomorphic) first integral for F if its indeterminacy (zeros) set is contained in
Sing(F ) and its level hypersurfaces contain the leaves of F .
Recently, Cerveau and Lins Neto proved the following result.
Theorem 5.6 (Cerveau-Lins Neto [4]). Let F be a germ at 0 ∈ Cn , n ≥ 3, of
holomorphic codimension one foliation tangent to a germ of an irreducible real
analytic hypersurface M . Then F has a non-constant meromorphic first integral.
In our context, we prove the following result.
Theorem 5.7. Let M be a germ at 0 ∈ Cn , n ≥ 3 of an irreducible real-analytic
Levi-flat hypersurfaces such that Sing(M ) = L := {z1 = z2 = . . . = zn−1 = 0}.
Suppose that:
(1) Every point in Sing(M ) is a Segre nondegenerate singularity.
(2) The Levi-foliation L on M ∗ extends to a holomorphic foliation F in some
neighborhood of M .
Then there exists f ∈ On and a real-analytic curve γ ⊂ C such that M = f −1 (γ).
Proof. Since the Levi-foliation L on M ∗ extends to a holomorphic foliation F , we
can apply directly Theorem 5.6, this means that F has a non-constant meromorphic
first integral f = g/h, where g and h are relatively prime. We asserts that f is
holomorphic. In fact, if f is purely meromorphic, we have that for all ζ ∈ C, the
complex hypersurfaces Vζ = {g(z) − ζh(z) = 0} contains leaves of F . In particular,
M contains an infinitely many of hypersurfaces Vζ , because M is closed and F
is tangent to M . Set Λ := {ζ ∈ C : Vζ ⊂ M }. Note also that the foliation F
 LEVI-FLAT HYPERSURFACES WITH AN ISOLATED LINE SINGULARITY 9

is singular at L, so that If := {h = g = 0} the indeterminacy set of f intersect

L. Therefore, we have a point q at If ∩ L which would be a Segre degenerate
singularity, because q ∈ Vζ , for all ζ ∈ Λ. It is a contradiction and the assertion is
proved.
The foliation F is defined by df = 0, f ∈ On and is tangent to M . Without
loss generality, we can assume that f is an irreducible germ in On . According to a
remark of Brunella [3, pg. 8], there exists a real-analytic curve γ ⊂ C through the
origin such that M = f −1 (γ). 
Remark 5.8. In [11], J. Lebl gave conditions for the Levi-foliation on M ∗ does
extended to a holomorphic foliation. One could be considered these hypothesis and
establish a theorem more refined. Note also that if Sing(M ) is a germ of smooth
complex curve, it is possible adapted the proof of Theorem 5.7. In general, the
holomorphic extension problem for the Levi-foliation of a Levi-flat real-analytic
hypersurface remains open and is of independent interest, for more details see [8].

Acknowledgments.– This work was partially supported by PRPq - Universidade

Federal de Minas Gerias UFMG 2013 and FAPEMIG APQ-00371-13. I would like
to thank Maurı́cio Corrêa JR for his comments and suggestions, and the referee for
pointing out corrections.

References
[1] V.I. Arnold: Normal forms of functions near degenerate critical points, the Weyl groups
Ak , Dk , Ek and Lagrangian singularities. Funkcional. Anal. i Priložen. (6), no. 4, pp. 3-25,
(1972).
[2] D. Burns, X. Gong: Singular Levi-flat real analytic hypersurfaces. Amer. J. Math. 121, no.
1, pp. 23-53, (1999).
[3] M. Brunella: Some remarks on meromorphic first integrals. Enseign. Math. (2), 58 (3-4):
315-324, (2012).
[4] D. Cerveau, A. Lins Neto: Local Levi-Flat hypersurfaces invariants by a codimension one
holomorphic foliation. Amer. J. Math. 133, no. 3, pp. 677-716, (2011).
[5] A. H. Durfee: Fifteen characterizations of rational double points and simple critical points.
Enseign. Math. 25, pp. 131-163, (1979).
[6] A. Fernández-Pérez: Singular Levi-flat hypersurfaces. An approach through holomorphic
foliations. Ph. D. Thesis IMPA - Brazil (2010).
[7] A. Fernández-Pérez: On normal forms of singular Levi-flat real analytic hypersurfaces. Bull.
Braz. Math. Soc. 42 (1), pp. 75-85, (2011).
[8] A. Fernández-Pérez: On Levi-flat hypersurfaces with generic real singularities. J. Geom. Anal.
v. 23, pp. 2020-2033, (2013).
[9] A. Fernández-Pérez: Normal forms of Levi-flat hypersurfaces with Arnold type singularities.
Ann. Sc. Norm. Super. Pisa Cl. Sci. (2014, to appear) doi:10.2422/2036-2145.201112 003
[10] J. Lebl: Algebraic Levi-flat hypervarieties in complex projective space. J. Geom. Anal. 22
(2), 410-432, (2012).
[11] J. Lebl: Singular set of a Levi-flat hypersurface is Levi-flat. Math. Ann. 355. no. 3. pp.
1177-1199, (2013).
[12] F. Loray: Pseudo-groupe d’une singularité de feuilletage holomorphe en dimension deux.
Avaliable in http://hal.archives-ouvertures.fr/ccsd-00016434
[13] J.F. Mattei, R. Moussu: Holonomie et intégrales premières. Ann. Ec. Norm. Sup. 13, pp.
469-523, (1980).
[14] D. Siersma: Isolated line singularity. Proc. of Symposia in Pure Math. (2) 40, pp. 485-496,
(1983).
[15] A. Zaharia: Characterizations of simple isolated line singularities. Canad. Math. Bull. Vol.
42 (4), pp. 499-506, (1999).

Departamento de Matemática, Universidade Federal de Minas Gerais, UFMG

Current address: Av. Antônio Carlos, 6627 C.P. 702, 30123-970 - Belo Horizonte - MG, Brazil.
E-mail address: arturofp@mat.ufmg.br